Managing Reverse Logistics or Reversing Logistics

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/34742326
Managing reverse logistics or reversing logistics management? /
Thesis · January 2004

Source: OAI
CITATIONS READS
192 3,163
1 author:
Marisa P. de Brito
Breda University of Applied Sciences
68 PUBLICATIONS 2,446 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Humanitarian Logistics View project
Entrepreneurship in Leisure, Tourism and Hospitality View project
All content following this page was uploaded by Marisa P. de Brito on 11 January 2018.
The user has requested enhancement of the downloaded file.

M A R I S A P. D E B R I T O
Managing Reverse
Logistics or Reversing
Logistics Management?
Managing Reverse Logistics
or
Reversing Logistics Management?
Marisa P. de Brito
ERIM PhD Series Research in Management, 35
Erasmus Research Institute of Management (ERIM)
Erasmus University Rotterdam
Internet: http://www.erim.eur.nl
Cover photo: ‘Reverse’

http://images.google.nl
originally from
pantransit.reptiles.org/images/nsorted/photo/animal/
ISBN 90-5892-058-6

c 2003, Marisa P. de Brito
All rights reserved. No part of this publication may be reproduced
or transmitted in any form or by any means, electronic or material,
including photocopying, recording, or by any information storage
and retrieval system, without permission from the author.
Managing Reverse Logistics
or
Reversing Logistics Management?
Beheersing van retourlogistiek

of
omgekeerde beheersing van logistiek?
Thesis
to obtain the degree of Doctor from

the Erasmus University Rotterdam
by command of the
Rector Magnificus
Prof.dr. S.W.J. Lamberts
and according to the decision of the Doctorate Board
The public defense shall be held on

Thursday 12 February, 2004 at 16.00 hrs.
by
Marisa Paula de Brito Pereira Maduro
born at Luanda, Angola
Promoters: Prof.dr.ir. R. Dekker
Prof.dr. M. B. M. de Koster
Other members: Prof.dr.ir. J.A.E.E. van Nunen

Prof.dr.ir. L.N. van Wassenhove
Prof.dr. T. Spengler
Acknowledgements
Finishing a PhD dissertation is certainly a moment to cheer. My PhD tra-

jectory was full of many other happy moments because along the way I had
the privilege to work with highly skilled colleagues, to enjoy the friendship
of many and the love of the ones most close to me. I will not mention you
all, neither I will acknowledge you enough, but I can assure you that my true
acknowledgements are ampler than these few pages.
First of all, I would like to thank my promoters. I want to thank Rommert
Dekker for giving me enough freedom in carrying out my research and for
welcoming my initiatives. I thank René de Koster for all his enthusiasm and
for making me feel we really worked as a team.
I also would like to acknowledge the contribution of the co-authors of my
papers, Beril Toktay, Dan Guide, Erwin van der Laan, Luk van Wassenhove
and Simme Douwe Flapper: I have learned much with all of you! Others
have also directly contributed to my research in various forms, as follows:
Dr. Krever provided me with the CERN data and Lehel Nagy helped me to
get the data from the database; Eric Porras Musalem assisted me in organiz-
ing the data of the refinery; Julien Mostard was a helpful contact with the
mail-order-company; Marcin Paweska gave me a hand with information about
European legislation and Ruud Teunter and Pelin Bayindir co-monitored the
NGT study.
During the past years of research I was affiliated with the Econometric
Institute (EI) from Rotterdam School of Economics, and I belonged as a
PhD candidate to the Erasmus Research Institute of Management (ERIM).
Both Institutes were very supportive during my PhD track, either at the
professional or at the social level (leuke uitjes, hoor! ).
My research project was also very much embedded in the European Work-
ing Group for Reverse Logistics (RevLog), which was financially supported by
the European Commission. I want to thank all RevLoggers for sharing with
me their expertise, for the more and less formal conversations, and for the
great time we had in every single workshop (I will never forget the Pedipaper
in Costa da Caparica!) .
Belonging to RevLog gave me the opportunity to visit INSEAD and to
work on a research project with Luk van Wassenhove, who I would like to
thank for his prompt hospitality. During my time there I could experience
the professionalism of the staff and the warmness of the PhD students. My
warm thanks go especially to Evrim.
During my PhD’s ups and downs, I always had a special source of enjoy-
ment and support: “the girls” and “the boys.” Anna, Daina, Fernanda, Gabi,
6
Helga, Julia, Laura, Lenny, Viara, ... : cheers for all the lunches, dinners,
cocktails, parties, laughs, for the serious and less-serious conversations, and
especially for all the non-sense that made us wiser. Dennis, Julien, Eric, Ger-
ard, Guillermo, Manolo, Paolo, Ruud, Willem, Wybe, ... : urra for all the
barbecues, picnics, movies, sportive events, and so forth.
I mainly met those mentioned above during my PhD trajectory. Yet, I
would like to mention, what I believe was the conception of a potential PhD in
the Netherlands: my exchange period as an undergraduate student in the aca-
demic year 1996/97, at the Econometric Institute of the Erasmus University
Rotterdam. Then, I wrote my thesis on inventory control with Hans Frenk
and closely watched by Marcel Kleijn. The proof of the success is that this is
one of the areas I deal with in this PhD dissertation. I would like to thank once
again the Department of Statistics and Operations Research (DEIO), Faculty
of Sciences, University of Lisbon, for facilitating the exchange. In particular, I
would like to thank the support and encouragement of my Professors Joaquim
Gromicho, Ivette Gomes, Antónia Turkman and António Rodrigues.
I also want to acknowledge the precious financial support of the Portuguese
Foundation for the Development of Science and Technology, Fundação Para
a Ciência e a Tecnologia (FCT).
I want to express my gratitude to my family, above all my parents: por
tudo, obrigada! In particular, I want to mention what I have learned from
my close family, which has proved most relevant for my research work. Thus:
Pai, obrigada for impregnating me with the “what-if” state of mind, which
makes so natural the act of thinking beyond current reality. Mãe, obrigada for
your multi-functional way of carrying life, (a precious inherited talent, which
helped me a lot in writing this thesis!). Obrigada mano Zé for influencing me
on “always knowing exactly how things are/ work/ etc,” which started in my
early youth with looking up words in the dictionary and their whereabouts in
the encyclopedia. Obrigada mano Rui for teaching me to give my max while
working in a team because, after all, every single thing is about people.
Finally, I would like to acknowledge who was much supportive during my
PhD as a friend, as a partner, as a colleague, and even as a co-author: Er-
win. We had so many moments related with my PhD that I deeply cherish.
To mention only a few: going to conferences together, participating in group
discussions, making presentations together and most especially our disagree-
ments about the Quans vs. the Quals in Science. Even the latter were very
much enjoyable!
For all of you: ’brigada, dank, dziekuje, gracias, grazie, kiitt, ksznet, tank,
tesekkr !
Contents
1 Introduction 17
1.1 Reverse Logistics: definition and scope . . . . . . . . . . . . . . 19
1.1.1 Definition and a brief history . . . . . . . . . . . . . . . 19
1.1.2 Delineation and scope . . . . . . . . . . . . . . . . . . . 20
1.1.3 The European dimension . . . . . . . . . . . . . . . . . 22
1.2 Literature on Reverse Logistics . . . . . . . . . . . . . . . . . . 23
1.2.1 Books on reverse logistics . . . . . . . . . . . . . . . . . 24
1.2.2 PhD theses on reverse logistics . . . . . . . . . . . . . . 25
1.2.3 Review articles on reverse logistics . . . . . . . . . . . . 26
1.2.4 Some literature on areas related to reverse logistics . . . 27
1.3 About the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.3.1 The structure . . . . . . . . . . . . . . . . . . . . . . . . 28
1.3.2 The process: aims, objectives and methods . . . . . . . 29
1.3.3 Research paradigms and this thesis . . . . . . . . . . . . 36
2 A Framework for Reverse Logistics 39

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.3 Reverse Logistics: a review of literature on theoretical devel-
opments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.4 Reverse Logistics: the basic dimensions . . . . . . . . . . . . . 45
2.5 Reverse Logistics: why? what? how? who? . . . . . . . . . . . 48
2.5.1 Why-returning: drivers . . . . . . . . . . . . . . . . . . 48
2.5.2 Why-returning: return reasons . . . . . . . . . . . . . . 53
2.5.3 What: types and characteristics . . . . . . . . . . . . . . 55
2.5.4 How: processes and recovery options . . . . . . . . . . . 59
2.5.5 Who: actors and roles . . . . . . . . . . . . . . . . . . . 65
2.6 The framework . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.7 Summary and conclusions . . . . . . . . . . . . . . . . . . . . . 77
7
8 Contents
3 Reverse Logistics: a review of case studies 79

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.3 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
3.4 Case studies on Reverse Logistics Networks . . . . . . . . . . . 84
3.4.1 The case studies . . . . . . . . . . . . . . . . . . . . . . 84
3.4.2 Preliminary insights . . . . . . . . . . . . . . . . . . . . 88
3.5 Case studies on Reverse Logistics Relationships . . . . . . . . . 89
3.5.1 The case studies . . . . . . . . . . . . . . . . . . . . . . 89
3.5.2 Preliminary Insights . . . . . . . . . . . . . . . . . . . . 92
3.6 Case studies on Inventory Management . . . . . . . . . . . . . 95
3.6.1 The case studies . . . . . . . . . . . . . . . . . . . . . . 95
3.7 Case studies on Planning and Control
of recovery activities . . . . . . . . . . . . . . . . . . . . . . . . 99
3.7.1 The case studies . . . . . . . . . . . . . . . . . . . . . . 100
3.8 Case studies on IT for Reverse Logistics . . . . . . . . . . . . . 105
3.8.1 The case studies . . . . . . . . . . . . . . . . . . . . . . 105
3.9 Overall insights . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.10 Summary and research opportunities . . . . . . . . . . . . . . . 115
3.10.1 Research opportunities . . . . . . . . . . . . . . . . . . . 119
4 Return Handling: decision-making and quantitative support123

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.2 Return Handling . . . . . . . . . . . . . . . . . . . . . . . . . . 124
4.3 Return Handling: an illustrative case . . . . . . . . . . . . . . . 125
4.4 Return Handling: a framework of decisions . . . . . . . . . . . 127
4.4.1 Long-term decisions . . . . . . . . . . . . . . . . . . . . 127
4.4.2 Medium-term decisions . . . . . . . . . . . . . . . . . . 130
4.4.3 Short-term decisions . . . . . . . . . . . . . . . . . . . . 133
4.5 Models and research opportunities . . . . . . . . . . . . . . . . 133
4.5.1 Long-term decisions . . . . . . . . . . . . . . . . . . . . 134
4.5.2 Medium-term decisions . . . . . . . . . . . . . . . . . . 135
4.5.3 Short-term decisions . . . . . . . . . . . . . . . . . . . . 141
Contents 9
5 Return Handling: an exploratory study 147

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
5.2 Handling returns . . . . . . . . . . . . . . . . . . . . . . . . . . 149
5.3 The case studies: brief description and design . . . . . . . . . . 151
5.4 The case studies: analysis and comparison . . . . . . . . . . . . 153
5.4.1 Contextual factors . . . . . . . . . . . . . . . . . . . . . 153
5.4.2 The process . . . . . . . . . . . . . . . . . . . . . . . . . 157
5.4.3 Manpower, space and associated costs . . . . . . . . . . 161
5.5 Discussion and implications . . . . . . . . . . . . . . . . . . . . 164
5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
5.6.1 Recommendations for practice . . . . . . . . . . . . . . 172
5.6.2 Research opportunities . . . . . . . . . . . . . . . . . . . 174
6 Inventory Management: a critical analysis of assumptions 177

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
6.2 Inventory models with returns:
a literature review . . . . . . . . . . . . . . . . . . . . . . . . . 178
6.2.1 Deterministic models . . . . . . . . . . . . . . . . . . . . 179
6.2.2 Stochastic models . . . . . . . . . . . . . . . . . . . . . 179
6.3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
6.3.1 Testing HCPP . . . . . . . . . . . . . . . . . . . . . . . 184
6.3.2 Testing exponentiality of the time to return . . . . . . . 184
6.3.3 Statistical tests: more details . . . . . . . . . . . . . . . 185
6.4 The data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
6.5 Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
6.6.1 Theoretical insights . . . . . . . . . . . . . . . . . . . . 200
6.6.2 Managerial implications . . . . . . . . . . . . . . . . . . 201
6.6.3 Research needs . . . . . . . . . . . . . . . . . . . . . . . 202
7 Inventory Management with product returns: the value of

information 205
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
7.2 (Mis)information in practice . . . . . . . . . . . . . . . . . . . . 206
7.3 Literature review . . . . . . . . . . . . . . . . . . . . . . . . . . 208
7.4 Forecasting methods . . . . . . . . . . . . . . . . . . . . . . . . 209
7.5 Forecasting performance . . . . . . . . . . . . . . . . . . . . . . 214
7.5.1 Analysis regarding the expectation of lead time
net demand . . . . . . . . . . . . . . . . . . . . . . . . . 215
7.5.2 Analysis regarding the variance of lead time net demand 216
10 Contents
7.6 Cost performance . . . . . . . . . . . . . . . . . . . . . . . . . . 219

7.6.1 Experimental design . . . . . . . . . . . . . . . . . . . . 219
7.6.2 Numerical study . . . . . . . . . . . . . . . . . . . . . . 220
7.7 Discussion and managerial implications . . . . . . . . . . . . . 228
7.8 Summary of conclusions and further research . . . . . . . . . . 230
8 Reverse Logistics research: what does the future bring? 245

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
8.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
8.2.1 An exploratory study . . . . . . . . . . . . . . . . . . . 247
8.2.2 The Delphi study . . . . . . . . . . . . . . . . . . . . . . 247
8.3 An exploratory study . . . . . . . . . . . . . . . . . . . . . . . . 248
8.3.1 Implementation and discussion . . . . . . . . . . . . . . 248
8.3.2 Observations and input for the Delphi Study . . . . . . 251
8.4 Design and implementation of the Delphi . . . . . . . . . . . . 252
8.5 Reverse Logistics researchers and the panel . . . . . . . . . . . 253
8.6 Reverse Logistics: past, present and future . . . . . . . . . . . . 254
8.6.1 Research . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
8.6.2 Trends and breakthroughs . . . . . . . . . . . . . . . . . 257
8.6.3 Pedagogy . . . . . . . . . . . . . . . . . . . . . . . . . . 259
8.7 The senior, the junior and the accidental . . . . . . . . . . . . . 261
8.8 European vs. North American researchers . . . . . . . . . . . . 262
8.9 A glimpse on the rest of the world . . . . . . . . . . . . . . . . 266
9 Concluding remarks 271

9.2 Raising questions and further research . . . . . . . . . . . . . . 282
Bibliography 287
Samenvatting (Summary in Dutch) 313
Curriculum Vitae 319

List of Figures
1.1 The body of the thesis . . . . . . . . . . . . . . . . . . . . . . . 30
2.1 Building the framework . . . . . . . . . . . . . . . . . . . . . . 42

2.2 The five basic dimensions of reverse logistics. . . . . . . . . . . 47
2.3 Driving triangle for reverse logistics. . . . . . . . . . . . . . . . 49
2.4 The typology ‘drivers’ for the dimension ‘why-receiving’. . . . . 52
2.5 The typology ‘return reasons’ for the dimension ‘why-returning’. 56
2.6 The typologies ‘product characteristics’ and ‘product types’ for
the dimension ‘what’. . . . . . . . . . . . . . . . . . . . . . . . 60
2.7 The typologies ‘processes’ and ‘recovery options’ for the dimen-
sion ‘how’. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
2.8 Recovery option inverted pyramid . . . . . . . . . . . . . . . . 64
2.9 Reverse logistics processes . . . . . . . . . . . . . . . . . . . . . 65
2.10 The typologies ‘actors’ and ‘roles’ for the dimension ‘who’. . . . 66
2.11 Who is who in reverse logistics . . . . . . . . . . . . . . . . . . 67
2.12 A framework for reverse logistics. . . . . . . . . . . . . . . . . . 69
2.13 Illustrative example of by-product returns. . . . . . . . . . . . . 70
2.14 Illustrative example of functional returns. . . . . . . . . . . . . 71
2.15 Illustrative example of reimbursement returns. . . . . . . . . . 72
2.16 Illustrative example of service returns. . . . . . . . . . . . . . . 73
2.17 Illustrative example of end-of-use returns. . . . . . . . . . . . . 74
2.18 Illustrative example of end-of-life returns. . . . . . . . . . . . . 75
3.1 IT support for Reverse Logistics in different phases of the life-

cycle of the product. . . . . . . . . . . . . . . . . . . . . . . . . 110
4.1 Return handling in the warehouse. . . . . . . . . . . . . . . . . 124
5.1 Reverse flows in the supply chain. . . . . . . . . . . . . . . . . . 148

5.2 The process of handling returns. . . . . . . . . . . . . . . . . . 150
5.3 The process of handling returns, with and without returns. . . 163
11
12 List of Figures
5.4 Decision-scheme on general long- and medium-term warehous-

ing decisions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
5.5 Decision-making scheme on combining vs. separating forward
and reverse flows during transport, receipt and storage. . . . . 174
6.1 Statistical analysis scheme to test if a process is an HCPP. . . . 185

6.2 Stabilizer gas: date of purchase vs. purchased quantity. . . . . 192
6.3 Large agenda: date of purchase vs. purchased quantity. . . . . 192
6.4 Large agenda: dot-plot of the purchased quantity. . . . . . . . . 195
6.5 Large agenda: histogram of the purchased quantity. . . . . . . 195
6.6 Group 2 of products: four probability plots. . . . . . . . . . . . 196
6.7 MOC Fashion 2000: histogram of the return lag. . . . . . . . . 197
6.8 RF: histogram of the return lag. . . . . . . . . . . . . . . . . . 198
7.1 Relation between data and decision-making. . . . . . . . . . . . 207

7.2 Method A (it uses the overall return probability to compute
the expected returns coming from demand during the lead time.211
7.3 Method B (it uses past demand and return probabilities p0j s to
compute returns coming from past demand and it employees
expected values for returns coming from demand during the
lead time). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
7.4 Method C (it uses information on past demand and past re-
turns to compute the returns coming from past demand; it
uses information on current demand to compute returns com-
ing from period t; and it uses expected value to compute returns
coming from demand during the lead time). . . . . . . . . . . . 212
7.5 Method D (it uses information on past demand and observed
returns from each past demand to compute the returns coming
from past demand; it uses information on current demand to
compute returns coming from period t; and it uses expected
value to compute returns coming from demand during the lead
time). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
7.6 Preference regions for F{V } in terms of a and Ri ; πi = βi = 0.6,
p = 0.7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
7.7 Preference regions in terms of a and Ri for various values of πi
and p. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
7.8 Misinformation (-30%,+30%) of the expected time-to-return
(Geometric time-to-return distribution, p = 0.8, q = 0.6, µD =
30, cvD = 0.2, L = 4, h = 1, b = 50). . . . . . . . . . . . . . . . . 225
List of Figures 13
7.9 Performance of Method A both under perfect correlation and

zero correlation (Geometric time-to-return distribution, p =
0.5, q = 0.6, µD = 30, cvD = 0.2, L = 4, h = 1, b = 50). . . . . . . 228
8.1 Representation of the sheet of paper given to the participants. . 250

8.2 Past research contributions: relative ratings; 100% corresponds
to the highest sum of ratings . . . . . . . . . . . . . . . . . . . 263
8.3 Current research limitations: relative ratings; 100% corresponds
to the highest sum of ratings. . . . . . . . . . . . . . . . . . . . 263
8.4 Future research opportunities: relative ratings; 100% corre-
sponds to the highest sum of ratings. . . . . . . . . . . . . . . . 265
8.5 Achieving relevance for Reverse Logistics. . . . . . . . . . . . . 269
List of Tables
1.1 Recycling quotas in Europe (data collected in 2003 with * being

a prognosis). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.2 List of aims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.3 Motivation, objectives, methods and aims (by project) . . . . . 32
1.4 Basic beliefs of four main research paradigms. . . . . . . . . . . 37
3.1 Key words used for the search of case studies . . . . . . . . . . 81

3.2 United Nations classifications for Industry (see http://esa.un.org) 83
3.3 United Nations classifications for Product (see http://esa.un.org) 83
3.4 Case Studies on Reverse Logistics Networks . . . . . . . . . . . 85
3.5 Case Studies on Relationships . . . . . . . . . . . . . . . . . . . 90
3.6 Case studies: return-reason vs. incentive . . . . . . . . . . . . . 94
3.7 Case Studies on Inventory Management . . . . . . . . . . . . . 96
3.8 Case Studies on Production, Planning and Control . . . . . . . 101
3.9 Case Studies on IT . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.10 Case studies: IT tools, requirements and benefits for reverse
logistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
3.11 The cases: return reasons vs. recovery options . . . . . . . . . . 112
3.12 The cases: drivers vs. return reasons . . . . . . . . . . . . . . . 113
3.13 The kind of elements to be considered in more detail . . . . . . 115
4.1 A framework of decisions for return handling. . . . . . . . . . . 129

4.2 Research opportunities for long-term decisions . . . . . . . . . . 135
4.3 Research opportunities for medium-term decisions. . . . . . . . 140
4.4 Research opportunities for short-term decisions . . . . . . . . . 143
5.1 Outgoing and return flows (in roll cages, per week) at the food
retailers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
5.2 Outgoing and return flows (in roll cages, per year) at SC1, SC2
and SC3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
5.3 Return type and policy at nine retailers. . . . . . . . . . . . . . 158
14
List of Tables 15
5.4 The return process at the nine retailers. . . . . . . . . . . . . . 160

5.5 Manpower (in FTE) and space (m2 ) at the retailers with stores.161
5.6 Simple estimation of the costs for mail order companies, with
and without returns in euros (see Figure 5.3). . . . . . . . . . . 163
5.7 Material flow per manpower (in FTE): forward vs. return flow. 166
5.8 Material flow per area (in m2 ): forward vs. reverse flow. . . . . 167
5.9 Decision on whether to combine (C) vs. to separate (S) the
forward and the reverse flows. . . . . . . . . . . . . . . . . . . . 168
5.10 To combine (C) vs. to separate (S) the forward and the reverse
flows: critical factors and implications. . . . . . . . . . . . . . . 171
6.1 The assumptions of the literature on inventory models with

returns. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
6.2 Statistical analysis of two products of CERN’s database. . . . . 191
6.3 Outcome of the investigation whether, or not, time until return
is negative exponentially distributed. . . . . . . . . . . . . . . . 194
7.1 Notation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210

7.2 Comparison of Methods A–D in the case of perfect information;
Geometric time-to-return distribution; µD = 30. ∗ ) results
below 1% are not significant. . . . . . . . . . . . . . . . . . . . 221
7.3 Comparison of Methods A–D in the case of perfect information;
Uniform time-to-return distribution; µD = 30. ∗ ) results below
1% are not significant. . . . . . . . . . . . . . . . . . . . . . . . 222
7.4 Comparison of Methods B–D in the case of misestimation of
the return probability; Geometric time-to-return distribution;
µD = 30. ∗ ) results below 1% are not significant. . . . . . . . . 223
7.5 Comparison of Methods B–D in the case of misestimation of the
return probability; Uniform time-to-return distribution; µD =
30. ∗ ) results below 1% are not significant. . . . . . . . . . . . . 224
7.6 Comparison of Methods B–D in the case of misestimation of
the expected time-to-return; Geometric time-to-return distri-
bution; µD = 30. ∗ ) results below 1% are not significant. . . . . 226
7.7 Comparison of Methods B–D in the case of misestimation of the
expected time-to-return; Uniform time-to-return distribution;
µD = 30. ∗ ) results below 1% are not significant. . . . . . . . . 227
8.1 Employed arguments during NGT. . . . . . . . . . . . . . . . . 250

8.2 Topics covered by the Delphi questionnaire. . . . . . . . . . . . 252
8.3 Reverse Logistics: describing three kinds of researchers. . . . . 254
8.4 Reverse Logistics research: past, present and future. . . . . . . 255
16 List of Tables
8.5 Top research trends and breakthroughs with high impact and
likelihood. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
8.6 Top pedagogic challenges and utilities with high priority. . . . . 259
8.7 Realism, rigor and links: average ratings. . . . . . . . . . . . . 261
8.8 Top research trends: North America vs. Europe. . . . . . . . . 264
8.9 Top research breakthroughs with high likelihood: North Amer-
ica vs. Europe. . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
Chapter 1
Introduction
“A long trip starts with the first step”

Lao Tsé
Twenty years ago, supply chains were diligently fine-tuning the logistics from
raw materials to the end customer. Today an increasing flow of products is
going back in the chain.
Thus, companies have to manage reverse logistics as well. Yet, logis-
tic managers typically think “forward,” putting the emphasis on moving the
goods to the clients. The issue is, whether or not, companies can go on, and
manage reverse logistics, with a purely forward-based thinking. Or, perhaps,
there is a need for a considerable share in focus with respect to reverse flows.
The question is then: is it a matter of simply managing reverse logistics or of
reversing logistics management?
Reverse logistics practices are in the position of being an asset rather
than a liability. When a consumer gets rid of a product, this does not mean
that the product is valueless. For instance, computers might contain precious
metals such as gold, palladium, platinum, and silver, and these materials are
still intact when the computer reaches obsolescence. Actually, one metric ton
of electronic scrap from personal computers (PC’s) has more gold than that
extracted from a 17 ton of gold ore. Taking into account that millions of PC’s
become obsolete every year, it is not surprising that many consider obsolete
computers a “Gold Mine” (USGS, 2001). Global Investment Recovery, Inc. is
one of the many companies providing services in this area. Global Investment
Recovery, Inc. was funded in 1992 and it processes millions of circuit boards
every year (see Global Investment Recovery, 2003).
17
18 Chapter 1. Introduction
Many other examples can be given, as the role of reverse logistics is in-
creasing in the whole range of industries, covering electronic goods, pharma-
ceutical products, beverages and so on. For more than a decade, Kodak sells
remanufactured single-use photo cameras and ReCellular has a mobile phone
remanufacturing business; all sorts of companies handle many of their prod-
ucts in reusable packaging, like Coca-cola (refillable bottles), Philip Morris
(pallets), and Estée Lauder’s Aveda (merchandize displays), see Kodak, 2003;
Recellular, 2003; Coca-cola, 2003; Andriesse, 1999.
Companies are also investing on information technology dedicated to re-
verse logistics, like Estée Lauder’s, which recovered the investment in one
year. Other companies, like L’Oreal push high quantities of products to the
retailers, because what is not sold can always be brought back (see Meyer,
1999; Coenen, 2000).
Reverse logistics is therefore a key competence in modern supply chains.
Accordingly, the importance of reverse logistics is widely recognized. Reverse
logistics has renowned professional organizations, like the Reverse Logistics
Executive Council (RLEC) in the U.S., which collaborates with academia.
In Europe, the European Commission has shown, for long, interest in the
development of the field by sponsoring international scientific projects on
reverse logistics, such as RevLog and Reloop (see Revlog, 2003; ReLoop,
2003).
This thesis is from the beginning rooted in one of these pan-European
projects, namely RevLog, the European Working Group on Reverse Logistics.
This thesis contributes to a better understanding of reverse logistics. We
bring insights to reverse logistics decision-making and to the field of reverse
logistics as a whole. With respect to the initial question, is it a matter of
simply managing reverse logistics or of reversing logistics management?, we
encourage the reader to bear it in mind along the whole thesis. We will come
back to it in the last chapter.
In the next section we put forward the definition and scope of reverse
logistics. Then, we review some of the literature on the topic, including key
monographs, PhD theses and reviews. After that, we present the structure
of the thesis, and we put forward the aims, objectives and the methods em-
ployed.
1.1. Reverse Logistics: definition and scope 19
1.1 Reverse Logistics: definition and scope

1.1.1 Definition and a brief history
“In the sweat of your face you shall eat bread

Till you return to the ground,
For out of it you were taken;
For dust you are,
And to dust you shall return.”
Genesis 3:19
Though the idea of reverse logistics dates from long ago (as the citation
above attests), the naming is difficult to trace with exactness. Though sys-
tematically related with recycling, terms like Reverse Channels or Reverse
Flow already emerge in scientific literature of the seventies (Guiltinan and
Nwokoye, 1974; Ginter and Starling, 1978).
During the eighties, the definition was inspired by the movement of flows
against the traditional flows in the supply chain, or as put by Lambert and
Stock, 1981 “going the wrong way” (see also Murphy, 1986 and Murphy and
Poist, 1989).
In the early nineties, a formal definition of Reverse Logistics was put to-
gether by the Council of Logistics Management, stressing the recovery aspects
of reverse logistics (Stock, 1992):
“...the term often used to refer to the role of logistics in recycling, waste
disposal, and management of hazardous materials; a broader perspective in-
cludes all issues relating to logistics activities to be carried out in source re-
duction, recycling, substitution, reuse of materials and disposal.”
The previous definition is rather broad, as it is manifest in the following
excerpts “the role of logistics in ... all relating ... activities.” (see also Kopicky
et al., 1998). Besides this, it has origins in a waste management standpoint.
In the same year Pohlen and Farris (1992) define Reverse Logistics, giving
again notice of a direction in a distribution channel, as follows:
“...the movement of goods from a consumer towards a producer in a chan-
nel of distribution.”
Carter and Ellram (1998) keep the concept linked to environmental pur-
poses, as follows: “the process whereby companies can become environmen-
tally efficient through recycling, reusing, and reducing the amount of materials
used.”
At the end of the nineties, Rogers and Tibben-Lembke (1999) portrayed

Reverse Logistics by stressing the goal and the intrinsic (logistics) processes:
“The process of planning, implementing, and controlling the efficient, cost-
effective flow of raw materials, in-process inventory, finished goods, and re-
lated information from the point of consumption to the point of origin for the
purpose of recapturing value or proper disposal.”
The European Working Group on Reverse Logistics, (see De Brito and
Dekker, 2004), puts forward the following definition, which is used in this
thesis:
“The process of planning, implementing and controlling backward flows
of raw materials, in-process inventory, packaging and finished goods, from a
manufacturing, distribution or use point, to a point of recovery or point of
proper disposal”.
This perspective on Reverse Logistics keeps the essence of the definition as
put forward by Rogers and Tibben-Lembke (1999), which is logistics. However
it generalizes “point of consumption” to “a manufacturing, distribution or
use point” and “point of origin” to “a point of recovery or point of proper
disposal.” In this way we give margin to return flows that were not consumed
first (for instance, stock adjustments due to overstocks or spare parts which
were not used). We employ the expression “point of recovery” instead of
“point of origin” since flows may go back to other points of recovery than
the original (e.g. collected computer chips may enter another chain). Besides
this, “point of recovery” stresses the distinction we want to make between
reverse logistics and pure waste management activities (see next section).
Furthermore, we include the reverse direction through the term “backward
flows,” to exclude what can be considered as forward recovery. For instance,
when a consumer gives his/her personal computer to the neighbor.
In summary, the definition of Reverse Logistics has changed over time,
starting with a sense of reverse direction, going through an overemphasis on
environmental aspects, coming back to the original pillars of the concept,
and finally widening its scope. For other discussions on the evolution of the
definition of reverse logistics, we refer to Rogers and Tibben-Lembke (2001)
and to Fernandéz (2003).
1.1.2 Delineation and scope

Since Reverse Logistics is a relatively new research and empirical area, the
reader may encounter other terms in the literature, like reversed logistics,
1.1. Reverse Logistics: definition and scope 21
return logistics and retro logistics or reverse distribution, sometimes referring

to roughly the same. In fact, the diversity of definitions with respect to
recovery practices is a well-recognized source of misunderstandings both in
research as in practice (Melissen and De Ron, 1999).
It is important to observe that Reverse Logistics is different from waste
management as the latter mainly refers to collecting and processing products
or materials that are to be discarded. The crux in this matter is the definition
of waste. This is a major issue, as the term has severe legal consequences,
e.g. with respect to regulations on import/export of waste. Reverse Logistics
concentrates on those streams where there is some value to be recovered and
the outcome enters a (new) supply chain.
Reverse Logistics also differs from Green Logistics as the latter considers
environmental aspects to all logistics activities and it has been focused specif-
ically on forward logistic, i.e. from producer to customer (see Rodrigue et
al., 2001). The prominent environmental issues in logistics are consumption
of nonrenewable natural resources, air emissions, congestion and road usage,
noise pollution, and both hazardous and non-hazardous waste disposal (see
Camm, 2001).
Industrial Ecology is another field that relates to Reverse Logistics. As
explained in Garner and Keoleian (1995), Industrial Ecology is primarily ded-
icated to the study of the interactions between industrial systems and the
environment. The underlying aim is to change the linearity of industrial sys-
tems, from raw materials to waste, into a cyclical system, i.e. with product
or materials recovery. In the latter, Reverse Logistics undoubtedly plays a
major role.
Finally, reverse logistics can be seen as part of Sustainable Development.
The latter has been defined by Brundland (1998) in a report to the European
Union as “to meet the needs of the present without compromising the ability
of future generations to meet their own needs.” In fact one could regard
Reverse Logistics as the implementation at the company level of processes
that guarantee that society uses and re-uses both efficiently and effectively
all the value which has been put into products.
The border between Forward Logistics, from raw materials to end user,
and Reverse Logistics, from e.g. the end user to recovery or to a new user, is
not strictly defined. One can wonder about for instance what “raw materials”
are in certain supply chains. For instance, used/recovered glass constitutes
considerable material input for new production of glass.
A holistic view on supply chains combining both forward and reverse logis-
tics is embraced by the concept of Closed-Loop Supply Chains (Guide and van
Wassenhove, 2003). Recovery practices are framed within the supply chain,
and the encircling aspect of the process as a whole is therefore stressed: hav-
ing either 1) a physical closed-loop: to the original user (see Fleischmann et
al., 1997); or 2) a functional closed-loop: to the original functionality; or 3)
an open-loop, when neither the original user or original functionality are in
the reverse logistics process. The denomination closed-loop supply chains em-
phasizes the importance of coordinating the forward with the reverse streams.
Reverse Logistics is different but relates with other fields of research such
as green logistics and industrial ecology, as discussed above. In the academic
world, these fields inhabit separate islands, but they lay in the same sea. The
knowledge in each field has been growing quite independently of the knowledge
in related fields, but commonalities persist due to the similarities of the overall
environment. We believe that we only have to gain by establishing structural
links among the fields, even though we have to cross a period of confusion as
indicated in the previous section.
1.1.3 The European dimension

In Europe, the European Union has stimulated sustainable development for
years, including recovery practices through environmental legislation. Recent
developments show that this legislation and its consequences are ever increas-
ing. Priority product flows include: cars (Directive 00/53/EC), electronics
(Directive 02/96/EC) and packaging (Directive 99/31/EC). Table 1.1 summa-
rizes the state-of-the-art of recycling quotas (required, realized or expected)
in some of the EU countries (see Jordan et al., 2001; European Environmental
Agency, 2002; Auto Recycling Nederland, 2003; Arge Altauto, 2003; Doppelt
and Nelson, 2001; Europa, 2003).
Though the Member States are also encouraged to introduce preven-
tion and reuse systems, they have some more freedom on this than on the
mandatory recycling quotas (Greece, Ireland and Portugal are allowed to
set lower targets). Concerning packaging waste, for instance, The Nether-
lands, Spain and Belgium launched prevention targets (some in the form of
covenants), while Denmark, Germany and Portugal introduced reuse targets.
Joint recycling-reuse targets are in force in Austria and Finland (the latter
also has explicit prevention targets).
The link between reverse logistics activities and employment has not yet
been thoroughly investigated. We do know that some of the recovery activ-
ities like recycling are labor intensive. Studies have indicated that recycling
activities create 5 to 7 times the number of jobs than incineration and more
than 10 times than land filling operations. (See Austrian Federal Chamber of
1.2. Literature on Reverse Logistics 23
Table 1.1: Recycling quotas in Europe (data collected in 2003 with * being a
prognosis).
EU Directive France Germany The Netherlands

(required quotas) (realized quotas)
Cars
85% (2006)* - 85% (2002) 85% (2002)
95% (2015)* - 95% (2007)*
Electronics
70%-90% (2006)* - - 45%-75% (2002)
Packaging
50% (2001) 45% (1998) 65% (1998) 60% (1998)
60% (2006)*
70% (2011)*
Labor, 1998a and 1998b; Desproges and David, 1998). The number of peo-
ple working in recovery activities is estimated to exceed 3 million (see Laxe,
1999). Besides this, the European Commission realizes that recovery activi-
ties offer an opportunity to create new distinct jobs, which are not going to
be a copy of already existing employment (see Europa, 2003). Furthermore,
these jobs are a privileged vehicle for reintegration of workers in professional
life. The number of social jobs (state financed) in recycling and reuse has
been estimated to surpass 35 000 in Western Europe (see ACRR, 2003).
In Europe, there are initiatives like the Association of Cities and Regions
for Recycling (ACRR). Yet, it is lacking a pan-European professional organi-
zation dedicated to Reverse Logistics. In contrast, in the U.S., there is the
Reverse Logistics Executive Council (RLEC). In addition, other organiza-
tions like The American Production and Inventory Control Society (APICS),
and The Council of Logistics Management (CLM) are very active on reverse
logistics (see as well Section 1.2.1).
Concluding, in Europe reverse logistics related activities play a threefold
role for sustainable development: environmentally, economically and socially.
1.2 Literature on Reverse Logistics

This section provides a review of some literature on reverse logistics. We
consider key monographs that are important references for the study of reverse
logistics, recent PhD theses on the topic, and reviews. Furthermore and to
give an idea of opportunities for multidisciplinary reverse logistics studies, we

present a snap overview of research in related areas.
1.2.1 Books on reverse logistics

In the nineties, the Council of Logistics Management (CLM) published three
relevant studies on reverse logistics. The first one, actually one of the first
studies making a thorough exam of reverse logistics, introduced the subject
from a waste reduction perspective (Stock, 1992). Based on interviews with
US industry and governmental organizations, the author concluded that re-
verse logistics was still at its very beginning. Accordingly, the companies were
rather reactive in dealing with it. Shortly after this first study, CLM pub-
lished the second study, now dedicated to the opportunities offered by reverse
logistics in the context of reuse and recycling (Kopicky et al., 1993). The
authors concluded that reverse logistics was evolving very fast with many
companies having pioneer programs on waste reduction. After introducing
the field, and having showed the opportunities, CLM brought forward a third
study dedicated to the implementation and development of reverse logistics
(Stock et al., 1998). This study gave form to relevant issues for successful
reverse logistics programs, like management and control, measurement and
finance.
Kostecki (1998) took a marketing approach to reflect on the optimal use
of consumer durable products. The approach linked consumers’ preferences
and the business challenge of reuse. The book reported on several business
examples of product stewardship programmes such as the ones of Kodak and
Xerox.
Also in the late nineties, Rogers and Tibben-Lembke (1999) conducted
interviews in the U.S. with 150 managers and a survey with hundreds of
others. The study focused on the economic and supply chain issues of reverse
logistics and how to reduce costs related with product returns.
Three more books are likely to join this year the collection of key-references
for reverse logistics research: Guide and van Wassenhove edit a book about
Closed-Loop Supply Chains, and the European Working Group on Reverse
Logistics, RevLog, brings out two more books, one on quantitative methods
and the other on cases from practice (Dekker et al., 2003; Flapper et al.,
2003). Next, we describe in more detail the contents of these three books.
Guide and van Wassenhove (2003) closely consider the business aspects
of closed-loop supply chains. The book handles subjects such as the recov-
ery options and the relations between parties, namely questions related with
contracting. Particular issues include logistics, production planning and in-
ventory control in the supply chain. Other embraced aspects are forecasting,
information technology, technology diffusion and product design, where the
fulcrum is the customer. From a more business perspective, the economics
of closed-loop supply chains and the issue of acquisition management are dis-
cussed.
One of the RevLog books is dedicated to quantitative modelling as for
aiding reverse logistics decision-making (Dekker et al., 2003). Models are
brought together, compared with traditional literature, related to each other
and linked to real life examples. The models cover collection, distribution,
inventory control, production planning and supply chain management issues.
The book embraces particular subjects such as forecasting of returns, network
design, vehicle routing, return handling, lot-sizing decisions, co-ordination,
environmental and information management. The other RevLog book (Flap-
per et al., 2003) is on managing closed-loop supply chains and it is a collection
of case studies with the respective lessons for managing reverse logistics prac-
tices.
1.2.2 PhD theses on reverse logistics

There have been a number of PhD theses on the field of reverse logistics.
Next, we describe a sample of those.
Thierry (1995) studied the impact of product recovery management in the
electronics and car industries. Special attention was given to the impact on
product design, logistics, and relations between actors. The author made use
of two real life cases: a copy remanufacturing company and product recovery
by the car manufacturer BMW.
Jahre (1995) investigated the performance of collection and recycling sys-
tems of household waste, with specific emphasizes on packaging materials.
Two main aspects were taken into account: the degree of separation at the
source and co-collection. The author employed data provided by the Euro-
pean Recovery and Recycling Association (ERRA).
Van der Laan (1997) analyzed the effects of remanufacturing on inven-
tory control. Special attention went to coordination mechanisms between the
manufacturing and remanufacturing operations and to which actions deal to
cost-efficiency. The findings were tested with real data from Volkswagen on
car parts remanufacturing.
Krikke (1998) addressed the determination of recovery strategies and the
logistics network design. For each of the issues, a case study was discussed in
detail. On the recovery strategies, for the recycling of computer monitors, the
author considerd Roteb, the municipal waste company of the city of Rotter-
dam, in the Netherlands. On logistics network design, the author dealt with
the copier reverse logistics network of the international company Océ, with
headquarters in Venlo, the Netherlands.
Fleischmann (2000) dealt with quantitative models for reverse logistics
network design and for inventory management with returns. In particular,
the thesis inquired under which conditions a network can be split into two
separate networks, the forward and the reverse. In addition, the practical
issues arising from reverse logistics were illustrated with a case study at IBM.
Beullens (2001) focused on the use of Operations Research (OR) tools for
supporting the facility location decision, process planning and vehicle routing
in reverse logistics. All the issues were illustrated with real life examples.
Kobeissi (2001) considered the evaluation of options for recovery of end-
of-life products. The thesis described the required resources and the different
activities of the evaluation process. Also the dissertation of Landrieu (2001) is
about end-of-life products. The focus was on the collection strategies, taking
into account the product, or the geographic zone of collection. The measures
of performance were maximization of profit and the satisfaction of the client.
Besides this, particular informational issues were considered as follows: how
to integrate information in the collection schemes and what is the impact on
efficiency.
Brodin (2002) analyzed the influence of both the product and the relation-
ship between actors on the efficiency of logistics systems for recycling. To do
so, the author made use of 1) a case study of the electronics recycling in the
Netherlands; 2) a survey of the Swedish recycling market; and 3) interviews
with Ericsson, Telia (large customer of telecommunications equipment), Stena
Technoworld (recycling company) and Tekpak (a US package producer).
As expected, each of the reviewed PhD dissertation dealt with a set of par-
ticular issues within reverse logistics. A diversity of subjects can be found such
as recycling systems, inventory control of remanufacturing, recovery strate-
gies, network design and vehicle routing. Furthermore, most of the theses
made a link with practice. The theses included studies on the electronics and
communication industries (IBM, Ericsson), automobile industry (BMW and
Volkswagen), and packaging industry (Tekpak), among others.
1.2.3 Review articles on reverse logistics

Fleischmann et al. (1997) compiled a review on Operations Research (OR)
models for reverse logistics. The review is organized in terms of distribution,
inventory control and production planning models. The authors pointed out
the need for comprehensive approaches, taking into account both the economic
and the ecological aspects in the supply chain.

Gungor and Gupta (1999) presented an exhaustive review of more than
300 articles or books on environmentally conscious manufacturing and prod-
uct recovery. The authors observe that environmental issues are becoming
popular among society, governments and industry. In addition, they identify
the need to develop both qualitative and quantitative tools to support proper
environmental conscious manufacturing decision-making.
Carter and Ellram (1998) put together a review of transportation and
packaging, purchasing and other literature on reverse logistics. The authors
deduce from the literature that there were both internal and external factors
affecting the behavior of organizations with respect to reverse logistics. Next,
they build up a model with a combination of determinants for reverse logistics,
and based on it they indicate opportunities for future investigation.
Bras and McIntosh (1999) reviewed the literature specifically dedicated
to remanufacturing. The authors distinguish between descriptive and devel-
opmental work, i.e. work simply meant to characterize the current state and
the future of remanufacturing and work aiming at improving remanufacturing
aspects. The authors conclude that the work up to then intended: to describe
practice; to motivate remanufacturing; to improve product and design for re-
manufacturing; to learn from traditional manufacturing; and, to improve the
process.
De Brito et al. (2003) put forward a review of case studies on reverse
logistics, covering various aspects like network design, inventory control and
information technology. This review and its findings are central in Chapter 3
of this thesis.
1.2.4 Some literature on areas related to reverse logistics

Here we present a snap overview of other research work that, though it is
related with reverse logistics, it is focused mainly on other research area.
This gives an idea of the existing opportunities for multidisciplinary reverse
logistics studies.
Van Nes (2003) focus on product design as a mean of influencing product
lifetime. The author first checks whether, or not, it is desirable to extend the
life cycle of products. For most of the products investigated an increase in
lifetime is desirable. The next step in the research is directed to provide an
understanding of the motivation of consumers to substitute their products.
Finally, the author provides insights on how to extend the product lifetime
by changing product design.
Beukering (2001) investigates the economic and the environmental impor-

tance of recycling for international trade mechanisms. The author discusses
the trade of waste paper for recycling in India, the recycling and trade in
waste plastics in China and the trade and recycling of truck tyres in Europe.
Hirsch et al. (1998) show how a simulation tool, called LOCOMOTIVE
can aid on reverse logistics decision making. The tool is able to simulate
the whole life cycle and to assess environmental impacts of different logis-
tics operations within a globally distributed production network. The tool
is meant to support middle and upper management, assisting on strategic
decision-making.
Stroeker (1995) gives the marketing and economics perspective on second-
hand consumer durable goods from. Demographic and lifestyle consumer
characteristics are taken into account as context variables as the price struc-
ture and the bargaining process in second-hand markets were studied. Fur-
thermore, the author presents a model to predict the purchase of second-hand
durable goods.
This section on work related to reverse logistics accommodates studies
on research areas such as Design (for end-of-life), Trade (with respect to
recycling), Computer Science (simulation tools to support decision-making),
and Economics and Marketing (analysis of second-hand markets).
1.3 About the thesis

1.3.1 The structure
Besides the existing areas “affiliated” to reverse logistics, the previous review
of literature shows that there are many subjects of research inherent to reverse
logistics. The review embraced studies on recovery systems; studies dedicated
to a particular form of recovery like recycling, or to a specific decision-area
like network design, inventory management or vehicle routing; another set
reviewed work was on structuring the field, or on pulling lessons from real-life
examples.
The set of topics we could elect to carry out research is very vast. We
opted to become engaged with a number of projects, divided in 4 blocks, as
follows (see Figure 1.1):
• Part I on the theoretical development of the field;
• Part II on decision-making and return handling;
• Part III on decision-making and inventory management;

1.3. About the thesis 29
• Part IV on the future development of the field.
In the reverse logistics field, return handling is an emerging cluster of

knowledge, while inventory management is a well-established cluster of knowl-
edge. Both offer many possibilities for research. As the first gives the oppor-
tunity to explore fresh issues, the second poses the challenge of being able
to contribute to on-going discussions. This explains our choices of topics for
blocks II and III.
Next, we explain more about how this structure was born, which are
the aims, what are the particular objectives and which was the employed
methodology to achieve them.
1.3.2 The process: aims, objectives and methods

With this doctoral thesis, we committed to make a contribution according to
the following two main lines:
• 1) reverse logistics as a research/academic field;
• 2) reverse logistics decision-making and practice.
Next, we present the correspondent aims.
Aims
To put together the aims we employ two sorts of input: on the one hand,
researchers with expertise on reverse logistics; and on the other hand, reverse
logistics (related) literature. Both the contact with experienced researchers
and a scan of the literature lead to the following conclusions:
• reverse logistics is a young field, growing fast, but with scattered and
fragmented theoretical contributions (see e.g. Melissen and Ron, 1999);
• besides this, there is a strong need to better understand reverse logistics

decision-making processes and to learn from them;
Aligned with our first main line, reverse logistics as a research/academic

field, we establish the following aims: to structure reverse logistics as a re-
search field. Furthermore, since reverse logistics is a young field of research, we
also plan to look ahead, and to conjecture about the future development of the
field. With respect to the second main line, reverse logistics decision-making
Chapter 1. Introduction
Figure 1.1: The body of the thesis

30
Table 1.2: List of aims

Guideline 1:
Reverse logistics as a research/academic field
i) structuring reverse logistics as a research field;

ii) conjecturing about the future development of reverse logistics as a research field;
Guideline 2:
Reverse logistics decision-making and practice
iii) better understanding of reverse logistics practices;

iv) structuring and supporting reverse logistics decision-making;
and practice, the following aims are put together: to better understand re-
verse logistics practices and to structure and support decision-making. Table
1.2 summarizes this.
This exploratory phase of setting the aims is a mean to elaborate the
geographic map of our “research expedition”. To be more specific on the trail
we will walk on, we elaborate specific objectives, which are described next.
Objectives and employed methods

In this design phase, we intensified the contacts with other researchers, as
well as the analysis of existing literature. Above all, researchers associated
with the RevLog group were of invaluable help (see RevLog, 2003). With this
input, we could bring focus to the overall aims, and therefore we are able of
specifying the objectives.
Table 1.3 summarizes the collection of research projects we carry out,
together with the motivation, the specific objectives, the employed methods,
and the key-aims. We refer to the employed methods only briefly, as more
details are given in each chapter. The projects are not presented in a strict
chronological order.
PART I- The Field
Chapter 2. A framework for Reverse Logistics

While going through the literature, it is clear that there is no formal frame-
work for the field as a whole. The existing contributions have been proposed
intentionally for sub-areas of the field, or they were informally presented as a
pretext to explain a specific topic. This is not surprising, as reverse logistics
is a young field of research. Despite of this fact, the field is growing fast
Chapter 1. Introduction
Table 1.3: Motivation, objectives, methods and aims (by project)

Motivation Main research objectives Correspondent aims Main methods
Part I - The Field
Chapter 2. A framework for Reverse Logistics
Structuring the field To structure the field and to provide an i) Review of literature, case studies
overall understanding of reverse logistics review & accumulated knowledge
Chapter 3. Reverse Logistics: a review of case studies
Need to better understand reverse To provide a diversity of reverse logistics iii) Case studies review & content
logistics practices activities (and to help constructing the analysis
framework)
Part II- Decision Making and Return Handling
Chapter 4. Return Handling: models for decision-making
Emerging cluster of knowledge To put forward the state-of-the-art of mod- iv) Critical literature review
Lack of decision-making models els to support return handling
Chapter 5. Return Handling: an exploratory study
Emerging cluster of knowledge Re- To identify the main factors contributing iii) & iv) (Comparative) case study research
search opportunity to whether or not to combine forward and
reverse flows during return handling
Part III- Decision Making and Inventory Management
Chapter 6. Inventory Management: a critical analysis of assumptions
Lack of empirical evidence to sup- To inquiry the validity of the common as- iii) & iv) Literature review and statistical
port common assumptions sumptions in the inventory management analysis
literature with product returns
Chapter 7. Inventory Management with product returns: the value of information
Literature assumes that more in- To assess the impact of (mis)information iv) Mathematical analysis & simula-
formation leads to better forecasts on inventory management tion
Part IV- The Future of the Field
Chapter 8. Reverse Logistics research: what does the future bring?
Reverse logistics is developing fast. To assess which future do academicians ex- ii) Nominal Group Technique & Del-
A look ahead pect for reverse logistics as a research field phi study
32
demanding to be formally structured. With this research project, we make

a close re-examination of the field and we propose a framework, providing a
formal structure and with it an overall understanding about reverse logistics
management and decision-making.
To build-up the framework we use three basic inputs: 1) a review of the
literature dealing with the organization of the field throughout classifications,
2) an interactive process with a review of case studies as a tool to ground
the development of the framework in empirical evidence, and 3) generated
knowledge.
By the latter we refer to all the knowledge acquired during this doctoral
trajectory. The whole process of conducting other research projects provided
“data” to build the framework. This “data” takes the form of accumulated
knowledge, and it results from collaborating with senior researchers, from
attending conferences and from carrying out several other different projects
about reverse logistics. Another source of knowledge is non-scientific litera-
ture, as well as the unavoidable awareness of real-life reverse logistics situa-
tions.
Chapter 2 gives insights on the diversity of reverse logistics systems and on
its complexity by putting together a framework for reverse logistics. Reverse
logistics is characterized via four basic aspects: why companies get involved
with reverse logistics and why products go back in the supply chain; how are
products recovered and which are the main recovery options; what is actually
being returned; and finally, who are the main players in a reverse logistics
system. More on the methodology behind this research project, can be found
in a section of Chapter 2. Another version of this chapter has appeared in De
Brito and Dekker (2003).
Please bear in mind that the research projects are not presented in chrono-
logical order.
The motivation for this research project is in essence the need to better un-
derstand reverse logistics practices. One finds however quite some literature
on case studies applied to a diversity of industries. This literature is however
scattered over journals for very different research communities. Besides this,
in countries in the forefront of reverse logistics a substantial number of case
studies have been published in other languages than the lingua franca and are
therefore not accessible for the majority of the research community. To give
a fair idea of the diversity of reverse logistics practices, we employ a content
analysis of more than sixty real reverse logistics cases. The analysis is based
on the framework of Chapter 2. Thus, Chapter 3 is also a life test of the
applicability of the framework, and of the completeness of the classifications

proposed by the framework. Chapter 3 presents the case studies according
the following decision-making focus: Network Structures, Relationships, In-
ventory Management, and Planning and Control. We remark particular issues
and unanswered questions, which we link with a research agenda. For some
of these decision making groups, some of the aforementioned questions are
predominantly important. We also discuss that and we provide research di-
rections to test these premisses. A previous version of Chapter 3 can be found
in De Brito, et al. (2003).
PART II- Decision-making and return handling
Chapter 4. Return Handling: models for decision-making

Return handling is an emerging cluster of knowledge. Since this cluster is
in the looming phase, there are not many models supporting decision-making
in this area. This motivates an investigation on the opportunities that forward
quantitative models for material handling offer for return handling models.
To do so, we put together a decision-making framework for return handling
and we proceed to a critical analysis of the literature. The chapter reviews
the main findings of general material handling literature giving suggestions
to adapt such models for return handling. An early version has appeared in
De Brito and De Koster (2004).
Chapter 5. Return Handling: an exploratory study
The details of return handling operations have been so far overlooked in
the scientific literature. More specifically, there is a need to better understand
decision-making regarding the transport, the actual handling and the storage
of product returns. In particular, insights are missing on the factors influ-
encing the decision of combining, or not, forward and reverse flows during
these processes. Chapter 5 focuses on this by comparing the return handling
processes of nine retailer warehouses. The analysis gives insights into critical
factors, complicating issues, possible simplifying solutions and practical im-
plications. The chapter ends with a discussion of propositions that can feed
future studies on return handling efficiency. This research project was pub-
lished in the International Journal of Retail and Distribution Management
(De Koster et al., 2002).
PART III- Decision-making and inventory management
Chapter 6. Inventory Management: a critical analysis of assumptions

In contrast with return handling, inventory management with product
returns is, for some time, a well-established cluster of knowledge. Much lit-
erature is available on the subject. An analysis of this literature shows that
certain assumptions are very often used. At the same time, the trustworthi-
ness of these assumptions has been repeatedly challenged due to the lack of
empirical evidence. With this research project we inquire on the validity of
the common assumptions, by proceeding to a statistical analysis of real data.
Chapter 6 checks empirically the validity of common assumptions in the
literature dealing with inventory control with product returns. This chapter
first proposes a general methodology to check the assumptions empirically
and then it describes actual practice in companies with respect to informa-
tion storage on returns and inventory control. Furthermore, the methodology
is employed to real data from three companies and the practical implications
are discussed. The companies are the European Organization for Nuclear Re-
search (CERN) in Switzerland, and a mail order company and a refinery, both
in the Netherlands. The findings have practical relevance, for instance with
respect to information management on inventory systems with returns. The
findings of this research were earlier published in the International Journal of
Production Economics (De Brito and Dekker, 2003).
Chapter 7. Inventory Management and product returns: the value of in-
formation
Chapter 7 evaluates the impact of misinformation for inventory systems
with product returns. If one could exactly know how much is going to be re-
turned and when, one would certainly benefit from incorporating this perfect
information a priori in the management of production, inventory, and distri-
bution. In practice, one has to attempt to forecast the timing and the amount
of product returns, by hypothesizing about the return flow properties. To do
so, historic data on demand and returns can be used. The available literature
on information and inventory management with product returns commonly
1) assumes known return probabilities; or 2) considers specific cases where
the most-informed method leads to the best forecast. Chapter 7 identifies
situations in which the most informed method does not necessarily lead to
the best performance, investigating the impact on inventory related costs. See
also related publications: De Brito and van der Laan (2003) and Toktay et
al. (2004).
PART IV- The future of the field
Chapter 8. Reverse logistics research: what does the future bring?

Since reverse logistics is a rapidly progressing field, with this research
project we look ahead and are able to conjecture on the future development
of reverse logistics. This project is developed in two phases. The first and ex-
ploratory phase employs the Nominal Group Technique. The second is based
on a Delphi study with an international panel of academics working on re-
verse logistics issues. Some of the outcomes are recommendations concerning
research and teaching about reverse logistics. In this way, this study assists
decision-making concerning proposals, and research priorities with respect to
reverse logistics. Furthermore, the chapter reports on the contrasts and/or
similarities between the European and North American researchers’ perspec-
tives. The exploratory study is also reported in De Brito (2003) and the
second part is in De Brito and van Wassenhove (2003).
Finally Chapter 9 summarizes the main findings, discusses the limitations

and proposes guidelines for future research on reverse logistics.
1.3.3 Research paradigms and this thesis

No research is conducted without a set of underlying beliefs, held more or less
consciously by the one carrying it out. Besides this, others will also judge his
or her research based on a set of own beliefs.
Table 1.4 summarizes the basic beliefs of the four main streams of paradigms,
which have been traditionally identified in the literature (for more details see
Guba and Lincoln, 1994; Tashakkori and Teddlie, 1998; Healy and Perry,
2000).
Instead of strictly following one paradigm, this research takes a pragmatic
approach in the sense that the research question is primordial in each research
project, and the method is chosen accordingly. In this context, pragmatism
is an attractive paradigm because it welcomes a priory any methodology,
including mixed methodologies. Besides this, it stays clear of interminable
and ineffectual discussions on metaphysical concepts like “truth” or “reality.”
Furthermore, pragmatism has also been proposed as a departing reference for
decision-making in management (see Fontrodona, 2002). Finally it exempts
the reader from an in-depth knowledge on extensive criteria to assess the
quality of the research. This is because the criterion for assessing the quality
of pragmatic inquire relies on asking whether, or not, the employed methods
fulfill the objective of answering the research question.
Table 1.4: Basic beliefs of four main research paradigms.
Paradigm Main
/Element Ontology Epistemology Methodology
Positivism Single reality and Researcher and re- Hypotheses testing
apprehendable ality are indepen-
dent (findings are
true)
Critical Realism Single reality but Partial objectivism Falsification of hy-
imperfectly appre- (findings are prob- potheses
hendable ably true)
Critical theory Reality is shaped Value-ladenness Dialogical/ dialec-
(e.g. by cultural (findings depend tical (researcher
values) and be- of researchers’ changes reality)
comes more clear values)
over time
Constructivism There are multiple researcher and re- Hermeneutical/
and constructed reality are indivisible dialectical (re-
alities (created findings) searcher is an
active participant
in the world being
investigated)
Chapter 2
A Framework for Reverse

Logistics
“We adore chaos because we love to produce order.”

M. C. Escher
2.1 Introduction
The objective of this chapter is to provide an overall understanding of reverse
logistics by structuring the field. In fact, we put together a framework for
reverse logistics. A framework is “a basic conceptional structure” (see the
Merrian Encyclopedia, 2003). Ergo, here we put forward a basic conceptional
structure for reverse logistics, i.e. we identify the elementary ingredients
(dimensions) of reverse logistics, we structure them, and we describe their
relation to each other. With the framework we provide a basic understanding
about reverse logistics situations, since we
1. identify the basic dimensions of reverse logistics,

2. classify different types for each dimension (we provide typologies),
3. explain the relations between the basic dimensions of reverse logistics,
and
4. illustrate how the characterization of the basic dimensions helps us to
identify the problematic issues and the associated decision-making areas
of a given reverse logistics situation.
39
40 Chapter 2. A Framework for Reverse Logistics
In spite of many genuine efforts, up to now the contributions to organize

the field are scattered and fragmented. This is not a surprise, since reverse
logistics is a young field of research. Despite of this, the field is growing
fast, demanding thus to be formally structured. This is because one cannot
provide substantial insights into reverse logistics without first being familiar
with its basics elements. Neither a theory on critical factors for successful
reverse logistics can be properly developed. Thus, the framework is valuable
as
1. a means to provide a holistic view on reverse logistics situations,
2. a tool to characterize the basic features of reverse logistics situations,
3. an instrument to understand the demands of different reverse logistics

situations with respect to the type of decisions that are crucial, and as
4. a necessary platform to provide further insights in the field, possibly

including the development of theory.
To build the framework we employ three sources of input: selective litera-

ture that brings structure to the field, a review of case studies (Chapter 3
of this thesis), and the knowledge on reverse logistics that was accumulated
throughout the whole PhD trajectory.
To extract data from the inputs, we proceed to a content analysis of the
aforementioned literature, an interactive process with the case study review
(which was carried out simultaneously), and a tacit intellectual record of the
basic dimensions of reverse logistics. The next section describes the method-
ology in more detail.
Summarizing, the main contributions of this chapter are as follows.
1. We put forward a framework for reverse logistics as a whole.
2. We build the framework with an underlying methodology.
3. We ground the typologies of the framework in empirical evidence (case

studies).
The remainder of the chapter is organized as follows. First, we clarify what

the methodology behind the project is (Section 2.2). Section 2.3 reviews
specific literature that contributes to the organization of the field. Next,
Section 2.4 puts together the skeleton of the framework by identifying the
basic dimensions of reverse logistics. Section 2.5 provides the “mass” for
2.2. Methodology 41
the skeleton by presenting the typologies for each of the basic dimensions.
Section 2.6 articulates the framework by describing the relations between the
basic dimensions and showing the research opportunities. Finally, Section 2.7
provides a summary and conclusions.
2.2 Methodology
The overall approach of this research project is very much inline with the
principles of Grounded Theory. As described in Strauss and Corbin (2000):
“Grounded theory is a general methodology for developing theory that is
grounded in data systematically gathered and analyzed.” Furthermore, the
authors explain that it is “a way of thinking about and conceptualizing data”
and an attempt to generate theory that is grounded in the data. We would
like to stress that it is a way of thinking about data, as well as a way of
conceptualizing data.
Since Grounded Theory grounds findings on data, it is thus an appropriate
methodology to build-up a framework, which is a means to understand the
field. To put together the framework, we make use of the following three
sources of data (see also Figure 2.1).
1. We employ existing literature,
- to re-examine the field,

- to identify the basic dimensions of reverse logistics, and
- to have a start-up for the typologies (when possible).
2. We employ the case study review in an interactive way, mainly as
- a means to empirically ground the typologies,

- a means to improve the completeness and adequateness of the ty-
pologies, and
- a means to refine the naming of the elements of each of the typolo-
gies.
3. We employ the knowledge accumulated during the whole PhD trajectory

to strengthen the research with another source of data (triangulation).
The first procedure is rather explicit. In Section 2.3 we explain how the
literature leads to the skeleton of the framework, i.e. the basic dimensions.
Figure 2.1: Building the framework
While putting forward the typologies (Section 2.5) we state what the con-
tribution is of the classifications in the existing literature. By typology we
mean a classification based on types or categories (see Merrian Encyclopedia,
2003).
Let us go into more details regarding the second and the third sets of
data. Interactive process refers to the progressive interaction and growth of
i ) the framework, and ii ) the review of the case studies. The analysis of the
case studies is based on the framework. Let us recall that the two research
projects were carried out simultaneously. The case studies are a tool to put
the adequateness of the typologies and its completeness to the test. This
allows the framework to be re-shaped and improved, when necessary. Also,
the interactive process provides a refinement in the naming of the elements
of the typologies. The naming of the elements originally was sometimes un-
suitable when faced with a specific case study. By carrying out the projects
simultaneously, we could ground the framework on empirical evidence.
Regarding the tacit record of basic dimensions, the whole doctoral tra-
jectory plays a role. It is the result of i ) cooperation with seniors and
young researchers, ii ) attending conferences, iii ) carrying out different re-
search projects on reverse logistics, and iv ) getting information from profes-
2.3. Reverse Logistics: a review of literature on theoretical developments 43
sional journals on the field, newspapers and internet sites. During the whole
PhD-trajectory there was awareness about this being the third source of data
for building the framework. It is important to note that this awareness is
actually the crux regarding this third source of data, so at all the moments
data could be gathered for the project. In addition, notice that a plurality of
approaches is employed, known as triangulation.
Triangulation is further achieved by carrying out research projects in two
different areas of the field: on the one end, research on an emerging cluster
of knowledge, and, on the other end, research in a well-established cluster of
knowledge. The contents of these projects can be found respectively in Part
II (Return Handling) and Part III (Inventory Management) of this thesis.
Furthermore, the criteria to choose the structure and to build-up the
framework can be summarized as follows (see Tashakkori and Teddlie, 1998;
Fleischman and Mumford, 1991).
- The structure has to be based on primary dimensions that characterize

reverse logistics.
- The framework has to be parsimonious, i.e. simplicity will be favored.
- The internal validity needs to be preserved by preserving consistency.
- The construct validity needs to be preserved. This is taken care of

by selecting the basic constructs, i.e. the dimensions of the framework
according to their meaningfulness to the understanding of the field.
2.3 Reverse Logistics: a review of literature on the-

oretical developments
The focus of this section is on literature contributing to structuring the field.
Thierry et al. (1995) describe strategic issues in product recovery manage-
ment. To facilitate the analysis, the authors first give a list of the options for
product recovery management. The options are characterized according to the
level of disassembly (product/ module/ part, etc.), the quality required (high/
upgrade to new, etc.), and the resulting product (parts/ modules/ materials,
etc.). The paper uses three case studies (BMW, IBM, and an anonymous
copier remanufacturer) to illustrate strategic product recovery management.
The authors do so by highlighting the changes in operations when the three
companies got engaged in recovery.
Fleischmann et al. (1997) present a review on quantitative models for

reverse logistics. Before coming to the review itself, the authors put forward
a classification of the situations in which reuse occurs. The classification is
given according to the following dimensions: motivation for reuse, the type
of recovered items, the form of reuse, and the involved actors. The paper
then goes over the available models to support reverse distribution, inventory
control in systems with return flows and production planning with reuse of
parts and materials.
Fuller and Allen (1997) put together a review of several important concepts
in the context of recycling of post-consumer solid waste. Thus, a classification
of the following recycling systems is given: manufacturer-integrated systems,
waste-hauler systems, specialized dealer-processor systems, forward retailer-
wholesaler systems, and temporary/facilitator systems. The authors argue
that the classification shows that the reverse distribution channels are making
a link between the source (post-consumer waste) and the market. The article
then continues with a discussion whether, or not, the reviewed variety of
systems are likely to be able to deal with the challenges of the future.
Carter and Ellram (1998) review the literature on reverse logistics to iden-
tify existing gaps and accordingly to propose an agenda for future investiga-
tion. While doing so, the authors identify drivers (regulators, customers,
and entrepreneurs), and constraints (such as stakeholder commitment, top
management, and vertical integration) regarding reverse logistics. The au-
thors distinguish between internal and external forces, i.e. company factors
and the task environment. Then a research agenda is built based on a set
of propositions, such as “successful reverse logistics programs require effec-
tive policy making,” or “top management support is necessary to ensure the
continued success of reverse logistics.”
Gungor and Gupta (1999) present an extensive review of the literature
(more than 300 articles and books) on environmentally conscious manufac-
turing and product recovery. They subdivide the literature in categories, out-
lining a framework. The authors look upon product recovery from the point
of view of environmentally conscious manufacturing. Both the regulations
and the pressure caused by government and customers are mentioned.
Goggin and Browne (2000) suggest a classification for resource recovery, fo-
cused on the electronics industry and on products at the end of their life. The
classification is given according to the following main lines: generic options,
operations (order or planning driven activities), and specific characteristics
(input and output complexity; degree of involvement by the OEM; collection,
2.4. Reverse Logistics: the basic dimensions 45
process and distribution complexity; and routing variability). Furthermore

the paper discusses electronic equipment manufacturing situations.
The existing categorizations in the field of reverse logistics, are
1. intentionally proposed for a solitarily restricted sub-area of the field, or
2. informally put forward to facilitate the analysis of a specific topic
Implicitly, these categorizations lack generalizability for reverse logistics

as a whole, as they constitute partial characterizations of reverse logistics
situations. Since, however reverse logistics is not yet a well-structured research
area, the aforementioned references are actually important seeds in organizing
the field. Next section describes how we use them in more detail.
Ultimately, let us emphasize that, in contrast with the previous literature,
this chapter does not focus on any specific form of reverse logistics. Instead,
we put forward overall basic dimensions of reverse logistics, such that a frame-
work on reverse logistics as a whole is put forward. Besides this, none of the
reviewed references explicitly describes which was the methodology to come
up with the suggested ingredients and the proposed classifications. Here, we
ground the framework on previous knowledge and on empirical evidence, as
described in the previous section, and as attested in the next sections.
2.4 Reverse Logistics: the basic dimensions

The reviewed literature brings order with respect to some identified ingredi-
ents in the universe of reverse logistics. As the review shows, for some ingre-
dients a typology is given, others are simply used to introduce the analysis.
Next we list the ingredients.
- motivation (Fleischmann, 1997)

- drivers (Carter and Ellram, 1998)
- environmental forces (Gungor and Gupta, 1999).
- type of items (Fleischmann, 1997)
- recovery options (Thierry, 1995)

- generic (recovery) options (Goggin and Browne, 2000)
- forms of reuse (Fleischmann, 1997)
- actors (Fleischmann, 1997)

- recycling systems classified according to the actors in the system (Fuller
and Allen, 1997)
Some of the ingredients are common to several of the references, though some-
times they are phrased differently. As the above list already suggests, the
ingredients can be clustered as follows.
• drivers (also called ‘motivation’ or ‘environmental forces’)
• type of items
• recovery options (also called ‘forms of reuse’)
• actors
A closer look shows that actually these are answers to the following funda-
mental questions:
• why is there recovery?
• what is recovered?
• how is the recovery done?
• who is doing the recovery?
These fundamental questions constitute basic dimensions of reverse logistics.

Yet, the literature does not give a formal classification of answers to all the
four questions. Besides this, though there are some formalized classifications
that are rather complete, others are not, or the structure behind it was never
explained. For instance, the list of recovery options put forward by Thierry et
al. (1995) is logically explained, rather complete, and generally accepted. In
contrast, the semi-formalized classification of type of items (see Fleischmann
et al., 1999) includes only packaging, spare parts, and consumer goods. Indus-
trial goods, and others, are kept out. This is not a surprise, as many authors
do not intend to formally structure the field. Besides this, the existing litera-
ture does not attempt to ground the classifications in empirical evidence for
reverse logistics as a whole.
Section 2.5 provides typologies for the basic dimensions of reverse logistics.
There we state what is the contribution from existing classifications, or we
logically argue how we arrive to them. To this end, the contribution of the
case study review is very important. The framework and the case study
2.4. Reverse Logistics: the basic dimensions 47
review were conducted in parallel interacting continuously with each other.

Thus, the typologies evolved through this interactive process. The typologies
presented next are in fact the final version. In the process, we strive to balance
exhaustiveness with meaningfulness for understanding reverse logistics as a
whole.
Before presenting the typologies, however, one can notice that the answer
to the first question, why, can be given from two perspectives: the companies
or the customers. Actually, the perspective of the customer is often informally
mentioned in the literature by naming the reverse flows like: end-of-life, end-
of-use, and so on. However, this has not been formalized.
Thus, we suggest to split the why question into two, as follows.
• Why-returning?
• Why-receiving?
Figure 2.2: The five basic dimensions of reverse logistics.
Summarizing, the skeleton of the framework is given by the following five

dimensions (see Figure 2.2):
• Why-receiving?
• Why-returning?
• What?
• How?
• Who?
2.5 Reverse Logistics: why? what? how? who?

This section deals with the fundamentals of reverse logistics by analyzing the
topic from five essential viewpoints: why-returning, why-receiving, what, how
and who. For each of these dimensions we provide one or more classification
of their elements. This section provides these typologies. In the next section
we go into more detail of how the dimensions are interrelated.
The following subsections are considered.
• Why-receiving: the forces driving companies and institutions towards

reverse logistics.
• Why-returning: the reasons why products are returned.
• What is being returned: product characteristics and product types.
• How are products recovered: processes and recovery options.
• Who is doing the recovery: the actors and their roles.
2.5.1 Why-returning: drivers

In the literature of reverse logistics, many authors point out driving factors
like economics, environmental legislation and environmental consciousness of
consumers (see e.g. RevLog, 2003). Generally, one can say that companies
get involved with Reverse Logistics, because i ) they can profit from it, or i )
they have to, or iii ) they feel socially impelled. Accordingly, we organize the
the typology driving forces according with the following elements:
• Economics (direct and indirect);
• Legislation;
• Corporate citizenship;
2.5. Reverse Logistics: why? what? how? who? 49
Figure 2.3: Driving triangle for reverse logistics.
Figure 2.3 depicts these three drivers as the driving triangle for Reverse
Logistics (compare with Carrol’s triple bottom line for corporate social re-
sponsibility, Carrol, 1979). Below we go into more detail regarding the three
drivers.
Economics
A reverse logistics program can bring direct gains to companies, such as
OEM’s, by dwindling on the use of raw materials, by adding value with re-
covery, and by reducing disposal costs. Independent parties have also entered
the reverse logistics arena, because of the financial opportunities offered in
the dispersed market of superfluous or discarded goods and materials. For
example, metal scrap brokers have made fortunes by collecting metal scrap
and offering it to steel works, which could reduce their production costs by
mingling metal scrap with virgin materials in their process. In the electronics
industry, many products have a short life-cycle, but at end-of-life they still
have a high intrinsic economic value. For instance, Recellular is a U.S. firm
that is taking economic advantage from recovery by trading in refurbished
cell phones (see Guide and van Wassenhove, 2001).
Even with no immediate expected profit, an organization can get more in-
volved with reverse logistics because of marketing, competition, and strategic
issues, from which indirect gains are expected. For instance, companies may
get involved with recovery as a strategic step to get prepared for future leg-
islation (see Louwers et al., 1999) or even to prevent legislation. In face of
competition, a company may engage in recovery to prevent other companies
from getting their technology, or to prevent them of entering the market. As
reported by Dijkhuizen (1997), one of the motives of IBM in getting involved
in recovery was to avoid that the brokers would do so.
Recovery can also be part of an image-building operation. For instance,
Canon has linked the copier recycling and cartridge recycling programs to the
“kyo-sei” philosophy, i.e. cooperative growth, proclaiming that Canon is for
“living and working together for the common good” (see Meijer, 1998; Canon,
2003). Recovery can also be used to improve customer’s or supplier’s relations.
An example is a tyre producer who also offers customers rethreading options
in order to reduce customer’s costs.
Summarizing, the economic driver embraces among others, the following di-
rect and indirect gains:
• Direct gains
- input materials
- cost reduction
- value added recovery
• Indirect gains
- anticipating/impeding legislation
- market protection
- green image
- improved customer/supplier relations
Legislation
Legislation here refers to any jurisdiction indicating that a company should
recover its products or take them back.
In many countries, home-shoppers are legally entitled to return the or-
dered merchandize (e.g. in the UK, see the Office for Fair Trading, 2003). Fur-
thermore, especially in Europe there has been an increase of environmentally-
related legislation, like recycling quotas, packaging regulations and take-back
responsibility for manufacturers. For instance, the European Directive on

Waste of Electronics and Electrical Equipment (WEEE) was approved by the
European Parliament in 2002 and sets the recovery quotas at 70%-90% (see
Europa, 2003). In fact, industries, like the electronics and the automobile in-
dustry, are under special legal pressure (see Bloemhof-Ruwaard et al., 2004).
Sometimes, companies participate “voluntarily” in covenants, either to deal
with (or to get prepared for) legislation.
Summarizing, the legal driver embraces the following two elements.
• Consumer rights
• Pro-environmental legislation
Corporate citizenship
Companies use this term ‘corporate citizenship’ to express that they respect
society out of good principles. Note, that there are also other terms used such
as social responsibility or ethics. Though the term ‘corporate citizenship’
many not be generally accepted yet, it fits well our purposes, i.e. it represents
when a firm is “feeling” socially impelled to act in certain way (see also Tichy
et al., 1997).
In the context of reverse logistics, corporate citizenship concerns a set of
values or principles that impels a company or an organization to become re-
sponsibly engaged with reverse logistics. For instance, the concern of Paul
Farrow, the founder of Walden Paddlers, Inc., with “the velocity at which
consumer products travel through the market to the landfill,” pushed him to
an innovative project of a 100%- recyclable kayak (see, Farrow et al., 2000).
Indeed, many firms have extensive programs on responsible corporate citizen-
ship where both social and environmental issues become the priorities (see
Shell 2003).
Figure 2.4 summarizes the typology ‘drivers’ for the dimension ‘why-receiving’
as described in this section.
Companies are often involved with reverse logistics for a mix of these
driving forces and in reality it is sometimes hard to set the boundary. For
example, in many countries customers have the right to return products pur-
chased via a distant seller like mail-order-companies or e-tailers. Thus, these
companies are legally obliged to give the customer the opportunity to return
merchandize. At the same time, this is also perceived as an opportunity to
attract customers, bringing potential benefits to the company.
Figure 2.4: The typology ‘drivers’ for the dimension ‘why-receiving’.

2.5.2 Why-returning: return reasons

Roughly speaking, products are returned or discarded because either they do
not function properly or because their function is no longer needed. These
return reasons are here presented in conformity with the usual supply chain
phases: starting with manufacturing, going to distribution until the products
reach the customer. Therefore, three groups are differentiated: manufacturing
returns, distribution returns and customer returns.
Manufacturing returns
Manufacturing returns are here defined as all those returns for which the need
for recovery of components or products is identified during the production
phase. This occurs for a variety of reasons. Raw materials may be left over,
intermediate or final products may fail quality checks and have to be reworked
and products may be left over during production, or by-products may result
from production. Raw material surplus and production leftovers represent
the ‘product not-needed’ category, while quality-control returns fit in the ‘do-
not-function’ category.
Summarizing, manufacturing returns include:
• raw material surplus
• quality-control returns
• production leftovers/by-products
Distribution Returns
Distribution/supply returns refers to all those returns that are initiated during
the distribution phase. It refers to product recalls, commercial returns, stock
adjustments, and functional returns.
Product recalls are products recollected because of safety or health prob-
lems with the products, and the manufacturer or a supplier is usually the
initiator, and not the customer (see Smith et al., 1996). Product recalls fall
in ‘distribution returns’ as they are usually initiated during this phase and
they are anyway specially demanding with respect to distribution.
B2B commercial returns are all those returns for which a buyer has a
contractual option to return products to the seller (see a review of supply
chain contracts at e.g. Tsay et al., 1999). This can refer to wrong/damaged
deliveries, or to unsold products that retailers or distributors return to, e.g.
the wholesaler or manufacturer. The latter include outdated products, i.e.
those products for which the shelf life is too short and may no longer be sold
(e.g. pharmaceutical products and food).
Stock adjustments occur when an actor in the chain re-distributes stocks.
For instance stock adjustments also occur between warehouses or shops for
instance in the case of seasonal products (see De Koster et al., 2002). Thus,
stock adjustments occur within a company while commercial returns involve
more than one company alone.
Finally, there are products for which their inherent function makes them go
back and forward in the chain. We suggest to call these as ‘functional returns’.
An obvious example is distribution carriers like pallets: their function is to
carry other products and they can serve this purpose several times (see e.g
Kroon and Vrijens, 1995 and Duhaime et al., 2000).
Summarizing, distribution returns comprehend:
• product recalls,
• B2B commercial returns (e.g. unsold products, wrong deliveries,)
• stock adjustments, and
• functional returns (distribution items/carriers/packaging).
Customer Returns
The third group consists of customer returns, i.e. those returns initiated
once the product has reached the final customer. Again, there is a variety of
reasons to return the products, viz.
• B2C commercial returns (reimbursement/other guarantees),
• warranty returns,
• service returns (repairs, spare-parts, etc.),
• end-of-use returns, and
• end-of-life returns.
As much as possible, the reasons have been listed according to the life cy-
cle of a product. B2C commercial returns, like reimbursement guarantees,
give customers the opportunity to change their minds about purchasing when
their needs or expectations are not met (usually shortly after having re-
ceived/acquired the product). The list of underlying causes is long. For
instance, with respect to clothes dissatisfaction it may be due to size, color,

fabric’s properties, and so forth. Independent of the underlying causes, when
a customer returns a new product, benefiting from a money-back-guarantee
or an equivalent, we are in the presence of B2C commercial returns. The
next two reasons, warranty and service returns, refer mostly to an incorrect
functioning of the product during use, or to a service that is associated with
the product and from which the customer can benefit.
Initially, customers benefiting from a warranty can return products that do
not (seem to) meet the promised quality standards. Sometimes, these returns
can be repaired. Otherwise, a customer may get a new product or his/her
money back after which the returned product can be recovered. After the
warranty period has expired, customers can still benefit from maintenance or
repair services, but they no longer have the right to get a substitute product
for free. Products can be repaired at the customer’s site or sent back for
repair. In the former case, returns commonly occur in the form of spare-
parts, since in advance it is hard to know precisely which components are
going to be needed for the repair.
End-of-use returns refer to those situations where the user has a return
opportunity at a certain life stage of the product. This refers to leased prod-
ucts and returnable containers like bottles, but also to returns to second-hand
markets as the one of Bibliofind, a division of Amazon.com for used books
(see Amazon.com, online).
Finally, end-of-life returns refer to those returns for which the product as
such is at the end of its economic or physical life. They are either returned to
the OEM because of legal product-take-back obligations, or other companies
like brokers, collect them for value-added recovery (see Section 2.5.4).
Figure 2.5 summarizes the typology ‘return reasons’ for reverse logistics in
the three stages of a supply chain: manufacturing, distribution, and customer.
One should note however that the distinction between these three stages might
be factitious. In practice it is not always easy to pinpoint exactly where
manufacturing ends and distribution starts, as value may be added when
some sort of distribution has already begun.
2.5.3 What: types and characteristics

A third viewpoint on reverse logistics can be obtained by considering what is
actually being discarded or returned. Standard logistics product properties,
such as size, weight, value, easiness of transportation and so on, play also a
role for reverse logistics.
Figure 2.5: The typology ‘return reasons’ for the dimension ‘why-returning’.
Furthermore, we identify three product characteristics that seem relevant

to the organization and to the profitability of reverse logistics systems, viz.
• composition;
- homogeneity
- disassembility
- testability
• deterioration
- economical
- physical
• use-pattern
- location
- intensity
- duration
- bulk use vs. individual
Composition
As highlighted by Gungor and Gupta (1999), product composition in terms of
the number of components and of materials is one of the many aspects to keep
in mind while designing products for recovery. Not only the number, but also
how the materials and components are put together, will affect the easiness of
disassembling, and re-processing them, and therefore the economics of reverse
logistics activities (Goggin and Browne, 2000 discuss product complexity).
The same holds for testability (e.g. in case of the recovery of mobile phones).
The presence of hazardous materials is also of prime relevance, as it demands
special treatment. The material heterogeneity of the product can play a role
in recovery, where one may try to separate streams of materials (see also
Krikke et al., 1999b). The size of the product has also been pointed out as a
significant factor for recovery systems (see Goggin and Browne, 2000). One
can imagine e.g. the impact of this aspect on transportation. Summing up,
the intrinsic characteristics of a product, like homogeneity, disassembility and
testability, are decisive for the recovery, since they affect the economics of the
whole process.
Deterioration
Next, product and materials undergo deterioration processes, viz. physical
(homogeneous or not) and economic. This strongly affects the recovery option.
Several questions have to be asked in order to evaluate the recovery potential
of a product: does the product age during use? (intrinsic deterioration); do
all parts age equally, or not? (homogeneity of deterioration); does the value of
the product decline fast? (economic deterioration). Deterioration eventually
causes a non-functioning status of the product, but also determines whether
there is enough functionality left to make further use of the product, either
as a whole or as parts.
In fact, products may become obsolete because their functionality becomes
outdated due to the introduction of new products in the market, as it happens
with computers. This will restrict the recovery options that are economically
viable. The same can be said for the intrinsic deterioration and whether or
not it is homogeneous. If a product is consumed totally during use, such as
gasoline, or if it ages fast, like a battery, or if some parts are very sensitive to
deterioration, reuse of the product as such is out of the question. If however,
only a part of the product deteriorates, then other recovery options like repair,
or parts retrieval and replacement, may be considered.
Use-pattern
The product use pattern, with respect to location, intensity, and duration
of use, is an important group of characteristics as it affects for instance the
collection phase (see Section 2.5.4). It will make a difference whether the
end-user is an individual or an institution (bulk-use), demanding different
locations for collection or different degrees of effort from the end-user (e.g.
bring the product to a collection-point), see also Chapter 3. The use can
also be less or more intensive. Let us consider the case of leased medical
equipment, which is commonly used for a small time period, and it is likely to
be leased again after proper operations like sterilization. Time is not the only
component describing intensity of use, but also the degree of consumption
during use. Consider for instance the example of reading a book. Quite
often one reads a book only once after the purchase and keeps it, but does
not do anything more with it. This has stimulated Amazon.com to start her
successful second-hand trading of books.
The characteristics of the product are related with the type of product in
question. The product type in fact gives the first global impression on the
potential status of the product when it reaches recovery. Zhiquiang (2003)
uses for instance a typology for products in order to sketch the planning of
reverse logistics activities. Fleischmann et al. (1997) distinguish the following

types: spare parts, packaging, and consumer goods. A natural addition is
the class of industrial goods, which in general are more complex and have a
different use pattern than consumer goods. Furthermore, by looking at the
United Nations (UN) classification of products (see UN, 2003), and at the
relevant characteristics for reverse logistics (as described previously) we have
to discriminate a few more classes of products, like: ores, oils, and chemicals;
civil objects, and other transportable materials. Everyone knows that civil
objects have a long useful life. Besides this, recovery has to be mostly in-
site as objects like bridges and roads are not easily removed and transported,
though there are cases where this has happened. Ores, oils, and chemicals
are a special category due to their common hazardous composition needing
specialized handling during any recovery process. Finally, a separate category
is of other materials, such as pulp, glass, and scraps. In sum, the following
main product categories can be discriminated:
• civil objects (buildings, dikes, bridges, roads, etc.),
• consumer goods (apparel, furniture, and a vast variety of goods),
• industrial goods (e.g. military and professional equipment, etc.),
• ores, oils, and chemicals,
• packaging and distribution items (like product carriers),
• spare-parts, and
• other materials (like pulp, glass, and scraps).
Figure 2.6 summarizes the typologies ‘product characteristics’ and ‘product

types’ for the what dimension.
2.5.4 How: processes and recovery options

This section deals with how reverse logistics works in practice, i.e. how is
value recovered from products.
Recovery is actually only one of the activities of reverse logistics. Other
processes include collection, inspection/testing selection, and sorting. Collec-
tion refers to bringing the products from the customer to a point of recovery.
At this point the products are inspected, i.e. their quality is assessed and a
decision is made on the type of recovery. Products can then be sorted and
routed according to the recovery decision. Thus, we have:
Figure 2.6: The typologies ‘product characteristics’ and ‘product types’ for
the dimension ‘what’.
• collection,
• inspection/ testing,
• selection,
• sorting, and
• recovery itself.
If the quality is “as-good-as-new,” products can be fed into the market

almost immediately through direct recovery. If not, other types of recovery
may be involved but now demanding more action, i.e. a form of re-processing.
Thus, we distinguish two groups of recovery, which we discuss next (Thierry
et al., 1995 have used the term ‘direct reuse’ and Fleischman et al., 1997
employed the term ‘ reused directly’).
• Direct Recovery
• Process Recovery
Direct Recovery
Note that re-use, re-sale and re-distribution are slightly different. Re-sale
applies to situations where the product is sold again. Re-use is a situation
in which the product is used again, but there is no purchase, e.g. unused
spare parts. Re-distribution refers to products like carriers, that are simply
distributed again and again. Direct recovery involves the three following
options:
- Re-use
- Re-sale
- Re-distribution
Process Recovery
Process recovery can involve several operations such as cleaning, disassem-
bling, and re-assembling. In spite of the detailed operations, process recovery
can be said to occur at different levels as explained next. For complete defi-
nitions, please see Thierry et al. (1995). For more on recovery/re-processing
levels, please have a look at Fleischmann et al. (1997), and Goggin and
Browne, (2000).
A product can be recovered as a whole, by being repaired at a product

level (repair). At module level, the product, e.g. a large installation, build-
ing, or other civil object, can get upgraded (refurbishment). In the case of
component recovery, products are dismantled and used and new parts can be
used either in the manufacturing of the same products or of different products
(remanufacturing). Re-processing may involve a guided selection process of
parts for recovery (retrieval). In the case of material recovery, products are
usually grinded and their materials are sorted out and grouped according to
quality criteria, so recycled materials can be raw material input for industry,
such as paper pulp and glass (recycling). Finally, in energy recovery products
are burned and the released energy is captured (incineration). If none of these
recovery processes occur, products are likely to go to landfill.
Summarizing, process recovery includes the following recovery options:
- repair (at product level),
- refurbishing (at module level),
- remanufacturing (at component level),
- retrieval (at part level),
- recycling (material level), and
- incineration (energy level).
Figure 2.7 gives the typologies ‘processes’ and ‘recovery options’ for the di-
mension ‘how’.
Figure 2.8 represents recovery options ordered regarding the level of recov-
ery (product as a whole; module; component; part; material; or, energy level).
This figure is in the form of an inverted pyramid, since repair and refurbishing
are forms of recovery at a more global level (at the top of the pyramid), while
recycling and incineration are forms of recovery in a more specific content
level (at the bottom of the pyramid). Please note that returns in any stage of
the supply chain (manufacturing, distribution and customer ) can be recovered
according to options from the top to the the bottom of the pyramid.
One can find overall similarities between the inverted pyramid of Figure
2.8 and Lansink’s waste hierarchy introduced in 1979 by this Dutch member
of parliament, being: prevention, re-use, recycling, and proper disposal (see
VVAV, 2003).
At first, one may think that recovery options at the top of the pyramid
are of high value and more environmentally friendly than options close to the
Figure 2.7: The typologies ‘processes’ and ‘recovery options’ for the dimension
‘how’.
bottom, which apparently recover less value from the products. It is important
to stress, however, that both thoughts are not necessarily true. Originally, the
Lansink’s hierarchy was put together regarding the environmental friendliness
of the recovery option. Yet, the hierarchy is not entirely true even with respect
to environmental friendliness. For instance, in the case of paper recycling
versus land filling, one may go against recycling by arguing that paper is
biodegradable and therefore land filling may require less energy than the de-
inking and the bleaching processes necessary for recycling. With respect to
the economic value of each recovery option, this depends for instance on the
existence of a matching market. This is because after recovery, products,
materials, and energy can go again to the market (see Figure 2.9). Thus, it
is possible that a used product has essentially no market value as such, but
is very valuable as a collection of spare parts.
Nonetheless, with respect to process recovery and if there is a market
for all the options, one is likely to select the options from top to bottom
(see Figure 2.9). This of course, if the product has enough quality to be
done so. The return reason plays a role here. If a customer is returning
the product because it needs to be ‘restored to working condition’ (service
returns), direct recovery is very unlikely. Note as well that the lines in the
figure separate recovery options that very much affect different parts of the
organization. Direct recovery mainly affects distribution, the middle group
Figure 2.8: Recovery option inverted pyramid

Figure 2.9: Reverse logistics processes
affects production planning, and the third group is most likely outsourced
by the organization. So, also depending of the recovery option, there are
opportunities in the reverse logistics arena for certain type of players. This is
further illustrated in Section 2.6.
2.5.5 Who: actors and roles

Essentially, the following players are the most crucial (see also the viewpoint
of Fuller and Allen, 1997):
• forward supply chain actors, (supplier, manufacturer, wholesaler, re-

tailer, and sector organizations),
• specialized reverse chain players (such as jobbers, recycling specialists,

dedicated sector organizations or foundations, pool operators, etc.),
• governmental institutions (EU, national governments, etc.), and
• opportunistic players (such as a charity organization).
These players can have the following roles:

• managing/organizing
• executing
• accommodating
Figure 2.10 gives the typologies ‘actors’ and ‘roles’ for the dimension ‘who’.
Figure 2.10: The typologies ‘actors’ and ‘roles’ for the dimension ‘who’.
At the managing level, parties are responsible for (organizing) the reverse
chain/network. Other players simply execute activities in the network. The
final role to add is the accommodator role, performed by both the sender/giver
and the future client (the market), without whom recovery would not make
much sense (see also Figure 2.11). The group of actors involved in reverse
logistic activities, such as collection and processing, are commonly indepen-
dent intermediaries, specific recovery companies (e.g. jobbers), reverse logistic
service providers, municipalities taking care of waste collection, and public-
private foundations created to take care of recovery. Each actor has different
2.6. The framework 67
objectives, e.g. a manufacturer may do recycling in order to prevent jobbers

reselling his products at a lower price (see Section 2.5.1). Furthermore, the
various parties may compete with each other.
Figure 2.11: Who is who in reverse logistics
Figure 2.11 gives an overall view on the type of players and their roles. The
parties at the top of the picture are responsible (or made responsible by leg-
islation) and/or they organize the network. They are from the forward chain,
like the OEM, or they are governmental entities like the European Union. At
the execution level, there are collectors, processors, and re-distributors, which
can be different parties. Between the managing and the execution level, there
are exchanges of information and capital.
2.6 The framework

In Section 2.4 we constructed the skeleton of the framework, by identifying
its five main dimensions: the return reasons (why-returning), driving forces
(why-receiving), the type of products and their characteristics (what), the
recovery processes and recovery options (how ), and the actors involved and
their roles (who). In addition, we provided “mass for the skeleton” by putting
forward a typology for each of the dimensions, in the previous section.
A detailed investigation of how reverse logistics situations are affected
by these five dimensions is not yet available in the literature. However, in
order to determine the situation itself, the combination of these dimensions

is crucial.
In a product recovery setting, product type and return reason are exoge-
nous entities, and they work as inputs. Given these inputs one has to decide
on how to recover and who is responsible for what. These decisions are also
influenced by the product recovery drivers. Figure 2.12 gives a schematic
representation of the relations between the five dimensions of the framework.
The elements in the typologies of input dimensions, such as why-returning,
are related with the elements of the output dimensions, such as how-recovery
options. Different return reasons, product characteristics, and different drivers
seem many times to lead to different recovery options and actors (see Chapter
3). Many factors play a role, however, and firms may end up with different
options. Thus, the same relations are not going to hold always. Nonethe-
less, below we give a number of typical situations. For each situation, we use
the framework to identify the problematic issues and the associated decision-
making areas.
By-product returns
Let us consider a typical situation of by-products in the pharmaceutical in-
dustry (see e.g. Inderfurth et al., 2004). During the manufacturing process,
chemical by-products are released (why-returning and what-type). These by-
products contain valuable substances that can again serve as input to the
production process. So, the manufacturer can directly gain from recovering
these substances (why-receiving). Please note that in this specific case, the
sender and receiver is one and the same organization, i.e. the manufacturer.
After retrieval the valuable substances are used again in the production
process (how-recovery option). However, the by-products deteriorate rapidly
(what-characteristics). This has an immediate impact on the way that pro-
cesses are organized. For instance, recovery is integrated with production
in the same production plant (how-processes and who). As a consequence,
both production and retrieval of substances compete for physical space and
manpower.
It is clear that the input dimensions, why-returning and what, have an im-
pact on the how and on the who. Furthermore, the combination of elements
defines the problematic issues and the associated decision-making areas. In
this case, coordination of production and retrieval operations is clearly essen-
tial. Figure 2.13 summarizes the above.
Figure 2.12: A framework for reverse logistics.

Figure 2.13: Illustrative example of by-product returns.
Functional returns
Functional returns are products that return due to their inherent function, like
distribution items (why-returning and what-type). A typical industry where
this type of returns occurs is the beverage industry.
A distribution item usually can be re-distributed several times, since it was
designed to be robust enough to protect product integrity in the first place
(what-characteristics). This makes re-distribution a natural and economically
viable recovery option (how). The drivers for recovery are economics and
possibly, also legislation in the form of recovery quotas, like it happens in
Europe (why-receiving).
Since the inherent function of distribution items is to go back and forward
in the chain, containers are carefully checked and cleaned, so they can be safely
redistributed (how-processes). Moreover, many actors are involved. Typically,
distribution items are pooled with several organizations. Thus forward players
as well as pool operators are usually involved (who).
Sometimes even parties from competing supply chains are involved. In this
setting, keeping control over the distribution items is problematic. One of the
issues is preventing leakage, i.e. preventing that items get lost or are stolen
throughout the chain. Note that the forward distribution depends heavily on
the availability of serviceable distribution items. So, one of the key-issues is

data collection (how many items are where) and forecasting how many items
come back when (see Figure 2.14).
Figure 2.14: Illustrative example of functional returns.
B2C commercial returns (reimbursements)

Let us consider the catalogue retail industry (see also Chapters 4 and 5, where
the catalogue retail industry is described in more detail).
After receiving the ordered product, the customer has a limited amount
of time to return it and claim reimbursement (why-returning). The industry
offers reimbursements. This is because, on the one hand, companies want to
improve customer relations, and, on the other hand, there is legal enforcement
(why-receiving). In principle, consumer goods like apparel (for which the
catalogue industry is huge) do not physically deteriorate in the meantime, but
due to seasonality the products suffer from economic deterioration (what).
A natural choice for recovery is direct recovery, by re-selling them. The
what-characteristics affect directly the way the company is going to organize
it. For instance, fashion apparel has to be re-sold during the same season
due to seasonal deterioration. Therefore, speed in re-selling is very important

(how). The products are likely to be brought back to the warehouse so its
integrity is checked and the products can be re-packaged (how-processes).
Thus, at least some actors of the forward chain are also involved in the reverse
chain (who-actors).
Considering the above, a crucial decision-making area is return handling
with the following being one of the issues: do the forward and reverse processes
(transportation, warehousing activities, etc.) need to be combined or not?
(see Figure 2.15). Chapter 5 further investigates this issue.
Figure 2.15: Illustrative example of reimbursement returns.
Service returns
In many B2B businesses, the original equipment manufacturer provides after-
sales service on valuable industrial products like e.g. aircrafts (why-returning
and what). After-sales service contracts are often very profitable, since this
service is very critical to maintain industrial systems and contracts are often
negotiated for the long-term (why receiving).
When a customer calls for an after-sales service due to a mal-function
of the equipment, a service engineer is sent to bring the equipment back to
working order, i.e. to repair it (who and how-recovery option). In order to

deal with the particular problem the engineer brings an inventory of tools and
spare parts.
A likely outcome is that some parts need to be replaced during the ser-
vice. Depending on their condition, replaced parts are returned for recovery.
The spare-parts that are not used by the engineer are returned and directly
restocked in the warehouse for spare parts (how).
Since we are dealing with industrial goods, service is very critical, and
therefore fast response times are crucial. This opens up opportunities for
3PL’s such as fast-courier services (who). It also puts major importance
on decision-making areas such as network design (centralization vs. decen-
tralization, network coverage) and inventory management such as inventory
deployment and accessibility (see Figure 2.16).
Figure 2.16: Illustrative example of service returns.
End-of-use returns
Consider the B2B leasing of industrial goods such as copiers (what-type).
Normally, after a lease contract expires, the copier is returned (why-returning).
Since these copiers are designed to be very durable, returned copiers are likely
to represent a high value (what-characteristics).
With these product characteristics, there is an excellent opportunity for
obtaining economic gain from remanufacturing (why-receiving and how-recovery
option). Typically, the recovery is done by the original manufacturer (who),
since they have the know-how and the resources to deal with these complex
machines. Since the recovery is done by the original manufacturer, it is nat-
ural to integrate the recovery with the production processes.
The questions are however to determine the the degree of integration and
how to do this efficiently. There is also a quality issue. Remanufacturing is
defined as ’the process that turns a product into as-good-as-new condition’.
Thus, how far should disassembly and recovery efforts go, so sufficient qual-
ity is guaranteed, but without wasting resources? In addition, the recovery
process would be more efficient in presence of design for remanufacturing (see
Figure 2.17).
Figure 2.17: Illustrative example of end-of-use returns.

End-of-life returns
Consider a well-known closed-loop recovery system, i.e. the paper industry.
Paper that is collected at the end-of-life can serve as pulp input for the indus-
try itself (what and why-returning). So, the industry has a direct economic
gain from the recovery of paper. In Europe, legislation is also a driver as
the paper industry is pushed to increase the percentage of paper recovered
(why-receiving). Since paper is a material, obviously, the only way to recover
it at the same time that serves as input for the paper industry, is recycling
(who and how-recovery option).
In paper-recycling, speed is not so much an issue, since we deal with a
material that does not deteriorate quickly. Therefore inventories and the recy-
cling process can be centralized. There is, however, a need for big investments
in recycling technology. Huge quantities need to be collected in order to take
advantage of economies of scale.
Acquisition is an issue, since you need the collaboration of the end-user
to separate the paper from the waste stream. Furthermore, the geographic
spread of paper-use calls for a very fine-branched network, which requires
even more investments (see Figure2.18).
Figure 2.18: Illustrative example of end-of-life returns.

Above, we illustrated with typical situations that filling the five dimensions
of the framework with information helps to identify the problematic issues
and the associated decision-making areas. The exact influence and relations
of the five dimensions of the framework presented here (why-receiving, why-
returning, how, what and who) is still an open question that requires more
investigation. There are nonetheless some leads in the literature, which we
discuss next.
Goggin and Browne (2000) developed typologies to assist to determine
the recovery level (in their case, remanufacturing, component recovery, and
material recovery) from product characteristics, like input product complex-
ity and output product complexity, together with some other factors. The
product complexity is defined after the number of levels and constituents in
the Bill of Materials describing the product. Future research can extend this
type of research. The authors restricted themselves to one kind of consumer
goods (electronic products) and two recovery options (reuse and refurbish-
ing). Future studies can embrace more recovery options and more product
types, covering a diversity of reverse logistics systems.
Krikke et al. (2003) study how value can be regained from returns. The
authors identify seven typical returns. They characterize them according to
the 1) features of the product (obsolescence, complexity, and remaining life-
time), 2) the influence of legislation, 3) the most associated recovery option,
and 4) market characteristics. Then, the authors try to match each reverse
flow with a specific type of reverse chain. They characterize this reverse chains
to 1) the main driver, 2) degree of centralization, 3) type of processes, and
4) the IT support. The main conclusion is that integration with the forward
chain provides the best opportunities, in particular when based on modular
reuse. The framework proposed in this chapter can be used in similar studies.
Chapter 3 indicates that per decision-making area, certain dimensions of

the framework have a predominant role in identifying the problematic issues.
The chapter puts forward the following predominant relations:
• Networks ⇔ Recovery option (How ) and type of Actors (Who)
• Relationships ⇔ Actors (Who)
• Inventory Management ⇔ Return reason (Why-returning) and Product

type (What)
• Production Planning ⇔ Recovery option (How ) and type of Actors

(Who)
2.7. Summary and conclusions 77
If one wants to get more in-depth details, build-up theory, or explain phenom-
ena, one has to go into more depth regarding some aspects of the framework.
2.7 Summary and conclusions

This chapter provided a framework for reverse logistics directly contributing
to
• structuring the field and

• providing an overall understanding of reverse logistics issues.
Five key dimensions were considered, with their respective typologies (see also
Figures 2.4, 2.5, 2.6, 2.7 and 2.10):
1. Why-receiving?, i.e. why companies get involved reverse logistics (drivers),

2. Why-returning?, i.e. the reasons for products being returned (return
reasons),
3. How?, i.e. how is recovery carried out (processes and recovery options),
4. What?, i.e. what is being returned (product characteristics and product
types)
5. Who?, i.e. who is doing the recovery (actors and their roles).
Finally, we have pointed out that the five dimensions and the typologies of the
reverse logistics framework not only give context to reverse logistics situations.
Also their combination determines to a large extent the kind of issues that
arise in implementing, monitoring, and managing such situations. This was
illustrated with a collection of typical reverse logistics situations.
The framework is also a means to raise questions. For instance, the pre-
cise influence of the five dimensions of the framework presented here (why-
receiving, (why-returning, how, what and who) is still an open question that
requires more investigation. Some of the work going in this direction, such
as Goggin and Browne (2000) can be extended by taking into account more
product types and return options. The typologies presented in this chapter
can provide a basis for this.
Furthermore, Chapter 3, a review of more than sixty case studies, touches
upon this issue. There, the framework is used to facilitate the analysis, shed-
ding more light on the relations between the five dimensions, and on their
importance with respect to characterizing reverse logistics situations.
We are confident that the framework presented here is going to be a plat-

form for making further progress in the field, by inspiring hypotheses that,
once tested, can provide the roots for reverse logistics theory.
Chapter 3
Reverse Logistics: a review of

case studies
“Practice is everything”
Periander
3.1 Introduction
This chapter reviews more than 60 case studies on the field of reverse logistics,
giving a fair idea of the diversity of reverse logistics practices. The review
not only gives an overview of real life situations, but it is also a source of
case support for researchers. This is important because one cannot provide
proper insights into reverse logistics without being familiar with how firms are
dealing with it in practice, what are the trade-offs during decision-making,
how decisions are being supported, and so on.
Field studies and surveys can be extremely helpful in getting this knowl-
edge. At the end of the nineties, Rogers and Tibben-Lembke (1999) presented
a broad collection of reverse logistics business practices, giving special atten-
tion to the U.S. experience, in which the authors carried out a comprehensive
questionnaire. One year later, Guide (2000) published a study on research
needs based on an extensive survey within the remanufacturing industry in
the U.S. Also, quite some case studies on reverse logistics have been described
in the literature dealing with different industries, recovery options and drivers.
This literature is however scattered over journals for very different research
communities. Besides this, in countries in the forefront of reverse logistics
79
80 Chapter 3. Reverse Logistics: a review of case studies
a substantial number of case studies have been published in other languages

than the lingua franca and are therefore not accessible for the majority of
the research community. This chapter can also be used as a check list for
researchers willing to conduct (comparative) case studies.
The analysis of the case studies is based on the framework of Chapter 2.
Concisely, according to the framework, a first characterization of reverse lo-
gistics situations starts with answering the following questions: what?, how ?,
who?, and why?. I.e., what products/materials are being returned? how are
they being recovered?, who is involved?, and why is all this happening?
Actually the two research projects, the framework and the case studies
review were carried out in parallel, and in an interactive way. The analysis
of the case studies is facilitated by the framework. At the same time, the
case studies put the adequateness of the typologies and its completeness at
test, leading to a refinement of the framework when necessary. So, on the
one hand, the case studies review is a means to ground the framework on
empirical evidence, and on the other hand, the framework is a means to “see
through” the case studies. The framework of Chapter 2 constitutes the “last
words” of the interaction between this research work and the framework.
We present the case studies according to the following decision-making
areas, widely used in Operations Management: Network Structures, Relation-
ships, Inventory Management, and Planning and Control. For each section
we describe and organize the cases and we put forward preliminary insights as
well as research gaps, which we link with a research agenda. For some of the
decision-making areas, some of the dimensions of the framework are predom-
inantly important. We also discuss this and we provide research directions
to further investigate this, by indicating the kind of elements that could be
considered in more detail, depending on the decision-making area.
The remainder of the chapter is organized as follows. Section 3.2 describes
the methodology employed for finding and classifying the case studies. Sec-
tion 3.3 provides statistics regarding the type of industry, the product and the
geographic area of the cases. After that, Sections 3.4 to 3.8 present the case
studies, and preliminary insights related to the state of the art of research in
the field. Section 3.9 presents the overall insights of this review. Finally in
Section 3.10 we present the conclusions and we put forward research direc-
tions. An earlier version of this work has been reported in De Brito et al.
(2003).
3.2. Methodology 81
3.2 Methodology
“Case study” here means a close analysis of the practice (in scientific liter-
ature), together with the circumstances and its characteristics leading to an
understanding of the situation within its own context (see Stake, 1995). Note
that this definition rules out mere examples that are short on details. The
approach also excludes the pure description of business practice in profes-
sional magazines. We do not contest the usefulness of professional magazines
in learning about reverse logistics practice (see e.g. Duijker et al., 1994-2000).
However, articles in professional magazines more likely than scientific litera-
ture include immaterial content and are biased with promotional aims. Thus,
our main source of literature is scientific literature. Yet, we includ refereed
professional literature like the Handbook of reverse logistics (Van Goor et al.,
1994-).
The search procedure was as follows. The search was carried out by three
researchers, one junior and two seniors. We inspected the Science Citation,
and the ABI/Inform online libraries, using the combination of key words listed
in Table 3.1.
Table 3.1: Key words used for the search of case studies
Asset recovery
By-products/byproducts Post-consumer Repair
Containers Producer responsibility Repairable
Co-products/co-products Product ownership Resale / re-sale
Core Product recovery Resell / re-sell
Defects Product stewardship (commercial) Return
Defective Reassembly Reuse/ re-use
Disassembly Rebuild Reutilization
Dismantling Recalls Reusable
Disposal Reclaim Reverse logistics
Downgrading Reclamation Rework
Energy recovery Reconditioning Salvage
Environment Re-consumption Secondary (market)
Garbage Recovery (product, asset) Separation
Gate keeping Recycling Source reduction
Green logistics Refill Take back
Material recovery Refillable Upgrading
Obsolete (stock) Refurbishing Value recovery
Outlet Remanufacturing Warranty
Overstock Repack Waste
We also survey literature that is especially dedicated to the topic, like the
proceedings of renowned conferences, namely the APICS Reman and IEEE
on Electronics and the Environment (see APICS, 2003, and IEEE, 2003).
The main search took place in early 2001 and continued till end 2002.
Note that we report only on the literature that fit the case study criteria,
i.e. leading to an understanding of reverse logistics issues, through a close
analysis of a real case.
To give the reader the overall reverse logistics context, we gathered, for
each case study, information on the five dimensions of reverse logistics, i.e.
what, how, who, and why-returning, why-receiving as introduced in Chapter
2.
In order to be engaged in, or to anticipate recovery activities, firms have
to consider several strategic, tactical and operational matters. At the strate-
gic level, the design of the recovery network has to be decided upon first. At
the tactical level, relationships with partners have to be developed. Relevant
issues here concern measures to ensure that the actors involved follow a cer-
tain behavior. At the operational level, inventories and the recovery activities
have to be managed and controlled (see Ganeshan et al., 1999; Fleischmann
et al, 1997). Information and communication technologies play an impor-
tant role to achieve this. Thus, we present the case studies according to
the following decision-making focus: Network Structures, Relationships, In-
ventory Management, and Planning and Control. Furthermore, there is an
overview of Information and Technology (IT) for reverse logistics. By IT
we denote technological means to process and transmit information (see IT
definitions by OECD, 2003, and NAICS, 2003). If a case focuses on multiple
decisions, then the case study is discussed in more than one section. With our
search procedure, we found 26 case studies on Network Structures,10 cases
on Relationships, 14 cases on Inventory Management, 25 cases on Production
Planning and Control, and 7 cases dedicated to IT issues.
3.3 Statistics
We have found more than sixty cases involving reverse logistics. Using the
United Nations classifications for Industry (see http://esa.un.org), roughly
60% of the cases are in the manufacturing category; about 20% are within
wholesale and retail trade and about 10% in construction. We have also
found cases in the following categories: transport and communication, public
administration and defense, and other community services (see Table 3.2).
We also use here the United Nations classifications for Products (see
http://esa.un.org). In Chapter 2 we actually expanded this classification
(what-types) as for thorough research, a more detailed typology is necessary.
For an overall perspective, we state here the statistics according to the United
3.3. Statistics 83
Table 3.2: United Nations classifications for Industry (see http://esa.un.org)

A-Agriculture, hunting and forestry
B-Fishing
C-Mining and quarrying
D-Manufacturing
E-Electricity, gas and water supply
F-Construction
G-Wholesale and retail trade
H-Hotels and restaurants
I-Transport, storage and communications
J-Financial intermediation
K-Real estate, renting and business activities
L-Public administration and defense
O-Other community, social and personal service activities
P-private households
Q-Extra-territorial organizations
Nations classification. We observe that almost half of them deal with metal
products, machinery and equipment. Around 30% of the products being pro-
cessed fall in the category of other transportable goods like wood, paper and
plastic products. Around 20% concern food, beverages, tobaccos, textiles and
apparel. Less than 10% fall in the category ores and minerals. Not surpris-
ingly, these numbers show that the majority of the cases are on products with
high value (see Table 3.3).
Table 3.3: United Nations classifications for Product (see http://esa.un.org)

1-Ores and minerals;
2-Food products, beverages, tobaccos; textiles, apparel and leather products
3-Other transportable goods, except metal products, machinery and equipment (e.g.
wood, paper and plastic products)
4- Metal products, machinery and equipment
The majority of the cases are from Europe. In fact, we report on 1 case
from South America ([79]), 2 cases from Asia ([46]; [134]), 17 case reports
from North America , ([16], [22], [76], [86], [95], [124], [128], [127], [129],
[130], [175], [185], [189], [209], [215] [249], [256]) and more than 40 cases from
Europe (note that some cases relate to more than one geographic area). The
unequal distribution of cases over geographic areas corresponds to the uneven
past development of reverse logistics research and practice in the different
continents. Next, we discuss the cases listed by subject.
3.4 Case studies on Reverse Logistics Networks

A main activity in reverse logistics is the collection of the products to be
recovered and the redistribution of the processed goods (see Section 3.2). Al-
though this problem apparently resembles the standard forward distribution
problem, there are also some differences. There are usually many points from
which goods need to be collected. In other words, the one-to-many structure
that characterizes the forward distribution of goods becomes a many-to-one
instead (Fleischmann et al., 2000). Besides this, cooperation of the sender is
much more needed and the goods tend to have a lower value than in forward
distribution. As reverse logistics is quite new, in many cases new networks
need to be constructed. Major issues in this respect are the determination
of the number of layers in the network, the number and location of depots
or intermediary points, the use of drop points in the collection, the issue of
integrating the reverse chain with the forward chain and finally the financing
of the network (see also Fleischmann et al., 2004).
3.4.1 The case studies

We report on 26 case studies, which are summarized in Table 3.4.
Below we discuss the cases organized according to the recovery option.
This is a common way of organizing the discussion of reverse logistics networks
(see Fleischmann et al., 1999).
The most often described recovery option is direct recovery, either re-
distribution or re-sale (12 cases), followed by recycling (11 cases, of which 8
describe public networks and 3 private networks), and then remanufacturing
(3 cases).
Networks for re-distribution/re-sale (12 cases)
Containers, packaging and refillable bottles are typical items which can be
re-used without extensive recovery operations. In the industrial market these
items are exchanged between several companies. Three cases were found.
Kroon and Vrijens [166] discuss the design of a logistics system for reusable
transportation packaging. The issues addressed are the role of all the parties
in the system, the economics of the system, the amount of containers needed
to sustain the system, the cost allocation to all parties and the locations of
the depots of the containers.
Del Castillo and Cochran [76] study the production planning, product
distribution and collection of re-usable containers and apply it to re-usable
bottles at a soft drink company in Mexico.
Table 3.4: Case Studies on Reverse Logistics Networks
Reference Product(in) Recovery Product(out) Sender Collector Processor Future Organizer/ Return Drivers
option customer Respons. reason
[166] dist. items re-distribut. (product-in) customers Nedloyd Nedloyd same market functional economics
(Europe)
[76] distribut. (product-in) retailers EMSA EMSA original EMSA functional economics
(N.A.) items chain
[86] distribut. re-distribut. distribut. Canada Canada Canada Canada Canada functional economics
(N.A.) items items Post Post Post Post Post
[65] distr. items re-distribut., distr. items supermarkets food chain food chain supermarkets food chain functional, economics
(3 cases, other mater. recycling materials specialist end-of-life
Europe)
[65] cons. goods re-sale cons. goods stores stores store stores stores stock adj., economics
(3 cases, dist. items stores chain chain chain chain chain functional
Europe) other mater. end-of-use
[65] cons. goods re-sale cons. goods customers mail order mail order customers mail order reimbursement economics
(3 cases, company company company company (legislation)
Eur.)
[164] ind. goods remanufact. (product-in) local filial Oce Oce same & new Oce end-of-use economics
(Europe)
[186] ind. & cons. remanufact., (product-in) households, specialized Canon Canon Canon end-of-life, economics
(Europe) goods recycling materials companies 3rd parties (F & UK) (F & UK) (F & UK) end-of-use (image)
[80] spare parts remanufact. spare parts customer return return same IBM service economics,
(Europe) centre centre customer legislation
[17] chemicals recycling other mater. households, municipalities, specialized Organiz. of end-of-life legislation
(Europe) (batteries) companies retailers, companies - importers (+incentives)
[67] cons. goods: recycling materials households municipalities specialized companies Organiz. of end-of-life legislation
(Europe) white goods retailers companies electronics (+ no dis-
3.4. Case studies on Reverse Logistics Networks
(NL) posal costs)

[274] dist. items recycling other mater. households, specialized specialized glass Glass end-of-life legislation
(Europe) (glass) (glass) firms players players industry Foundation
[263] other mater. recycling other mater. car certified selected glass Car end-of-life legislation
(Europe) (car wrecks) owner disassemblers recyclers industry Foundation economics
(NL)
[15] other mater. recycling other mater. waste Recyclers Recyclers Dep. of Recyclers end-of-life legislation
(Europe) (construction) (sand) processors consortium consortium Public consortium
Works
[157] other mater. recycling vs. materials households, opportunistic specialized pulp consortium end-of-life legislation
(Europe) (paper) incineration or energy businesses players industry Governm. economics
[46] other mater. recycling materials households public public - Kaohsiung’s end-of-life (corporate)
(Asia) (waste) authority authority municipality citizenship
[178] cons. goods recycling fibers, households, municipalities, professional road, dam carpet end-of-life economics
(Eur. & (carpets) materials companies specialists organization builders industry (legislation)
N.A.)
[209] cons. goods recycling other mater. business carpet Dupont Dupont end-of-life economics
(N.A.) carpets nylon fibres customers dealers
[234] other mater. recycling other mater. steel steel steel steel & other industry manufacturing corporate
(Europe) steel scraps (steel) industry industry industry industry government by-products citizenship
[234] other mater. recycling materials households demolition specialized industry government end-of-life corporate
(Europe) (construction) players recycler citizenship
85
Duhaime et al. [86] discuss the collection and distribution of returnable

containers for Canada Post. Again the inventory balance between the different
locations is a major problem. The last two cases also have production planning
and control issues and are also discussed in Section 3.6.
De Koster et al. [65] describe return handling operations of three food re-
tailers, three department chains, and three mail-order companies (see Chapter
5).
Networks for remanufacturing (3 cases)
Remanufacturing is typically applied to complex equipment or machinery
with many modules and parts. It is usually a labor-intensive activity, which
requires much testing. Fleischmann et al. (2000) makes a further distinction
into networks set up by the OEM and by independents, as in the latter case
there can be no integration with the forward chain. We found the following
cases.
Krikke et al. [164] discuss the remanufacturing of photocopiers. The
authors consider two options for the remanufacturing facility, one coinciding
with the manufacturing facility and one in a cheap labor country. They
evaluate the costs of both options, including the transportation effects.
Meijer [186] discusses the remanufacturing of used scanners, printers,
copiers, faxes at Canon.
Dijkhuizen [80] discusses the remanufacturing network of IBM. He deals
with the problem of where to re-process the products: in each country, or
centrally at one place in Europe.
Networks for Recycling (11 cases)
One can distinguish two types of networks depending on the organizer/ re-
sponsible party, viz. private and public. Fleischmann et al. (1999) and
Goggin and Browne (2000) have also made this kind of differentiation while
classifying networks for recovery.
Public Networks (7 cases)
There are several papers describing the set-up and organization of recycling
networks in The Netherlands.
Bartels [17] describes the Dutch nationwide system for the recycling of
batteries. Since 1995, it is legally required that all Dutch battery manu-
facturers and importers have the batteries they sell retrieved and recycled.
Stichting Batterijen (Stibat), was put together to make possible for Dutch
battery manufacturers and importers to fulfil these legal requirements. Stibat
is a non-profit organization, of which batteries manufacturers and importers
3.4. Case studies on Reverse Logistics Networks 87
are members. The Stibat Plan describing how batteries are collected and
processed was approved by the Dutch ministry of Environment (VROM).
De Koster et al. [67] deal with recycling of white and brown goods. White
goods are large consumer appliances, like washers, dryers, and refrigerators.
Brown goods are small domestic appliances, such as a video recorder. The
authors also describe the details of network for end-of-life large white goods in
De Koster et al. (2003). The consumer pays a disposal fee when purchasing
the good, and the sector organization is responsible for “distributing” the
money among the parties involved.
Van Notten [274] explains the glass recycling system in the Netherlands.
The rules are legally established by the Dutch Ministry for the Environment
(VROM). The glass industry formed a Glass Recycling Foundation to deal
with them. The authors discuss the bring and pickup systems for the collec-
tion of glass scrap from households.
Van Burik [263] describes the car-recycling nation-wide scheme in the
Netherlands. Again a society, Auto Recycling Nederland (ARN) was raised
by the car manufacturers, importers, and other players in the network of the
car industry to address car recycling. A tax on new purchased cars, paid by
the consumer to the car importer is used to pay for the recycling. Discarded
cars are first assembled at car dismantling centers. Resulting materials are
collected and centrally recycled. Recycling companies are paid for processing
volumes.
Barros et al [15] discuss the design of a network for the recycling of sand
case that comes free during the sieving of building waste, in the Netherlands.
Sand recycling in large-scale infrastructure projects, e.g. road construction,
is considered a potential alternative to land fill and furthermore aligned with
environmental legislation. The Department of Public Works is responsible for
the destination of the recycled sand, since it coordinates large-scale infrastruc-
ture projects. As the Dutch government aims to increase the recycling quota
of building waste from 70% to 90%, a consortium of construction waste pro-
cessing companies is interested in improving efficiency in the sand-recycling
network. The main problem tackled in the paper is the determination of the
number and location of depots of the sand.
Kleineidam et al. [157] consider the structure of the recycling network of
the paper industry in the Netherlands. Paper recycling in the Netherlands is
bounded by the rules put down by the Dutch Ministry for the Environment
(VROM). Companies selling or importing paper and cardboard in/to the
Dutch market are members of the Dutch Corporation of Paper Recycling.
This corporation is in charge of the control mechanisms. The recycling system
resembles a close-loop, with used paper being collected, processed, and turned
into pulp, which is raw material for the paper industry. Paper from households
is mainly collected by a network kept up by non-profit organizations. The
authors investigate the dynamic behavior of the chain and use it to evaluate
the system. In the analysis, both incineration costs and paper taxes are
considered.
Chang and Wei [46] discuss the recycling network for household waste of
the city of Kaohsiung in Taiwan, Waste management is maneuvered by the
Environmental Protection Bureau of the city. This bureau has the challenge
to have in place an effective planning recycling program. The authors deal
with the allocation of the recycling drop-off stations.
Private Networks (4 cases)
Louwers et al. [178] discuss the set up of a carpet recycling system. It
concerns a special type of carpets, of which the output is used as feedstock
in the chemical industry. Both the organization and the collection network
are discussed. Realff et al. [209] discuss a similar network, using the same
technology, but now in the U.S.
Spengler et al. [234] discuss two cases, one for recycling building debris
and one for the recycling of by products in the German steel industry. The
authors consider the effect of various variables, like the cooperation between
companies.
3.4.2 Preliminary insights

The outline of the previous section is organized around two dimensions of the
framework, as follows: the how and the who (see Chapter 2). In more detail,
we presented cases for
• Networks for re-distribution/re-sale,
• Networks for remanufacturing,
• Networks for Recycling - Public Networks,
• Networks for Recycling - Private Networks.
Typologies of the how and the who, also have been (more or less con-
sciously) employed in previous literature to distinguish reverse logistics net-
works (see Fleischmann et al., 1999; Goggin and Browne, 2000; Fleishmann
et al., 2000; and Fleischman et al., 2004). Thus, the how and the who are two
prominent elements when one characterizes reverse logistics networks. We
come back to this on Section 3.9.
3.5. Case studies on Reverse Logistics Relationships 89
In general, within reverse logistics networks, the models to determine the

location of facilities are based on deterministic integer programming models
(Fleischman et al., 2004). However, it is commonly accepted that reverse
logistics is characterized by much more uncertainty than in forward logistics
(see Dekker et al., 2004). With this in mind, there is space for stochastic
programming modelling with respect to reverse logistics network design (see
e.g. Listes and Dekker 2001).
3.5 Case studies on Reverse Logistics Relationships

It is typical for reverse logistics that several parties are involved, such as the
sender of the product, the collector, the processor and the organizer/initiator
(see Chapter 2). To stimulate/enforce a certain behavior of their partners,
parties in the reverse chain make use of various incentive tools. In this subsec-
tion we discuss our findings with respect to the tools that are used in practice
to stimulate/enforce a desired behavior of partners in the context of product
recovery. A summary of the cases is provided in Table 3.5.
A distinction should be made between two categories of incentives to stim-
ulate/enforce a certain behavior of other players in the supply chain:
1. incentives to acquire products for recovery, and

2. incentives to force others to accept the products a company wants to
get rid of.
To be more concrete, two examples follow: in the first situation, a producer

of toner cartridges may be interested in getting back its empty cartridges; in
the second situation, a company buying chemical raw materials in kegs, may
wish that a supplier takes back these kegs (avoiding in this way potentially
high disposal costs, once the kegs are empty). Accordingly, these companies
are looking for incentives to make “their wishes come true” but against the
lowest costs.
The cases embrace incentives regarding functional returns, service returns,
end-of-use returns and end-of-life returns.

We have found ten cases illustrating the employment of eight different incen-
tive tools, as follows:
1) Deposit fee: This fee may concern the product itself or the distribution
item like a bottle, box, and pallet. An example of the former is the deposit
Table 3.5: Case Studies on Relationships

Reference Product(in) Recovery Product(out) Sender Collector Processor Future Organizer Return Drivers
option customer reason
[16] cons. goods remanufact., (product-in) customers U.S. customers Unysis end-of-use economics
(toner (parts) of Unysis postal of (+ price) (customer
(N.A.) cartridges) retrieval service Unysis relations
[85] ind. goods remanufact., ind. goods owners Mercedes Daimler- owner Daimler- service economics
Chrysler
(Europe) (car refurbishing of Mercedes- -Benz deal- Chrysler of Mercedes- (Mercedes- (trade-in) (+ customer
ers
engines) -Benz engine Benz cars Benz) relations)
[94] Chemicals recycling other mobile specialists customers City of end-of-life
mater.user
(batteries) of batteries units, for raw Leicester (+ info.)
(Europe) stores, malls other mater. (UK)
[95] other mater. recycling (100% supplier of Kayak’s recreational Walden end-of-life corporate
(N.A.) (plastic (+ DfX, recycled) consumers recycled manufacturer kayak Paddlers citizenship
scraps) innovation) kayak resins market
[129] cons. goods remanufact. (product-in) customers U.S. customers Xerox end-of-use economics
(toner of postal of (+ price)
(N.A.) cartridges) Xerox service Xerox
[130] cons. goods remanufact. (product-in) cellular Recellular Recellular end-of-use economics
(cellular airtime (specialist) (specialist) (+price,
(N.A.) phones) providers ...)
[185] cons. goods recycling other mater. users of UPS several HP HP end-of-use economics
(N.A.) (toner toner specialists (+donation (+ to put
cartridges) cartridges by HP jobbers out
to WWF) of market
[277] dist. items re-distribut. (product-in) distribution Campina Campina Campina Campina functional economics
(Europe) (PC bottles, centres of (+ deposit
crates) supermarkets fee)
[282] other mater. recycling other mater. demolishers, 3rd party building RockWool RockWool end-of-life economics
(Europe) (rock wool) building logistics industry (+buy) (+ green
suppliers industry provider back) image)
[284] Chemicals recycling Chemicals households, UPS, several end-of-life
(Europe) (batteries) companies retailers, specialists (+ price)
municipalities corporation
90
fee that has to be paid when renting a car, whereas an example of the latter
is the deposit fee on certain bottles from glass and PET bottles (functional
returns), see e.g. Vroom et al [277].
2) Buy back option: At the moment that a product is sold, the buyer is
offered the possibility to sell back the product on a later moment in time.
Some conditions can be imposed, such as kilometers driven.
For instance, Ford has a program called Options, which gives the client
some advantages in exchanging his/her car, after two years, for another new
car Ford (end-of-use returns).
Rockwool Lapinus, the Dutch subsidiary of the Danish Rockwool Com-
pany, gives the client the opportunity to return rock wool for free or for a
lower price than he/she would else have to pay to get rid of it (end-of-life
returns), see Wijshof [282]. Numerous examples of buy back options for un-
used products (commercial returns) are presented in the literature, where no
explicit attention is paid to what happens to these products thereafter (see
e.g. Tsay (2001)).
3) Trade-in: One can only get a new product if a used one is returned.
This incentive is among others used by Daimler-Benz for the engines that
they produce for Mercedes-Benz passenger cars and small vans (Driesch et al,
[85]). Owners of a Mercedes-Benz (MB) passenger car or small van with an
MB engine can go to an authorized MB dealer in order to have their present
engine replaced by a reconditioned engine (service returns). The MB dealer
removes the engine and sends it to the central parts Distribution Center (DC)
of the center of Daimler-benz motor recycling (DB-MTR). From this DC, a
reconditioned engine is sent to the dealer where the engine is available within
24 hours. The MB-dealer puts the reconditioned engine in the MB passenger
car or van, after which its owner can use the car again.
4) Acquisition price: This is paid to the sender/giver when he or she
delivers a product for recovery. Some suppliers for toner cartridges, including
UNISYS, deliver their cartridge in a box that can be returned for free to
them either by post (Bartel [16]) or via another third party logistics service
provider like Hewlett Packard or Xerox do (see McGavis [185], Guide and
van Wassenhove, [129]). Recellular, a U.S. cellular phone remanufacturer
is also very active in setting prices to buy used mobile phones in the B2B
environment (see Guide and van Wassenhove [130]). These are cases of end-
of-use returns.
Also manufacturers of batteries use this type of incentive, see Yender [284].
Another example is Varta, the German battery manufacturer, that pays 50p
in the UK for every returned rechargeable battery (end-of-life returns) send
to a collection point (see Faria de Almeida and Robertson [94]).

5) Supplying timely and clear information about the recovery program:
How important this incentive is, is illustrated by a pilot system for the col-
lection of different types of batteries in Denmark and Germany (end-of-life
returns), where it turned out that it was too difficult for a number of suppliers
to understand clearly which type of battery they have (see Faria de Almeida
and Robertson [94]).
6) Power : As always, power can be used to force desired behavior. An
example is New England Foam of Windsor. One of the customers of this
company, Walden Paddlers uses her power as a customer to force her supplier,
New England Foam of Windsor, to take back their cardboard boxes (end-of-
life returns), see Farrow and Johnson. [95].
7) Environmental responsibility: The idea is to appeal to the consumer’s
environmental responsibility. This incentive is likely to require a lot of ad-
vertising, and its reliability is questionable, as is illustrated by the collection
of toner cartridges (end-of-use returns), by Hewlett Packard (see McGavis
[185]).
8) Social responsibility: The idea is to appeal to charity donors. Basically,
it works as follows: for each product received for recovery, a non-profit orga-
nization receives an amount of money. This incentive was used by Hewlett
Packard in the beginning of the nineties but it did not result in a satisfy-
ing number of returns (McGavis, [185]). A drawback of this incentive is
that the company is going to accept the products without any guarantee of
quality. This also applies to the previous tool, i.e. appealing for the con-
sumers’environmental responsibility.
Very recently, this incentive was revived by Recellular (see Recellular,
2003) in the U.S. and by the retailer chain Tesco in the U.K. (see Tesco,
2003), to acquire used mobile phones. All these are examples of end-of-use
returns.
3.5.2 Preliminary Insights
Among the ten cases, there are quite different incentive tools:
1) Deposit fees (Vroom et al.[277]),
2) Buy back options (Wijshof [282])
3) Trade-in (Driesch et al, [85]),
4) Acquisition price (Faria de Almeida and Robertson [94], Bartel [16],
McGavis [185], Guide and van Wassenhove [130], [129]; Yender [284]),
5) Timely and clear information (Faria de Almeida and Robertson [94]),

6) Power (Farrow and Johnson [95]),
7) Environmental responsibility (McGavis [185]), and
8) Social responsibility (McGavis [185])
In the beginning of this section we distinguished two type of incentives:

1) incentives to acquire products for recovery, and
2) incentives to force others to accept the products a company wants to
get rid of.
Actually only one among the ten cases found was on the latter type of
incentive. This is the case of Walden Paddlers using its power to force its
supplier of foam to take back the cardboard boxes (see Farrow and Johnson
[95]). Besides this, not all the other cases are examples of incentives exclu-
sively dedicated to acquire products for recovery. Many of the tools are also
meant to keep customers. One example is the buy back option offered by
Ford (www.ford.nl), when the customer buys another new Ford after 2 years.
Another example is the case of the engine trade-in option offered by Daimler
Chrysler (Driesch et al [85]). Other tools are not directly coupled to a selling
activity and they can involve more than just own customers. For instance,
a gift to a non-profit organization (see McGavis [185]) as used by Tesco to
get back used mobile phones. Or, deposit fees on beverages, where customers
from one shop may return the empty bottles in another shop. Actually, the
latter seems to be the only incentive that is specific for product recovery.
Other incentives can serve other objectives than recovery alone.
Next, we make some observations regarding the incentives in the cases and
the return reason why-returning. The cases do not report on manufacturing
returns. These are usually handled internally, or they are not a matter of
acquisition anyway. A similar argument can be used for the absence of stock
adjustments. None of the cases include product recalls, as in this situation
safety is a key-issue thus the parties involved have to take their responsibility.
Regarding B2B and B2C commercial returns, these are documented in the lit-
erature on contracts for unused products with a return option (see Anupindi
and Bassok, 1999; Corbett and Tang, 1999; Debo et al., 2004; Tsay et al.,
1999; Tsay, 2001; Lariviere, 1999). The cases presented here embraced incen-
tives to acquire products from functional, service, end-of-use and end-of-life
returns (see Table 3.6). The table shows that for several return reasons, it
seems to exist different incentives to choose from.
Although all the aforementioned incentives can be found in practice, there
are no models in the literature supporting the choice of incentive. The litera-
Table 3.6: Case studies: return-reason vs. incentive
Cases Return Reason Incentives

Acquisition
[277] Functional Deposit fee
[85] Service Trade-in
[16] End-of-use Acquisition Price
[185] End-of-use Acquisition Price, Environ-
mental Responsibility & So-
cial Responsability
[282] End-of-life Buy back option
[284] End-of-life Acquisition Price
[94] End-of-life Acquisition Price & supply-
ing timely and clear infor-
mation
Disposition
[95] End-of-life Power
ture deals with one or more aspects of those mentioned above. For instance,
1) literature on sales contracts with return options for unused products; and
2) research on the optimal acquisition price to realize a certain flow of prod-
ucts. The prime focus of the first group of literature is on what may be
gained by both sellers and customers by allowing customers to order more
under certain return options. Usually a fixed sales price is assumed for the
products that are taken back by the seller. Besides this, most of the literature
on contracts with a return option concern unused products (see Anupindi and
Bassok, 1999; Corbett and Tang, 1999; Debo et al., 2004; Tsay et al., 1999;
Tsay, 2001; Lariviere, 1999 [171]).
Klausner and Hendrickson (2000) present a mathematical model that
might be used for estimating the acquisition price. Guide et al (2001) present
a mathematical model to determine the optimal acquisition price for products
from the field as well as the selling price of these products. In the above two
models the time aspect is neglected, i.e. a steady state situation is considered.
This aforementioned literature can be used as a starting point to construct
a model to support the choice of which incentive to choose, under which
conditions.
Another research opportunity is searching explanatory variables with re-
spect to the choice of incentive. To start with, we suggest mapping who is
3.6. Case studies on Inventory Management 95
who in each network. This is because an important factor in choosing between

incentive tools is to be informed on the alternatives that other players have
and the costs associated with them. Therefore, the type of players and their
interrelations are likely to determine the type of incentive tools employed.
3.6 Case studies on Inventory Management

The 14 cases on inventory management within reverse logistics are given in
Table 3.6. They can be classified according to the return reasons (see Section
3.2). We did not find cases for all the return reasons. We did find cases
for functional returns, B2C commercial returns, service returns, end-of-use
returns and end-of-life returns. Omissions can be explained as follows. Man-
ufacturing returns are often treated in production planning contexts. Product
recalls are often special events, which are left out of consideration in inven-
tory management. Warranty returns have similar characteristics as repairs.
They differ mainly in the contractual side and accordingly one cannot expect
many publications with an inventory management focus. Stock adjustments
are somewhat similar to some of B2B commercial returns (bulk returns) and
might even not be distinguished from them in the case descriptions. Below
we go over the cases in more detail.
Functional returns (2 cases)

This category corresponds to distribution items, that is, products like con-
tainers, bottles, railcars and crates, which are used for distribution purposes.
On this we report two cases, viz. Swinkels and van Esch [242] and Del Castillo
and Cochran [76]. Here the location of the distribution items is a major issue
in the inventory decision.
Del Castillo and Cochran [76] study production and distribution planning
for products delivered in reusable containers. Their model includes trans-
portation of empty containers back to the plants. Availability of empty con-
tainers is modelled as a resource constraint for the production of the original
product. The model is applied to a case study of a soft drink company using
returnable bottles.
Swinkels and Van Esch [242] describe how the optimal stock of refillable
beer kegs is determined within Bavaria, a Dutch beer brewery.
Table 3.7: Case Studies on Inventory Management

[59] ind.goods re-use ind.goods internal customer CERN internal CERN reiumbursement,economics
(Europe) customers (bring-back) customers end-of-use
[59] cons. goods re-sale cons. goods customers mail order mail order customers mail order reimbursement economics,
(Europe) company company (same mar- company legislation
ket)
[59] spare parts repair, spare parts service service Refinery internal service economics
(Europe) (refinary) re-use engineers engineers customer
[76] distribut. re-distribut. (product-in) retailers Coca-Cola Coca-Cola original Coca Cola functional economics
(N.A.) items chain chain
[79] spare parts repair spare parts Caracas Caracas Caracas Caracas Caracas service economics
(S.A.) (railways) subway subway subway subway subway
[82] spare parts repair spare parts telephone Lucent Lucent same Lucent service economics
(Europe) (telephones) companies Technologies Technologies chain Technologies
Nederland Nederland Nederland
[101] ind. goods repair, ind. goods business IBM end-of-life economics
(Europe) (IBM) refurbish. spare-parts customers etc.
[154] cons. goods remanufact., (product-in) customers specialized specialized manufacturers end-of-life economics
(Europe) (power recycling other mater. party facility (pro-active)
tools)
[192] ind. goods repair, ind. goods UK UK UK UK UK service economics
(Europe) (aircraft en- refurbishing Air Force Air Force Air Force Air Force Air Force
gine)
[222] cons. goods reuse, (product-in) users Norwegian TAC or same Norwegian end-of-use economics,
(aid equip.) refurbish., Technical specialized market National service corporate
Aid
(Europe) retrieval,... Centers player Insurance citizenship
(TAC) Administra-
tion
[223] cons. goods re-sale cons. goods customer 3rd party Wehkamp same Wehkamp reimbursement economics
(Europe) log. provider (catalogue) market (legislation)
[242] distribut. re-distribut. distribut. business Bavaria Bavaria Bavaria functional economics
items items
(Europe) (beer kegs) customers agents Bavaria
[256] cons. goods remanufact. cons. goods consumer photo Kodak same Kodak service economics
(single-use shops, chain
(Europe, photo other
N.A.) cameras) retailers
[266] spare parts remanufact. spare parts Importer Volkswagen Volkswagen Importer Volkswagen service economics
(Europe) (cars) organizations (in Kassel) (in Kassel) organizations (in Kassel)
96
3.6. Case studies on Inventory Management 97
Commercial returns (3 cases)

Commercial returns occur in a B2B or in a B2C setting (see Section 3.2),
where the buyer has a right to return the product, usually within a certain
period. In the B2B setting, there are all kind of contracts lowering the risk
of e.g. the retailer by giving him the opportunity to return what is not sold.
In this situations, returns are likely to be in bulk and at the end of a season.
In the B2C setting, consumers are given, or they have by law, the right to
return a product if the product does not really meets his/her expectations.
Sanders et al. [223] describe how the inventories of products are controlled
within Wehkamp, a Dutch mail order company, selling all kinds of consumer
goods to the Dutch and Belgium market. Two types of products are dis-
tinguished: seasonal and non-seasonal products. For the first, the company
employs an amended version of the newsboy model taking into account re-
turns. For the latter, the inventory management is done according to an (R,
S) policy.
De Brito and Dekker [59] investigate the distribution of the return lag, i.e.
the time between the purchase and the return of an item, and its consequences
for inventory management (see Chapter 6). Two cases of commercial returns
are considered, viz. a mail order company and and the warehouse at the
center for nuclear research, CERN.
Service return (6 cases)
Within service systems like repair systems returns occur basically in two ways.
First of all, the products themselves may be brought or sent to a center for
repair. If the repair is successful, they are brought back, else, a new product
or system needs to be bought and the one that failed is discarded. Secondly,
if one needs a continuous functioning of the product or system, one may
directly restore functionality by replacing a spare part. The failed part is
then repaired, after which it will enter the inventory of spare parts. The cases
found are described below in detail.
Díaz and Fu [79] study a 2-echelon repairable item inventory model with
limited repair capacity. For several classes of arrival processes they develop an
analytic expression for the number of items in queue at the different stages of
the system. They analyze the impact of the capacity limitation and compare
the performance of their approach with an uncapacitated METRIC type of
model. Both models are applied to the case of spare parts management at
the Caracas subway system.
Donker and Van der Ploeg [82] describe how the optimal stock of reparable
service parts of telephone exchanges is determined within Lucent Technolo-
gies Netherlands. They use an amended METRIC model, where the service
measure is fill rate (i.e. the percentage of demand that can immediately be
fulfilled from stock) and there is no budget restriction for service parts.
Moffat [192] provides a brief summary of a Markov chain model for ana-
lyzing the performance of repair and maintenance policies of aircraft engines
at the Royal Air force.
Van der Laan [266] describes the remanufacturing chain of engines and
automotive parts for Volkswagen. It is somewhat similar to the engine re-
manufacturing case with Mercedes Benz in the previous section (see Driesch
et al. [85]).
A special example of service returns is the one described by Toktay et
al. [256] on Kodak’s single use camera. Customers return it, so they can
develop the film. Printed circuit boards for the production of these cameras
are either bought from overseas suppliers or remanufactured from the cameras
returned by the customers via photo laboratories. The issue is to determine
a cost-efficient order policy for the external supplies. Major difficulties arise
from the fact that return probabilities and market sojourn-time distribution
are largely unknown and difficult to observe. The authors propose a closed
queuing network model to address these issues. They assess the importance
of information on the returns for the control of the network.
De Brito and Dekker [59] investigate the distribution of the return lag, i.e.
the time between the purchase and the return of an item for a spare parts
warehouse at a petrochemical plant.
End-of-use returns (1 case)
This return reason concerns items that are only temporarily with the user.
The product may e.g. be leased or rented.
Rudi et al. [222] discuss the product recovery actions of the Norwegian na-
tional insurance administration. This public entity retrieves no longer needed
wheel chairs, hearing aids and similar products provided to people with hand-
icaps. They assess how many are needed to meet all demands.
End-of-life returns (2 cases)
Fleischmann [101] describes the dismantling of computers at the end of their
life-cycle into useable spare parts with IBM. This source nicely combines
with return obligations and it is a cheap source for spare parts for systems on
which one does not want to spend too much. The problems identified were a
lack of knowledge of the content of the computers, as well as, an insufficient
information system to handle the operations.
Klausner and Hendrickson [154] develop a model to determine the optimal
buy-back amount to guarantee a continuous flow to remanufacturing power-
3.7. Case studies on Planning and Control of recovery activities 99
tools. The authors apply the model to the actual voluntary take-back program
in Germany.

We have grouped the presentation of the cases according to the why-returning
typology, i.e. according to the return reason as follows:
• Functional returns,
• Commercial returns,
• Service returns,
• End-of-use,
• End-of-life
This is because this seems a natural way of grouping and discriminating

the reverse logistics issues raising from each inventory system. Other authors
have done roughly the same (see Dekker and van der Laan, 2003).
Many have defended that product data are essential for efficient handling
of returns. For instance, Kokkinaki et al. (2004) provide an example of the
value of information for disassembly. Other authors have investigated the
impact of data, with respect to returns, on inventory management perfor-
mance (Kelle and Silver, 1989 ; Toktay et al., 2000; De Brito and Van der
Laan, 2003). Yet, there is still room to model the impact of having a priori
information on what can be recovered, i.e. on which parts are likely to be
recoverable. In practice, techniques to forecast return behavior would have
to be enriched with broader explanatory variables. We refer to Toktay (2003)
for a discussion of factors influencing returns, which are potential explanatory
variables in advanced forecasting models.
Concluding, it remains to be investigated in which degree inventory sys-
tems’ characteristics like timing, quality and degree of control, are determined
by the return reason why-returning, as well as on the type of product (what).
3.7 Case studies on Planning and Control

of recovery activities
This subsection deals with the planning and control of the recovery activities,
i.e. collection for recovery and the recovery itself.
Part of the planning and control of product recovery concerns the planning
and control of supply of goods to be recovered in which incentives play an
important contextual role. This important aspect has already been dealt with
in subsection 3.5. In this subsection the incentives and their effect on supply
and acceptance are assumed to be given.
Planning and control of recovery activities is strongly related to inventory
management, the topic of the previous section. The latter includes the lev-
els triggering recovery activities. Hereafter it is assumed that the inventory
strategy has already been decided on. What are left are decisions concerning
the planning of activities, such as on lot sizes, re-processing priorities and
scheduling.
The 23 case studies on planning and control of product recovery activities
that we found in the literature (see Table 3.8) can be subdivided into case
studies dealing specifically with:
1) collection for recovery (13 cases), and
2) recovery itself (9 cases).
The cases are succinctly presented in Table 3.7.
Collection for recovery (13 cases)

In a number of case studies the lot size used for collection is given, usually
without explaining how these lot sizes have been determined. All the case
studies that we found concern end-of-use or end-of-life returns.
Andriesse [6] describes the Packaging Return System of Philip Morris for
reusable pallets and sheets used to cover pallets. Philip Morris Holland BV
and most of its suppliers agreed that the lot size for returning reusable pallets
should be a full truck.
Del Castillo and Cochran [76] describe the model used by EMSA, a soft
drink producer in Mexico City, for determining the quantities of refillable
bottles to be returned to the bottling plants. First the final customers return
the bottles to the stores, which return them to the depots.
Duhaime et al [86] present a model that is used by Canada Post to deter-
mine the number of empty containers that should be distributed and returned
each month, as well as the number of containers stored each month per region.
Klausner and Hendrickson [154] mention the lot size used for the collection
of power tools by Robert Bosch GmbH.
Table 3.8: Case Studies on Production, Planning and Control
[6] distr. items re-distribut. distr. items suppliers Philip Philip suppliers Philip functional legislation
(Europe) (pallets) Morris Morris Morris
[14] cons. goods recycling other mater. households, distributors Ortes Ortes end-of-life economics
(Europe) (PVC lamel- (PVC) firms of lamellas Lecluyse Lecluyse (image)
las)
[17] see table 3.4
[22] ind. goods repair, (product-in) NY City, M.-K. NY City, M.-K. service economics
(N.A.) (subway) refurbishing Chic. Tr. Chic. Tr.
[76] see table 3.7
[86] see Table 3.4
[85] see Table 3.5
[124] ind. goods remanufact. (product-in) Air force Air force US Navy Air force Air force
[128] (aircraft overhaul/repair
[127] equipment Depot
[134] other mater. recycling other mater. Producer Producer Producer Producer Producer manufacturing economics
(Asia) (glass scrap) (glass) of glass of glass of glass of glass of glass leftovers
[154] see Table 3.7
[165] cons. goods retrieval, parts & consumers Municipal Secondary Municipal legislation
(Europe) (PC moni- recycling other mater. Waste Com- markets Waste Com- economics
tors) pany pany
3.7. Case studies on Planning and Control
[215] spare parts remanufact. (product-in) Detroit Detroit

Diesel Diesel
(N.A.) (engines)
[225] other mater. recycling gypsum building 3rd party producers customers organiz. of end-of-life legislation
(Europe) (gypsum) companies LSP (gypsum) (gypsum) producers
[229] other mater. recycling (product-in) Trespa 3rd LSP, Trespa Builders Trespa manufacturing economics
(Europe) (wood) (buildings) demolishers leftovers (legislation)
[229] dist. items re-distribut. (product-in) builders, 3rd LSP, Trespa Trespa Trespa economics,
(Europe) (pallets) demolishers demolishers legislation
[234] see Table 3.4
[246] ... chemicals retrieval ... chemicals (manufacturing (manufacturing Schering AG (manufacturing Schering AG manufacturing
(Europe) (pharmaceut.) (pharmaceut.) process) process) (pharmaceut.) process by-products
[243] cons. goods recycling other mater. households municipalities, recycler users of government end-of-life corporate
(Europe) (white & retailers other mater. citizenship
brown)
[249] ind. goods remanufact. ind. goods Pratt&Hitney Pratt&Hitney
(N.A.) aircraft
engine
of recovery activities
[260] dist.items recycling other mater. households, municipalities specialized Metal in- Stichting legislation
dustries
(Europe) (tin plate) (tin) etc. player
[270] other mater. recycling other mater. construction specialized manufacturing corp. citiz.
(Europe) (construction) firms players leftovers legislation
[274] see Table 3.4
[282] see Table 3.4
101
Bartels [17] describes the Dutch nation wide collection and processing of
disposed batteries. Among others, attention is paid to the collection of batter-
ies at municipality collection points. These depots call one of the contracted
collectors to collect the batteries, what has to happen within one month.
Van Donk [270] describes the system set up by Nelis Utiliteitsbouw B.V.,
a Dutch construction company. The system has as objective to raise the level
of reuse of different types of materials by separating them at the building
locations. Among other things, attention is paid to the number and sizes of
containers used for collection. Whenever at a building location it is expected
that a container will be filled soon, a recycling company is called who replaces
the filled container by an empty one.
Van Notten [274] describes the recovery of glass in The Netherlands. The
bring and pickup systems for the collection of glass scrap from households
are discussed, including the sizes of the containers used and the collection
scheme, usually being once a week.
Schinkel [225] describes the Dutch nation wide system for the recycling
of gypsum. Attention is paid to the actual collection of gypsum via special
containers and bags.
’T Slot and Ploos van Amstel [243] describe the pilot project in the region
of Eindhoven, the Netherlands preceding the introduction of a nation wide
system for the collection and processing of discarded white and brown goods.
Among others attention is paid to the collection scheme at households (fixed,
once per month or quarter), and the collection frequency at sales points of
white and brown goods.
Ubbens [260] describes the recovery of metal from metal packaging mate-
rials in The Netherlands. Among others attention is paid to the number and
sizes of special containers for collecting from households.
Wijshof [282] describes the system setup by Rockwool Benelux in the
Netherlands for the collection and processing of rockwool production scrap
and waste, as well as for the rockwool disposed after use. Among others
attention is paid to the sizes of the bags that are used for collection, and the
number of bags that has to be filled before a third party is collecting them
for Rockwool. The disposer has to contact the third party to do this.
Simons [229] describes the recycling system set up by Trespa International
B.V. This company is a producer of sheets made from resins and wood fibers,
which are used in the building industry. The recycling system was estab-
lished to recycle(or alternatively to incinerate) these sheets, or parts of them.
Attention is paid to the collection of leftover sheets at building sites. These
leftovers are put into containers supplied by Trespa. The customer lets Trespa
know when a container is filled. Empty containers replace filled containers
3.7. Case studies on Planning and Control of recovery activities 103
when new Trespa sheets are delivered to the customer. The reusable pallets
used for the distribution of the Trespa sheets to the customers are collected
by third party logistics service providers (3rd LSP’s).
Bakkers and Ploos van Amstel Jr [14] describe the recycling system setup
by Ortes Lecluyse, a Dutch producer of PVC window blinds. Attention is
paid to the sizes of the containers used for collection and to the frequency
for emptying these containers. The containers are located at their direct
customers, and they are collected once a week, when new window blinds are
delivered.
Recovery itself (9 cases)
Two cases have been found on disassembly that deal with the sequence in
which products should be disassembled as well as the method and degree
of disassembly, i.e. the choice between destructive and nondestructive dis-
assembly and the extent of the disassembly. Krikke et al. [165] discuss the
disassembly of PC monitors, where it appears difficult to determine in advance
the level of disassembly. This is due to the heterogeneity of the products to
be recovered. Kobeissi (2001) determines the optimal disassembly plan for
washing machines.
Bentley et al [22] mention that Morrison-Knudsen uses MRP II to plan the
remanufacturing of subway/transit overhaul, but the authors do not explain
exactly how. The latter also holds for Robinson [215] who mentions the use of
MRP by Detroit Diesel Remanufacturing West, that remanufactures Detroit
Diesel engines.
Driesch et al [85] describe the car engine recovery network set up by
Mercedes-Benz For Europe, including the actual planning and control of the
recovery activities in the plant in Berlin. Among other things it is mentioned
that the disassembly, cleaning, test, remanufacturing and reassembly activi-
ties are dealt with in lots, and that the number of engines that are disassem-
bled is related to the number of reconditioned engines that are reassembled.
No further details are given, nor is explained how these lot sizes have been
determined.
Guide and Spencer [124] and Guide and Srivastava [125], [126] discuss
a method for rough cut capacity planning to the Air Force overhaul de-
pot. Guide and Srivastava [127] discuss a method to determine the inven-
tory buffers between the disassembly and the remanufacturing shop, and the
inventory buffer between the remanufacturing and the reassembly shop.
Thomas Jr. [249] mentions that the Pratt Whitney Aircraft remanufac-
turing facility in West Virginia uses MRP to schedule inspection and rebuild
of military and commercial aircraft engines. The batch size is one because
different engines have to go through different routings. The bottleneck is the

engine reassembly. Buffer time is used to protect this activity form variations
in foregoing activities. This time is determined via Linear Progamming, but
no formulas are given.
Spengler et al. [234] discuss an Mixed Integer Linear Programming model
used for the planning of processing components arising from the dismantling
of buildings in the Upper Rhine Valley.
Next there are two case studies where the resources used for production
are also partly used for processing as described.
Gupta and Chakraborty [134] describe the processing of glass scrap gen-
erated during the production of glass. A mathematical model is presented to
determine the optimal production lot size, taking into account the recycling
activities. Teunter et al [246] describe the mathematical model presently used
by Schering AG, a German producer of pharmaceutical products, including
the re-processing of by-products.

The case studies on planning and control of product recovery activities that
we found in literature can be subdivided into case studies dealing with:
• collection for recovery (13 cases)

- separate collection for recovery (Andriesse [6], Del Castillo and
Cochran [76], Duhaime et al [86], Klausner and Hendrickson [154], Bar-
tels [17], Van Donk [270], Van Notten [274], Schinkel [225], ’t Slot and
Ploos van Amstel [243], Ubbens [260], and Wijshof [282]).
- combined collection and distribution (Simons [229], Bakkers and
Ploos van Amstel Jr [14]).
• recovery (9 cases)
- separate recovery (Krikke et al. [165], Bakkers and Ploos [14],
Bentley et al. [22], Robison [215], Guide and Spencer [124]; Guide and
Srivastava [125]- [126]-Guide and Srivastava [127], Thomas Jr, [249],
Spengler et al. [234]).
- recovery combined with production (Gupta and Chakraborty [134],
Teunter et al. [246]).
Quite a number of planning and control concepts for product recovery have
been presented in academic literature. However, the supply of recoverables
3.8. Case studies on IT for Reverse Logistics 105
is often assumed to be autonomous, except for some literature on repair (for

an overview see e.g. Guide and Srivastava, 1997b) and remanufacturing (see
Guide, 2000; Guide et al., 2001; Guide and van Wassenhove, 2001; Minner
and Kiesmüller, 2002). No direct relation between sold/leased and recovered
products is assumed. Besides this, uncertainty has been incorporated only as
far as the arrival of products for recovery and the duration of repair related
activities are concerned. Uncertainty with respect to the result of the pro-
cessing activities has hardly been taken into account (for an exception, see
Souza et al, 2002)
Planning and control is all about deciding on how things are going to be
done and by whom. These are stressed aspects in the case studies. Naturally,
the most relevant elements of the framework for this decision-area are likely
to be precisely the how and the who. This poses a research opportunity. We
come back to it on Section 3.9.
3.8 Case studies on IT for Reverse Logistics

We have found six cases concentrating on various applications of IT for reverse
logistics activities (see Table 3.9). IT is used to support reverse logistics
during different stages of the life cycle of a product, namely manufacturing,
distribution and customer (see Kokkinaki et al., 2004 and Hendrickson et al,
2003).
Manufacturing (2 cases)
Regarding the phase of product development and manufacturing there are two
variables to consider within the what dimension: material content and product
structure. The materials that are used and how they are combined determine
the degree and the type of a potential recovery once the product is at the
end of its life. Marking parts with manufacture identification are also helpful
when a product has to be pulled out of the market due to defect, i.e. in case
of product recalls (see Smith, 1996). Many companies have already in place
product development programs encompassing design for the environment, for
recovery, for disassembly, and so on: generally called as Design for X, or just
DfX. This is the case for Xerox Europe as reported by Maslennikova and Foley
[183]. Xerox has an extensive Design-for-the-Environment program in place.
The design of each new component has to be accompanied with instructions
for the end-of-use.
Recovery can also be the starting point for product development, as is

the case for Walden Paddlers, who launched a 100% recycled kayak project
(Farrow et al. [95]). The project had to rely much on computer experiments
as no available design then suited recycled resins. The company was able to
attract a manufacturer to invest in advanced rotational molding technology
and to convince the supplier to proceed to further resins’ separation.
Distribution (3 cases)
Landers et al. [169] highlight the importance of tracking component’s orders
in the case of a closed-loop business telephones supply chain. The authors
use a concept called “virtual warehousing” where real-time information feeds
expeditious algorithms to support decisions. The use of IT leads to an im-
provement in stock levels, routing and picking processes when compared with
the pre-IT scenario. Xerox (Maslennikova and Foley [183]) uses bar code labels
to track packaging material with the aim of achieving resources’ preservation.
Fraunhofer IML has developed software to embed data on recovery pro-
cesses as reported in Nagel and Meyer [201]. The authors consider two chains
in Germany: the national refrigerator and the computer recycling networks.
Costs could be minimized with the optimization of the location of facilities,
vehicle routing and operations’ scheduling supported by the software. For the
case of the German computer-recycling network, transport volume (in tons
per km) could be reduced by almost 20%.
Customer (4 cases)
After the customer has accepted the product and starts using it, the product
may need maintenance. Xerox has a remote faulty detection system in place
called the Sixth Sense (see Maslennikova and Foley [183]). In some situations,
the problem is remotely identified and solved. Customers are assisted by a
multi-functional database that permits them to get thorough acquaintance
with product characteristics.
In the case of Nortel Networks, a Decision Support System (DSS) was
developed to assist remanufacturing (see Linton and Jonhson, [175]). The
tool permits to apprehend the interrelations between the production and the
remanufacturing of products. By this means both processes can be better
planned and controlled resulting in a more efficient allocation of resources.
Estée Lauder is another firm that has developed specialized software to
handle product returns (see Meyer [189]). The system checks the cosmetic
items for expiration date and damages. In this way recovery-related decisions
are accelerated. The software is linked to an automatic sorting system, which
smoothes away labor costs. Estée Lauder could reclaim the investment on IT
in one year’s time.
Table 3.9: Case Studies on IT
Reference (product- Recovery Product(out) Sender Collector Processor Customer Organizer Return Drivers
in)
option reason
[201] cons. goods remanufact., product(in) end-of-life forward & recovery end-of-life legislation
(Europe) refrigerators recycling materials user specialized plant economics
[201] cons. goods (parts) end-of-life VOBIS Covertronic Covertronic end-of-life economics
(Europe) computers retrieval user (retailer) (specialist) ((specialist)
[189] cons. goods re-sale (product-in) customer Estée Estée same Estée reimbursement legislation
(N.A.) cosmetics Lauder Lauder market or Lauder economics
employees
[95] o. materials recycling (100% supplier of Kayak’s recreational Walden end-of-life corporate
(N.A.) (plastic (+ DfX, recycled) consumers recycled manufacturer kayak Paddlers citizenship
3.8. Case studies on IT for Reverse Logistics
scraps) innovation) kayak resins market

[175] ind. goods remanufact. ind. goods customer customer customer (closed- Nortel service economics
(N.A.) (circuit service service loop) Networks
boards) of Nortel of Nortel
[154], cons. goods remanufact., cons. goods customers specialized specialized manufacturers end-of-life economics
[155] (power recycling o. materials party facility (pro-active)
tools)
(Europe)
[183] cons. goods refurbish., (product-in) customer Service Xerox (same Xerox service economics
(Europe) (Xerox) recycling, Engineers Europe chain) Europe legislation
(+ DfX) (Xerox Eu-
rope)
107
Nagel and Meyer [201] report real software development by the German
recycler Covertronic to read the configuration of a computer and to compute
costs and revenues of subsequent recovery. Based on this, an appropriate
bonus is offered to the final user when the computer is returned. Covertronic
operates this software in tandem with Vobis, a large computer retailer in Ger-
many. Another technology available is the so-called Electronic Data Loggers
(EDL, see also EUREKA, 2003). These devices are able to store data on
physical parameters, which can be retrieved later. The idea is to put them
into products or equipment (as is done for some coffee machines) and to reg-
ister information about heat or other parameters as they are used. Thus, at
the point of recovery, one could make use of this information to decide which
destiny to give to a certain product without first investing resources in disas-
sembling and testing components. Klausner et al. [155] have investigated the
benefits of collecting information via this chip technology in electric motors.
Likewise Simon et al. (2001) apply both steady state and transient models to
evaluate the benefits of using a data logger in washing machines.

Table 3.10 summarizes the IT tools to support reverse logistics activities de-
scribed in the case studies presented above.
The case studies illustrate IT applications in all the phases of the life
cycle of a product and show improvements for reverse logistics. E.g., Xerox
(Maslennikova and Foley [183]) has an integrated solution for reverse logistics
from product development to recovery or proper disposal. Furthermore, all
the case studies provide in one way or another an evaluation on the benefits
of IT. However, the investments on such technologies or on gathering infor-
mation are not reported, except for the Estée Lauder case (Meyer [189]) and
partially for Walden Paddlers (Farrow and Jonhson [95]). The real benefit
of IT investment is therefore difficult to assess as the case studies stress the
benefits while the investments and operational costs are left out. In face of
limited investment capacity, it would be helpful to know in which phase the
investment would have most earnings. In addition, if alternative technologies
are available, one can investigate which one is the best. To do so, one has
to take into account costs and benefits of collecting and managing data and
costs of investing and managing the technology.
The following literature is an excellent starting point, as the potential ad-
vantages or bottlenecks of some type of information are partially analyzed:
Kelle and Silver (1989), Inderfurth and Jensen (1999), Klausner et al., (1998)
Simon et al. (2001), De Brito and Van der Laan (2003), Ferrer and Ketzen-
3.8. Case studies on IT for Reverse Logistics 109
Table 3.10: Case studies: IT tools, requirements and benefits for reverse
logistics
IT Tool Information’s Type of support Life cycle phase

requirements
EDL for electric mo- Info. on potential Reuse decisions Customer
tors reuse cost savings) through
[154], [155]
DSS for end-of-use Info. on operations’ Cost optimization, Customer
costs and recycling facilities location,
revenues vehicle routing, etc.
[201] Distribution
Computer’s configu- Info. on operations’ “Setting buy-back Customer
ration reader costs &recycling rev- price”
enues
[201] Distribution,
Software specialized Product’s expiration Recovery-related de- Customer
on return handling data, damage check cisions
[189]
DSS for In-depth Remanufacturing Customer
remanufacturing information on
[175] processes -related decisions
DfX, remote main- Recovery options, en- Extensive data-base All phases
tenance, etc. vironment’s sustain- on products
(X=Recyclability) ability, and so forth.
[183]
DfX Further separation of Developing a 100% Manufacturing
(X=Recyclability) resins; recycled
[95] (& recyclable) Kayak
Figure 3.1: IT support for Reverse Logistics in different phases of the life-cycle
of the product.
3.9. Overall insights 111
berg(2003), Kiesmuller (2003), and [254]Toktay (2003).

Besides investments, IT tools are very demanding regarding data on re-
verse logistic processes and associated costs and earnings. This data is how-
ever often not available. Nagel and Meyer [201] declare that the lack of
information is a bottleneck, which complicates the management of recycling
systems. Furthermore, on the Nortel Networks case the DSS could not be de-
signed as desired due to a shortage of data on customer’s returns (see Linton
and Jonhson [175]). Besides this, to acquire and to manage the data is very
expensive.
3.9 Overall insights

The framework presents a structuring and classification of reverse logistics. In
our opinion the typologies in the framework are not only useful for describing
reverse logistics but also to establish relations between attributes of reverse
logistics. Below we present some findings.
Table 3.11 presents the cases organized by return reason (Why-returning)
and the recovery option (How ). No cases of recalls were found. This is
not surprising, since these cases are not typically documented in Operations
Management literature.
From the above table we make now some observations and remarks about the
return reason and the recovery option.
- Among the return flows that may occur during manufacturing, there are
three cases of manufacturing leftovers, and two cases of manufacturing
by-products. They are all associated with materials or chemicals and
therefore the corresponding recovery option is recycling.
- The B2C commercial returns are associated with re-sale and with direct
re-use. The latter is a form of internal returns, as the customers belong
to the same organization as the supplier of the products.
- Functional returns, by definition, are re-distributed. This actually justi-

fies the element ‘re-distribution’ in the recovery option typology (How)
that we suggested in the framework of Chapter 2.
- Service returns are mainly associated with repair, and to a lower degree
refurbishing, and remanufacturing. Let us recall that service returns
occur in the first place, because customers need it to be serviced. This
usually means ‘restore it to working order’.
Table 3.11: The cases: return reasons vs. recovery options

Return Reason Recovery Option Total The cases
End-of-life Recycling 23 [14], [15], [46], [65] (3
cases), [69], [94], [95],
[154]–[155], [157], [178],
[165], [186], [201], [225],
[209], [234], [243], [263],
[274], [282], [284],
Remanufact. 2 [154]–[155], [201]
Retrieval 1 [201]
Incineration 1 [157]
Service Repair 7 [22], [57], [79], [82], [101],
[124]–[127]–[128], [192]
Refurbishing 5 [22], [85], [101], [183],
[192]
Remanufact. 5 [80], [85], [175], [256],
[266]
Reuse 1 [57]
Recycling 1 [183]
Functional Re-distribution 11 [6], [65] (3 cases), [76],
[6], [86], [166], [229],
[242], [277]
End-of-use Remanufact. 6 [16], [85], [130], [123],
[164], [186]
Retrieval 3 [16], [85], [165]
Recycling 1 [17]
Reuse 1 [57]
B2C commercial re- Re-sale 5 [65] (3 cases), [223], [189]
turns
Re-use 1 [57]
Manufacturing Left- Recycling 3 [134], [229], [270]
overs
Manufacturing Recycling 2 [234], [246]
by-product
3.9. Overall insights 113
- End-of-use returns appear with various options, although remanufac-

turing is the more dominant one. This is not surprising, since typically
there is enough value to recover to make remanufacturing attractive.
- In contrast, end-of-life returns are highly correlated with recycling. This
is not surprising, as at the end of the product’s life, there is less value to
recover than in earlier stages. Please note, however, that if products or
their modules would be designed for remanufacturing, this option would
likely be more attractive, even at the end of life.
Table 3.12: The cases: drivers vs. return reasons

Driver Return Reason Total The cases
Economics Recycling 16 [201], [155]-[154], [178],
[185], [186], [183], [209],
[263],[282], [157], [65] (3
cases), [225], [134], [14]
Remanufacturing 10 [16], [85], [123], [130],
[164], [154]–[155], [175],
[186], [256], [266]
Redistribution 9 [65] (3 cases), [76], [86],
[166], [229], [242], [277]
Repair 6 [22], [57], [79], [82], [101],
[192]
Re-sale 6 [57], [65] (3 cases), [189],
[223]
Refurbishing 5 [22], [85], [101], [183],
[192]
Retrieval 2 [16], [201]
Re-use 1 [57]
Legislation Recycling 12 [15], [17], [69], [157],
[165], [183], [201], [225],
[260], [263], [270], [274]
Re-sale 3 [57], [189], [223]
Re-distribution 2 [6], [229]

Retrieval 1 [165]
Remanufacturing 1 [80]
Refurbishing 1 [183]
Corporate Recycling 5 [46], [95], [234], [243],
[270]
Citizenship
Table 3.12 exhibits the cases organized by driver vs. recovery option.
- Economics is associated with all recovery options, of which recycling

and remanufacturing are the most documented ones. This shows that
in principle all recovery options have the potential of being economically
viable.
- Legislation appears to be mainly associated with recycling. Some of

the cases are associated with re-sale. In those cases, legislation gives
customers the right to return purchased products. The above raises
the question whether environmental legislation is not able to stimulate
other forms of recovery. As we stated before, if design for recovery
would be more common, other recovery options than recycling would
likely be more attractive. Concluding, there is indeed an opportunity
to stimulate other forms of recovery through legislation. For instance,
legislation could also pro-actively focus on product design issues, rather
than dealing with recovery in a reactive way.
- There are not so many cases where the driver is corporate citizenship,
but all of them appear consistently associated with recycling. They can
be divided into community services and private initiatives.
For each of the decision-making areas, such as Networks, Relationships,

etc., some of the dimensions of the framework (see Chapter 2) are predomi-
nantly important. These were mentioned at the end of each section under the
preliminary insights. These dimensions are more or less consciously empha-
sized in the case studies presented here. We believe that this is because these
dimensions predominantly affect the organization of reverse logistics within
each of the discussed areas. We summarize the predominant dimensions per
decision-making area in Table 3.13.
To further investigate these relations, we suggest a dedicated case study
analysis. In such research, one may also consider to add typologies per dimen-
sion or to go into more depth regarding sub-characteristics of the typologies
in the framework. Please recall that the framework was constructed to be
parsimonious and to facilitate a holistic view of reverse logistics. The last
column of Table 3.13 exhibits the kinds of elements that could be consid-
ered in more detail, depending on the decision-making area. Furthermore, we
also put forward what kind of variables can be added in relation to a given
typology.
3.10. Summary and research opportunities 115
Table 3.13: The kind of elements to be considered in more detail
Decision making area Dominant dimensions Additional elements

Networks How & Who Value of the product;
Relationships Who The type of relation be-
tween the parties; Degree of
control in acquisition;
Inventory Management Why-returning & What Why-returning + used vs.
not; What -characteristics
(in detail);
Production planning How & Who How-processes (in detail);
Degree of involvement
of the sender (bring and
pickup systems); combined
with forward, vs. not;
3.10 Summary and research opportunities

We have found more than sixty cases involving reverse logistics. Roughly
60% of the cases are in the manufacturing category, about 20% are within
wholesale and retail trade and about 10% in construction.
Almost half of the cases deal with metal products, machinery and equip-
ment. Around 30% of the products being processed are materials like paper
and plastic. Around 20% concern food, beverages, tobaccos, textiles or ap-
parel. Less than 10% fell in the category ores and minerals. Not surprisingly,
these numbers show that the majority of the cases are on products with high
value.
With this review we
1. give a substantial overview of the diversity of real life reverse logistics

situations, and
2. provide a reference guide for researchers that look for case support.
We observed a lack of cases on reimbursement returns. However, the

growth of the catalogue industry including e-catalogues is posing new chal-
lenges. Ergo, field research in this area is likely to grow as well.
The analysis of the cases was based in the framework for reverse logistics
proposed in Chapter 2. Thus, for each case study we answered the following
questions:
- what products/materials are being returned?

- why are they returned?
- why is the company involved in recovery?
- how are the products being recovered?
- who are the parties involved?
This review was also a means to improve the quality of the framework (Chap-
ter 2). Both projects were carried out simultaneously and in an interactive
way. This means that on the one hand, the case studies review is a means
to ground the framework on empirical evidence, and on the other hand, the
framework is a means to further analyze the case studies. The analysis of
the case studies allowed us to establish relations between attributes of reverse
logistics. We come back to this later in this section.
We presented the case studies according to the following decision-making
focus: Network Structures, Relationships, Inventory Management, and Plan-
ning and Control. We remarked particular issues and latent questions, for
which we proposed topics for future research. For each group, some of the
aforementioned questions gain a dominant role, as summarized below. Fur-
thermore, there was an overview of Information and Technology (IT) for re-
verse logistics.
Reverse Logistics Networks
The section on Reverse Logistics Network Structures was organized around
the typologies on the how and the who (see Section 2). In more detail, we
presented cases for
• Networks for Re-distribution/Re-sale,
• Networks for Remanufacturing,
• Networks for Recycling - Public Networks,
Here, most of the cases fell under ‘Re-distribution’ and ‘Recycling - Public
Networks’. Only a few cases relate to private networks, but at this time we
do not know to which degree private networks for recycling are successful.
Reverse Logistics Relationships
The cases report quite different incentive tools to stimulate/enforce a certain
behavior of other players in the supply chain (see Section 3.5):
• Deposit fees
• Buy back options
• Trade-in
• Acquisition price
• Timely and clear information
• Power
• Environmental responsibility
• Social responsibility
Some incentives are also part of sales contracts, while others require the cus-
tomer to buy another product in exchange. There are also tools that are not
directly coupled to a selling activity like a gift to a non-profit organization.
Furthermore, it seems that only deposit fees are specific for product recov-
ery. The other mentioned tools are also used to attract or keep customers in
general.
Inventory Management for Reverse Logistics
We have grouped the presentation of the cases according to the why-returning
typology. The cases covered:
• Functional returns,
• End-of-use,
• End-of-life
We did not find cases for all the return reasons. Omissions can mainly be
explained, because some return reasons are treated in different contexts, like
production planning (see Section 3.6).
One of the complicating factors in Inventory Management is the uncer-
tainty in the quality of reverse flows. Many have defended that product data
are essential for the efficient handling of returns. This offers various research
opportunities (see the next section).
Planning and Control for Reverse Logistics

We found case studies on Planning and Control of product recovery in the
following categories:
• collection for recovery

- separate collection for recovery
- combined collection and distribution
• recovery
- separate recovery
- recovery combined with production
In the literature, the supply of recoverables is often assumed to be autonomous,

except for some literature on repair and remanufacturing. Besides this, un-
certainty has been incorporated only as far as the arrival of products for
recovery and the duration of recovery related activities are concerned. Uncer-
tainty with respect to the result of the processing activities has hardly been
taken into account. In practice, however, there is a lot of yield uncertainty.
IT for Reverse Logistics
We found cases reporting evidence that IT can be used in the following stages
of the life-cycle of a product:
• Manufacturing
- dfX
- labels
• Distribution
- software
- track & tracing
• Customer
- software
- decision support systems
The technology to process and transmit information seems to be available

with promising benefits for reverse logistics. The cases show, though, that
the lack of appropriate data is still the bottleneck in the implementation of
decision support systems.
For each of the above decision-making areas, some of the dimensions of the
framework (see Chapter 2) are predominantly important (see Sections 3.4–
3.8). We summarize the predominant dimensions per decision-making area as
follows
• Networks ⇔ Recovery option (How ) and Actor (Who)
• Relationships ⇔ Actor (Who)

type (What)
• Production Planning ⇔ Recovery option (How ) and Actor (Who)
In the next section we give directions to explore this further.
Besides the analysis per decision-making area, we also presented the cases
organized by return reason (why-returning) vs. recovery option (how ), and the
cases organized by driver vs. recovery option (see Section 3.9). This analysis
revealed some relations and issues. For instance, end-of-life returns appear
to be highly correlated with recycling. If products or their modules would
be designed for remanufacturing, however, remanufacturing would likely be
more attractive, even at the end of life. The main driver for recycling is
legislation. Actually, current legislation does not seem to be able to stimulate
other forms of recovery. However, if legislation would also pro-actively focus
on product design issues, rather than dealing with recovery in a reactive way,
other recovery options than recycling would likely be more stimulated.
3.10.1 Research opportunities

If one wants to get more in-depth details, build-up theory, or explain phenom-
ena, one has to go into more depth regarding some aspects of the framework.
In Table 3.13 we put forward the kind of elements that can be considered
to further analyze the prominent dimensions of the framework per decision
making area. In order to do this, a dedicated case study analysis should be
conducted. Next we continue with specific research opportunities regarding
the various decision making areas.
Reverse Logistics Networks
Commonly, the models to determine the location of reverse logistic facili-
ties are based on deterministic integer programming models. However, it is
commonly accepted that reverse logistics is characterized by much more un-

certainty than in forward logistics. With this in mind, there is room for other
tools like stochastic programming modelling for reverse logistics network de-
sign.
Reverse Logistics Relationships
We pointed out the lack of literature supporting the choice of incentive. For-
tunately, there is plenty of literature that can be used as a starting point.
For instance, 1) literature on sales contracts with return options for unused
products, and 2) research on the optimal acquisition price to realize a certain
flow of products.
Inventory Management for Reverse Logistics
Knowledge about returns is seen to be essential for efficient inventory man-
agement of returns. With respect to modelling, it is critical to question the
validity of the common assumptions in the literature (see Chapter 6). Fur-
thermore, it is important to investigate the impact on inventory management
of having a priori information on product returns (see Chapter 7). Moreover,
specialized forecasting techniques should be developed to deal with this issue.
Planning and Control for Reverse Logistics
In the literature, uncertainty with respect to recovery yield has hardly been
taken into account. In future research, yield uncertainty should be mod-
elled explicitly. Furthermore, return handling operations have been so far
overlooked in the scientific literature (see Chapter 4). On this, Chapter 5
provides insights regarding splitting vs. combining forward and reverse flows
during transport, actual handling and the storage.
IT for Reverse Logistics
Almost all the cases that focus on IT report on the benefits only, while dis-
regarding the required investment. It is important to broaden such partially
sighted reports.
In face of limited investment capacity, it would be helpful to know in
which phase the investment would have most earnings. In addition, when
alternative technologies are available, one can investigate which one is best.
To do so, one has to take into account the costs and benefits of collecting and
managing data and the costs of investing and managing the technology.
The majority of the case studies deal with one aspect of a real reverse
logistics situation but they do not give the overall business environment, which
makes insights rather one-dimensional. Thus, there is a need for conducting
more integral case study research, by mapping the business context together
with more broad information on critical factors, trade-offs and implications.
Besides this, the lack of theory for reverse logistics (see Dowlatshahi, 2000)
or even for supply chain management (Croom et al., 2000) adds to companies’
inability of knowing what matters in reverse logistics. Therefore the devel-
opment of theory should be on top of the research agenda to support reverse
logistics decisions.
Chapter 4
Return Handling:
decision-making and
quantitative support
Consistently wise decisions can only be made

by those whose wisdom is constantly challenged.
Theodore C. Sorensen
4.1 Introduction
Return rates can be very high and very costly if not handled properly (see
Meyer, 1999 and Morphy, 2001). In spite of this, some managers have dis-
regarded product returns or they handle returns extemporarily (see Meyer,
1999). The fact that quantitative methods barely exist to support return
handling decisions adds to this. In this chapter those issues are bridged, as
we
• put together a decision-framework for return handling, and
• identify models to support the decision-making process.
With respect to the second item, we often depart from existing quanti-
tative models on general material handling. This is because of the lack of
quantitative methods dedicated to return handling.
123
124 Chapter 4. Return Handling: decision-making and quantitative support
The remainder of the chapter is organized as follows. First, Section 4.2

gives a brief review of the return handling operations. Section 4.3 provides an
illustrative example. Subsequently, in Section 4.4 we put forward a decision-
making framework for return handling. After that, Section 4.5 identifies op-
portunities to support warehousing decisions with quantitative models. Sec-
tion 4.6 draws some overall conclusions.
This research was reported in De Brito and De Koster (2004).
4.2 Return Handling

Different actors in the supply chain face various return flow types (see the
framework in Chapter 2). Independently of the return flow type, the following
warehousing processes can be distinguished (see Figure 4.1):
• Receipt,
• Inspection, sorting & other handling,
• (Interim/Stock) Storage,
• Shipping.
Here we do not consider the last process in detail, since shipping is similar
to forward flows. In general, operations can differ per return type (see Chapter
2 for a description of return processes and operations). For instance products
that come back in an as-good-as-new state can be restocked to be sold again
(see De Brito and Dekker, 2003), while end-of-life products may only need
interim storage until they are sold to a recycling company.
Figure 4.1: Return handling in the warehouse.

4.3. Return Handling: an illustrative case 125
Material handling is among the most recent research issues within reverse
logistics, which in itself is a relatively new topic. Therefore, material handling
aspects (e.g. receipt, and storage) have not been dealt with in the literature
in great detail.
Yet, the coercion of re-use and recycling quotas in Europe (see EUROPA,
2003) has served as impetus for specific research on subsequent consequences
for return handling. For instance, Anderson et al. (1999) have investigated
how firms are coping with EU legislation for packaging recovery and Fernie
and Hart (2001) have considered the required assets to carry out the European
packaging regulations.
Furthermore, very recently, De Koster et al. (2002) investigated and dis-
cussed the return operations of nine retailers. This research is also reported
in Chapter 5.
4.3 Return Handling: an illustrative case

This section illustrates the return handling process throughout the operations
of a catalog company, here named CTC. Similar operations hold for mail
order companies in general (see also Chapter 5). We employ the framework
of Chapter 2 to generally structure the description of the case.
What, why, who and how?

The assortment consists of thousands of consumer goods, as fashion and hard-
ware articles. There are two selling seasons per year with different assort-
ments. The peak in demand is during the first weeks of the season, when
the new catalogue is out. Return rates in catalog business are usually large
(see Meyer, 1999). For CTC they vary from a few percent (mainly warranty
returns for furniture and large appliances), to larger percentages for small
appliances, and to much larger for fashion.
A main driver behind the returns is home-shopping legislation (consumer
rights legislation). Therefore, CTC allows customers to return products.
Apart from legislation, accepting returns at no cost to the customer is seen
as an important service element (improved customer relations).
Telephone operators sometimes stimulate customers to order more fashion
products than they really need (e.g. by suggesting to order the item in mul-
tiple sizes). They do this to increase the probability that the customer will
keep at least one of the items. This will however cause the customer to return
the sizes that do not fit. This is not per se a drawback if one understands the
following arguments. Return transportation and handling costs are free to
the customer. This also holds for the delivery cost, unless the customer buys
at least one item. Thus, to suggest multiple sizes may still lead to lower costs
for CTC than when the customer places multiple orders of one unit each,
keeping only one of them.
The home delivery of customer orders is outsourced to a Third Party
Logistics Provider (3PL) with a very dense network. When a customer wants
to return an item, CTC informs the 3PL, which schedules the pick-up in a
route carried out in the area of the customer (who-actors). If such a route is
not planned yet, the pick-up is delayed, or the customer can decide to bring
it to a nearby post office and send it back herself. Pick-ups go back again via
the sorting centers with large trucks to CTC. The same trucks are used for
pick-ups of the new orders at CTC’s warehouse.
The return handling operations of the warehouse of fashion and small
appliances is further described in more detail. The focus is on collection,
inspection and recovery (storing for re-sale) (how-processes).
• Collection of returns at customers has been integrated with the distribu-

tion of products to customers. The same third party logistics company
carries out both processes and the route planning is based on both types
of orders, although return collection has a slightly lower priority. Even
the return drop-offs at CTC are integrated: the truck that brings re-
turned items also collects new orders for distribution.
• Returned products are inspected and the receipt is confirmed in the

information system, where it is decided whether the client should be
credited or not (based on the condition of the product, payments made,
and date of original purchase). The products are labelled, which permits
to trace the number of times each one of them were returned. Such a
label is however disguised so the customer cannot detect whether, or not,
she is receiving a product that was once returned by another customer.
• Returned products with a ‘as-good-as-new’ status are re-sold. They are

first consolidated with new products on a location in a forward pick area,
whereas newly purchased items are stored in a reserved area. Other than
‘as-good-as-new’ products are either returned to vendors or to brokers.
Considering additional decisions made by CTC to control the returns, here

follow some observations:
• The return handling process is completely separated from the regular

receipts of suppliers, with dedicated workstations and handling equip-
4.4. Return Handling: a framework of decisions 127
ment like conveyors. For administrative handling, a tailor-made soft-

ware module is used.
• The forward storage differs from the regular receipts. For regular re-
ceipts, one product carrier (usually a pallet) consists of one product.
Returns on the other hand are regularly brought to the storage area in
rolling bins.
• In the picking process, no explicit priority is given to returned items.
4.4 Return Handling: a framework of decisions

In the last section, the company CTC served to illustrate some of the re-
turn handling related decisions. This section formally addresses the decision-
making process according to:
• long-term decisions
• medium-term decisions
• short-term decisions
Table 4.1 provides an overview of the decision-making process. One should

note that the decision topics have many interdependencies. In practice, when
deliberating over a decision, the whole set of decisions should be kept in mind.
4.4.1 Long-term decisions

In supply chain management, the design of the supply chain is often con-
sidered to be a strategic, or long-term, decision (see Ganeshan et al., 1999).
Analogously we consider here that the design of the warehouse is a long-
term decision for return handling. Intimately related to design decisions is
the outsourcing decision, which is therefore also considered to be a long-term
one.
Facility Layout & Design

Facility layout and design is essentially a long-term decision regarding ware-
housing. Companies have to ensure that sufficient storage and handling ca-
pacity is going to be allowed for return handling and, additionally, how the
space in the facility is going to be organized. For instance, companies have
to decide if a facility dedicated to return handling is preferable, or, if returns
are going to be handled in a separate area of the warehouse facility. By in-

tegrating return and forward handling, resources can be shared on the one
hand, but on the other hand handling complexity increases.
In the illustrative example, returns were handled in the same facility, but
in a separate area dedicated to returns. Actually, this is observed in retailers
of other industries as well, as reported in the comparative study of Chapter 5.
From the analysis there, a threshold on return volume is likely to determine
whether, or not, returns are allocated to a separate area: above the threshold
it is more efficient to handle returns in a dedicated area. In practice, one even
finds companies with a separate facility dedicated to returns, such as Quelle
in Germany, Sears and Kmart in the US, which all have to deal with huge
amount of returns (Chain Store Age, 2002).
Fleischmann et al. (2004) further explains how the decision of opening
a separate facility for return handling is affected by the amount and type of
returns. An obvious trade off is between investments and capacity for return
handling. In addition, the location of a return handling area within a facility
affects the internal handling systems and therefore the related costs.
Concluding, return handling should be explicitly taken into account during
the design of the warehouse.
Outsourcing
In the illustrative case, CTC keeps the warehousing operations in-house. In
this case, outsourcing would be difficult since the activities take place within
the warehouse and are closely related with order picking and storage activities.
A determinant factor is whether, or not, the return process can be clearly
separated from the forward process. If so, return handling may be out-
sourced. Examples include Albert Heijn (the return warehouse in Pijnacker,
the Netherlands) and Sears and Kmart in the US, where the returns are han-
dled by Genco (see De Koster and Neuteboom, 2001; Chain Store Age, 2002).
Other determinant factors are the value and type of product as well as the
availability of experienced third parties. At this moment such companies are
emerging. Examples include Genco and Universal Solutions Incorporated (see
e.g. Chain Store Age, 2002). The advantage of outsourcing is that one can
focus on its own competence and let other matters to proper professionals.
On the other hand, when outsourcing, a company will always lose a certain
degree of control over the return handling operations.
Long-term decisions
• Facility layout & design
- physical facilities for return handling
• Outsourcing
- Outsourcing of return handling operations (transportation and warehousing)
Medium-term decisions
• Integrated operations
- return policy (rules and responsibilities)

- reusable packaging
• Inventory Management
- storage policy of returned products
• Internal (return) transportation
- type of distribution items (product carriers)

- automatization of vehicles
- type of vehicles
• Information Systems
- IT systems to support the return handling

- type of information
Short-term decisions
• Inventory, storage and order picking
- controlling return storage

- storage of returns vs. order picking
• Vehicle planning and scheduling
- route selection
Table 4.1: A framework of decisions for return handling.

4.4.2 Medium-term decisions

Often, medium-term decisions concern integrated operations, transportation,
and information systems (see Ganeshan et al., 1999). In addition, we make
a distinction between inventory management (medium-term) and inventory
control (short-term). The latter is considered in the next section.
Integrated Operations
Here we consider mainly two sets of decisions,with respect to i) return policy,
and ii) reusable packaging.
Actors in the chain are assigned with responsibilities for returns. In the
case of chain stores, either a central facility (often the distribution center)
collects and processes the products and materials that have been returned
from the stores, or the stores take care of it themselves.
According to IGD (2002), in the UK all major food retailers recycle pack-
aging materials in a central facility: either a distribution center or a separate
recycling unit. Actors in the chain also have to establish whether different
types of returns incur different levels of accountability. For instance, the man-
ufacturer takes the responsibility for all the processes related with end-of-life
returns but no responsibility for reimbursement returns.
Another issue to decide upon is the rules for the return policy. For in-
stance, stores have to know which products can or must be returned to the
warehouse, in which situations, or how to grant permission for this. In the il-
lustrative example, the mail-order-company does not have much freedom with
respect to the return policy, as it is a legal requirement to allow for returns.
The urge to reduce packaging materials leads to the decision whether,
or not, packaging materials are going to be replaced by reusable packages.
Reusable packages will demand a higher initial investment, but have the in-
herent benefit of the repetitive use. Many industries choose to reduce pack-
aging waste by using standardized boxes (Schiffeleers, 2001), or by replacing
packaging materials by other materials involving less material handling.
Some retailers opt to start recycling their own packaging materials. Albert
Heijn, for example, recycles wrap foil into plastic bags, integrating the forward
and reverse flows for a practical purpose. Yet, this demands changes in its
operations as follows. Stores have to separate the wrap foil, to put it in roll
cages, and store it until collection. After recycling by a third party, bags are
re-sent to the warehouse to be integrated with the distribution.
Inventory Management
Returned products may have a quality status that differs from new ones.
Depending on the status and the timing, they may be, or not, sold in the
same market. Both newly supplied and as-good-as-new returned merchandize
are going to be stored for future sales as long as the season runs and taking
into account the shelf life of the product. The two streams can be stored
separately or the merchandize can be consolidated in the same location.
In the case of CTC, returned items are stored together with new mer-
chandize. In the comparative analysis of Chapter 5, other retailers than mail-
order-companies have mixed and hybrid storage policies. For instance, one
retailer would first store returns at a separate location and consolidate them
only when the location’s capacity would be exceeded. The analysis led to the
conclusion that separate storage policies are mainly found among retailers
that wish for a high degree of control over returns. Returns can be consoli-
dated if the future market is the same as for newly supplied merchandize. The
‘future market’ is also a critical factor for other returns than the ‘as-good-
as-new’ ones. Unsold products are usually consolidated in interim storage
locations per vendor or potential ‘customer’. A well-known example is unsold
books that go back to their respective publishers. End-of-life returns are also
usually separated and held together by potential brokers.
Internal (return) transport

An important issue in internal transport is the choice of reusable transport
carriers. Choosing the type depends on the willingness of the parties in the
chain to adopt one of the available standards. This is usually a complex pro-
cess, where power and trust play a major role. Once a standard is in place,
the decision that is left to take is the amount needed. Product carriers require
collection, transportation back to the warehouse, checking, storage, and possi-
bly cleaning before being used again. In order to limit the amount of product
carriers needed, they should rotate rapidly. Since the timing and quality of
such returns are difficult to anticipate, many companies have searched for
ways to reduce such uncertainty.
Carriers can also leak from the system. A way to monitor this is to
invest on tracking and tracing systems. Chapter 3 reviews incentive tools
to persuade parties behaving in a desired way. A common incentive is to
charge a deposit fee between the different parties in the chain, especially if
the material has some intrinsic value. Each receiver must pay this fee to its
supplier or a central party (see Lüstrenbach, 1993 for a review of systems
for returnable carriers). Often, when multiple companies participate in such
networks, there is a central organization that tracks the ownership of carriers

and registers financial transactions. The deposit fee not only prevents items
from being lost but it also provides a natural mechanism to motivate careful
handling and therefore a minimum quality is ensured per returned item, as
transactions have to be tracked.
Several international pools for pallets exist in Europe, like Deutsche Bahn
(DB) and Chep. Many manufacturers supply their products on such pallets.
In the case that the reusable packaging can be shared between different users,
the benefits can sometimes be huge, since large savings can be achieved in the
numbers needed (Koehorst et al., 1999). For this reason, in the Netherlands
and in the UK food retailers have switched to reusable container systems
(IGD, 2002). In food retailing, reusable crates are mostly used for product
categories such as produce and chilled, where warehouse stocks are small and
sales volumes are high. This means the crates are used intensively, reducing
the cost per trip. However, wholesalers and retailers with many suppliers
and different pallet types have considerable work in handling, sorting and
storing pallets returned from their customers (retail organizations or stores).
In the case of large return volumes, mechanization or automation may become
economically attractive. Choices to be made are the degree of mechanization
and the appropriate equipment for internal transport and storage.
Information Systems
One has to determine which information system is going to be used to register
product returns. Kokinakki et al. (2004) examine the overall needs and ex-
isting supporting technology for managing information in the reverse logistics
context.
Commercial software particularly designed for supporting return handling
is lacking (Caldwell, 1999). The commonly used ERP packages generally
lack the ability to properly deal with returns (De Kool, 2002). At this stage
decision-makers may consider the in-house development of dedicated software.
This has been the case at CTC and Estée Lauder (see Meyer, 1999). In the
latter case, the specialized software system checks returns for expiration date
and damages, speeding up return handling. Besides this, the software is linked
to an automatic sorting system, which has lowered labor costs.
In general, the type and the horizon of information to be registered has
to be determined, as well as how decisions with respect to returns can be
supported. Also attention can be given to potential abusive returns (see
Schmidt et al, 1999) and how this affects warehousing operations.
In the case of CTC, though law enforces accepting returns, returned mer-
4.5. Models and research opportunities 133
chandize is first checked and only then it is decided whether or not the client
should be credited for it.
4.4.3 Short-term decisions

Inventory, storage, order picking, and vehicle planning and schedul-
ing
When storing returns, this has to be planned in coordination with forward
operations such as order picking and internal transport. In the case of CTC,
it is crucial to put returns back in inventory timely to prevent stock-outs.
Another aspect is the organization of the picking process. In the case of CTC
no explicit priority is given to returned items over regular stock. There are,
however, cases where priority to returns might be given (see Chapter 5).
The former discussion put together a decision framework of long-, medium-,

and short-term warehousing decisions with respect to return handling.
Long-term decisions embrace facility design and outsourcing. At a more
tactical level, one finds the topics of integrated operations, inventory manage-
ment, internal (return) transport, and information systems. Within a shorter
horizon, one has also the control of inventory and order picking, and the
planning and scheduling of vehicles for internal transport.
The discussion tackled one topic at a time, but the complete framework
has to be taken into account at any decision instant. Subsequent decisions are
circumscribed by previous decision-making. E.g., the layout of a warehouse
restricts the possibilities regarding order picking and storing. Or, if part of
the warehouse is likely to be outsourced, the facility design should take this
into account.
The remainder of this chapter brings into focus the available and potential
models to support the decision-making process. To do so, we use the decision-
framework to go over the existing models, and to identify opportunities for
extensions. The decision-framework is also used in Chapter 5 to structure
general recommendations to practitioners (see Section 5.6.1).
4.5 Models and research opportunities

Quantitative models on return handling are practically non-existent. This
section puts forward the state-of-the-art on quantitative support for material
handling expounding opportunities for return handling models. An early
overview of the use of OR tools in material handling can be found in Matson
and White (1982).
The list of the main research areas in warehousing with respect to return
handling can be drawn from the previous section (see also Van den Berg and
Zijm, 1999). For information systems for return handling, see Kokkinaki et
al., 2004. Models for the remainder of the topics are now discussed. For a
more detailed discussion, see De Brito and De Koster (2004).
4.5.1 Long-term decisions

Facility layout and design
A successful research area is facility layout. A well-known approach is Muther’s
(1973) Systematic Layout Planning concept, for which many improvement
algorithms are known (see Tompkins et al., 1996; Bozer et al, 1994; and
Goetschalckx, 1992). The major objective in the underlying models is to
minimize the total cost of daily transport between the different areas in the
facility. For some facility types special layout models have been developed
with the reduction of travel time as primary objective (Bartholdi and Gue,
2000; and Roodbergen, 2001).
An important shortcoming in the models is that many other relevant re-
strictions and objectives are not taken into account, such as congestion re-
duction. The models and solution methods can be applied straightforwardly
to facilities with return flows and handling areas. The existence of dedicated
areas to returns do not have to change the models fundamentally. However,
it is important to incorporate returns operations from the start in layout
decisions, since the solutions may change dramatically when compared to a
situation without returns (think of crossing flows, or travel distances). It
would be interesting to study the impact that taking into account returns has
on layout. This is a topic that has not received attention from researchers so
far.
Outsourcing
Literature on process outsourcing is mostly of a qualitative or quantitative em-
pirical nature. Although much literature deals with outsourcing of warehous-
ing and transport to logistics service providers (see for example Rabinovich
et al., 1999; Van Laarhoven et al., 2000), academics have not thoroughly
addressed the outsourcing decision of return handling so far. Among the ex-
ceptions are the works by Krumwiede and Sheu (2002) and Meade and Sarkis
(2002). In the first, a model is presented to aid third-party logistics providers
to decide whether, or not, to enter the reverse logistics market. Meade and
Sarkis (2002) address the selection of third-party logistics providers. The au-
thors claim that the factors to take into account when selecting a third-party
reverse logistics provider, differ from traditional factors for supplier choice.
Table 4.2: Research opportunities for long-term decisions

Research opportunity Supporting literature
To extend the models for SLP models: Tompkins et
Facility layout and de- Systematic Layout Plan- al., 1996; Bozer et al.,
sign ning (SLP) 1994; Goetschalckx, 1992;
Bartholdi and Gue, 2000;
Roodbergen, 2001
To intensify research in Krumwiede and Sheu,

Outsourcing this area 2002; Meade and Sarkis,
2002
Table 4.2 summarizes the research opportunities for long-term decisions

and the literature that can be of support.
4.5.2 Medium-term decisions

Integrating operations: return policy
Distant sellers and other retailers usually accept merchandize back up to
a number of weeks after delivery or purchase. CTC puts the threshold at
about two weeks and does not charge the client for returns. In practice, we
find various charging schemes including partial refundable costs. Schemes to
charge returns and some of the associated dilemmas have been studied in the
literature, like how to control opportunistic returns (see Hess et al., 1996).
Researchers have established relations between return policies and a number
of elements, among which: salvage value (Davis et al., 1995 and 1998, Emmons
et al., 1998), mismatching probabilities (Davis et al., 1995; Hess et al., 1996),
speed of consumption (Davis et al., 1998), product’s value to the consumer
(Davis et al., 1995; Hess et al., 1996), and product’s quality (Moorthy and
Srinivasan, 1995; Wood, 2001).
However, factors like the maximum contractual return period have not
always been explicitly involved in the discussion. In environments where the
customer is entitled by law to return merchandize free of charge (see Office of
Fair Trading, 2003), the freedom the retailer has is on fixing the contractual
return period together with the shipping and returning charges, and price, as
follows:
• t, maximum contractual return period,
• p, price of the product,
• csc , shipping charges, and
• crc , returning charges.
The utility function of the retailer, here named Ur , is a function of addi-

tional factors, like:
• ch , handling/warehousing costs,
• crh , return handling costs,
• x, quantity shipped, and
• y, quantity returned,
For the retailer, the decision variables are t, csc , crc and p, while ch , crh are
input variables, and x and y are output variables.
The utility functions of customers i= 1,..., m, here identified as ui will
depend of factors, such as:
• p, price of the product,
• t, maximum contractual return period,
• q, quality of the product,
• csc , shipping charges,
• crc , returning charges.
Naturally, customers will mostly prefer low prices, a long return period,
high quality products, and low shipping charges. Based on these factors, each
customer i decides on
• xi , the quantity ordered by client i, and
• yi , the quantity returned by client i.

All the actors involved intend to maximize their utility function. The
retailer is the leader as he/she sets a priori some of the parameters, such as t,
p, csc and crc . In turn, customers will react to the values of these parameters
and maximize their utilities.
The problem can be written in the following general form:
maxp,t,csc ,crc Ur (p, t, csc , crc ; y, x, ch , crh )
s.t.
y = summi=1 argmax ui (yi ; p, t, xi , q, csc , crc )
x = summi=1 argmax ui (xi ; p, t, yi , q, csc , crc )
xi ≤ yi
Summing up, for a given strategy that the retailer may choose, customers
will have a ‘best’ strategy to follow. The retailer has to anticipate these
reactions while setting the decision variables.
The aforementioned model gives a holistic approach to return policies by
giving plenty of room to incorporate multiple critical factors. The format of
the problem resembles a Mathematical Programming problem with Equilib-
rium Constraints (MPEC), see Luo et al., 1996.
The MPEC is closely linked to the Stackelberg game-theoretical models
(see e.g. Debo et al., 2004 on coordination of closed-loop supply chains).
Solving the MPEC, demands first proper estimation of the utility functions
and inherent relations between individual factors. The following literature
on return policies can assist on this. The model by Hess et al. (1996) com-
prises price, the value that the product has to the consumer, and the ship-
ping costs. The authors infer from the results that the seller has an incen-
tive to set a non-refundable charge to avoid opportunistic behavior and that
higher prices will have higher non-refundable charges. Moorthy and Srini-
vasan (1995) investigate the role of return policies in signaling quality (e.g.
customers may associate a priori high quality products with tolerant product
policies). The authors consider the following variables: price, matching prob-
ability (between customer’s wishes and product), depending on quality and
various costs both for the retailer and for the costumer. Somewhat opposed
to Hess et al. (1996)’s point of view, Moorthy and Srinivasan (1995) suggest
that high quality sellers (in principle associated with high prices) should con-
sider taking up buyer’s transaction costs. This puts into evidence the need
for more research in this area. Davis et al. (1998) have specifically computed
the optimal level of ‘hassle’ for return policies. They included factors such as
the trial value of the product. They concluded that a retailer tends to offer
a low-hassle return policy when the product cannot be consumed during the
trial period, when there are opportunities of cross-selling, and the returned
merchandize has high salvage value. Though in these references, the maxi-
mum contractual return period has not been included as a decision variable,
the references are still helpful to draw relations between other variables.
Furthermore experiences similarly to the ones described by Wood (2001)
can be conducted to tune the parameters’ relations. Wood (2001) carried
out experiences with undergraduate students for two types of return policies:
lenient vs. restricted. The author examined the deliberation time for the
decisions of 1) order, and 2) keep-or-return. The author further examines
the consumer’s evaluation of quality and the likelihood of continued search
while waiting for the product. The study shows that continued search is more
probable in the case of restricted policies rather than lenient policies. One
can judge this to be contradictory. Yet, the author explains it by defending
that product ownership is more a perception than the factual possession.
Concluding, the point is that setting a return policy is a complex matter
because of the many hidden elements that can influence profitability. Quan-
titative support on the topic is more than welcome.
Integrated operations: reusable packaging

In order to determine economic benefits, the needed number of reusable pack-
ages and the distribution over the network has to be determined. Kroon and
Vrijens (1995) and Duhaime et al. (2000) use respectively a location-allocation
and a minimum-cost-flow model to minimize the empty reusable packaging.
To determine the environmental impact of reusable crates, Life Cycle Analy-
sis (LCA) can be used (see Pappis et al., 2004 and Bloemhof-Ruwaard et al.,
2004). For every process (production, handling, transport) or material usage
the impact on the environment (for example global warming or acidification)
can be established by calculating the emissions of different components. It is
still necessary to choose one of the existing methods to calculate these emis-
sions. A priori there is no “best method” as its appropriateness depends on
the production cycle, the transport technologies, the country, and so on.
Inventory Management
If return rates are high, such as in distant selling businesses, it becomes nec-
essary to manage return stocks explicitly. One can also look at this from a
completely different perspective: remote sellers are overselling to customers
that return merchandize (and possibly underselling to customers that do not
return merchandize).
One can think about making two categories of customers: the ones that
(systematically) return merchandize vs. the ones that do not. Or better, to
discriminate between less and more profitable customers taking into account
purchased and returned merchandize.
The above is an obvious simplification of reality since many other cat-
egories can be distinguished. For instance, catalog companies carry large
amounts of historical customer data that can be employed to draw the profile
of each customer class.
Existing inventory models with priority customers can be stretched to
this new application (see Kleijn and Dekker, 1999). Research has showed
that ‘... in some cases it is optimal to ... reserve inventory for possible orders
from higher-margin customers’ (Cattani and Souza, 2003). In other words, it
may be optimal to reserve inventory for customers that do not systematically
return merchandize. This nurtures confidence in the opportunities that this
sort of models can offer to retailers like catalog companies. Return-handling
costs can be explicitly utilized to draw the line between the two classes. In this
way, return-handling costs are plugged in explicitly, bringing more realism to
inventory management with product returns.
There are many other issues on inventory management with product re-
turns for instance in remanufacturing settings. Recent reviews can be found
in Dekker et al. (2004).
Internal (return) transportation

Operations Research models (stochastic models, mixed-integer linear pro-
gramming models or simulation) can be applied successfully in the evaluation
of material handling systems, in particular estimating the number and type
of vehicles needed. By comparing multiple scenarios, with multiple types of
material handling systems, an evaluation of the best system can be made.
The underlying models try to determine the number of vehicles of a certain
type in a certain facility. Examples are the evaluation of single-load versus
multiple load vehicles (Van der Meer, 2000), or lifting versus non-lifting ve-
hicles (Vis, 2002). Other, related, design areas are vehicle transport track
design, choice of pick-up and delivery points, design of deadlock-free tracks,
track claim design, or design of battery loading areas. The paper by Goetz
and Egbelu (1990) is an example. Material handling systems used for both
forward and return flows may have to meet different requirements. Depending
on the return volume, separate systems may be needed, that are better fit for
return handling.
Information systems
On this we refer to Kokkinaki et al. (2004).
Table 4.3: Research opportunities for medium-term decisions.

To model this as the Math- - MPEC: Luo et. al, 1996
Integrated operations: ematical Programming
return policy problem with Equilibrium
constraints (MPEC)
- models on relations be-
tween key variables: Hess
et al., 1996; Moorthy and
Srinivasan, 1995; Hess et
al., 1996; Moorthy and
Srinivasan, 1995; Davis et
al., 1998;
- empirical evidence on re-
lation between key vari-
ables: Wood, 2001
Life Cycle Analysis (LCA) - Operations Research

and Eco-eco Models models: Kroon and Vri-
jens, 1995; Duhaime et al.
Integrated operations: (2000)
reusable packaging
- LCA models: Pappis
et al., 2004; Bloemhof-
Ruwaard et al., 2004
To model two classes of - Inventory Management

customers: less and more with priority customers:
profitable taking into ac- Kleijn and Dekker, 1999;
Inventory Management count purchased and re- Cattani and Souza, 2003
turned merchandize.
To employ returns explic- - evaluation of material

itly in existing models handling systems: Van der
Meer, 2000; Vis, 2002;
Internal (return) trans- Goetz and Egbelu, 1990
portation
see Kokkinaki et al., 2004

Information systems
Table 4.3 summarizes the research opportunities and the literature that
can be supportive for medium-term decisions.
4.5.3 Short-term decisions

Inventory, storage and order picking models
This topic includes particular problems within warehouses, such as order
batching, order picking, routing pickers in a warehouse, warehouse zoning
(dividing a picking area in zones to achieve certain objectives), product stor-
age allocation, forward versus reserve storage area decisions.
Many papers deal with warehouse planning and control. There are sev-
eral overview papers in this area; recent ones are by Van den Berg (1999),
Rouwenhorst et al. (2000) and Wäscher (2002). The following topics have
been addressed in Operations Research literature: product to storage alloca-
tion, Heskett (1963); order batching, wave picking, De Koster et al. (1999);
routing order pickers in a warehouse, Roodbergen (2001); warehouse zoning,
worker balancing, De Koster (1994); Bartholdi and Eisenstein (1996); and
Bartholdi et al. (1999); forward - reserve problem, Hackman and Platzman
(1990); pallet or container loading, Coffman et al. (2000).
All these topics are strongly interdependent in their impact on warehouse
performance (e.g. order through time). There is some literature on combi-
nations of these above-mentioned problems (see Gademann et al., 2001; and
Roodbergen, 2001). Recently, storage and routing received some more atten-
tion (see, for example, Dekker et al. 2002).
Currently, no models exist that explicitly include return flows. Here is a
list of relevant issues that should have priority:
• Whether or not to set-up separate storage areas for returns, or to con-

solidate them with existing stock. Return volume seems to be a decisive
factor, as it is concluded in Chapter 5: above a threshold, returns are
going to be dealt with in a separate area. Quantitative models can bring
insight on what is missing: for instance to identify where the threshold
is. The trade-off is in the size of the storage area and order pick travel
time, but also in the cost of a warehouse of increased size. Storing re-
turns on separate locations may save time in storage, but potentially
requires more space, which in turn increases pick times.
• How long to buffer product returns before consolidating them on stock?

Important factors that play a role in such decision are the quantity,
variety and timing of the returned items. The larger the stored batch
size is, the more efficient the storage process becomes, while on the
other hand the buffer space may become overcrowded and the stock is
not available in time.
• Storing returns versus order picking: how to sequence and route both re-
turn storage with order picking? In many occasions, return job schedul-
ing issues play a role, since picks can only be carried out when the
(returned) products are at their location (this is the case at CTC, see
section 4.3). On the other hand picks are much more urgent, because
of due times for shipment. This problem has received no attention from
researchers so far.
Vehicle planning and scheduling

A large part of the literature on material handling equipment covers plan-
ning and control issues of unit-load automated storage and retrieval systems,
i.e. AS/RS (see Bozer and White, 1984; Bozer and White, 1984, 1990; and
Hausman et al., 1976).
Another large group of papers is on planning and control of automated
guided vehicle (AGV) systems. These internal transport systems are suited to
work in highly dynamic, where the horizon over which information is available
of loads that need transportation is usually short. On a slightly longer horizon,
information may be known in advance, but is uncertain. Load arrivals are
stochastic and also transportation times are stochastic, due to congestion and
intersection control policies.
The following activities have to be carried out by a controller of the sys-
tem: dispatching of vehicles to pick up certain loads, selecting the route,
scheduling the vehicle, and dispatching vehicles to parking positions. When
returns are involved, the planning and scheduling of such material handling
equipment will change. Returns can be modelled as another type of (storage
or transport) job that has to be carried out. However, particular-sequencing
restrictions will apply. For example storage jobs resulting from returns must
be carried out prior to retrieval jobs. Or, for unit load handling machines, it
must be attempted to combine jobs in a double play to increase machine uti-
lization. Vis (2002) has developed a new control concept to schedule storage
and retrieval for a unite load AS/RS working in multiple aisles. Unit loads are
for instance pallets. However, this does not apply necessarily to the situation
of returns as products are mostly returned item by item (or at least in small
quantities). Products in small quantities are normally stored or retrieved by
a worker travelling with proper equipment. As Vis (2002) describes, in this
case retrieval and storage requests are not combined.
Table 4.4: Research opportunities for short-term decisions

- to investigate whether, or Chapter 5
Inventory, storage and not to separate storage ar-
order picking models eas
- to investigate how long (not explicit)
buffer product returns
- storing returns vs. order (not explicit)

picking
To extend Automated Bozer and White, 1984;

Vehicle planning and Storage/ Retrieval Sys- Bozer and White, 1984,
scheduling tems (AS/RS) models 1990; Hausman et al.,
1976; Iris, 2002
Table 4.4 summarizes the research opportunities and the literature that
can be supportive for short-term decisions.
In conclusion, there is a shortage of quantitative models to support return
handling decisions. At the same time, there are many opportunities to extend
existing forward models or to initiate new research paradigms. Presumably,
these quantitative models will not be straightforward and more has to be
learned on return processes before they can be developed and implemented
successfully.

Handling return flows involves unique processes such as collection, inspection,
selecting, sorting, and recovery (see also Chapter 2). This demands additional
decisions on a larger number of issues.
This chapter served the following purposes:
• to put together a decision-framework for return handling, and
• to identify potential models to support the decision-making process.
The decision-making framework for return handling was organized from

long- to short-term decisions, as follows.
• Long-term decisions
- facility layout & design

- outsourcing
• Medium-term decisions
- integrating operations (return policy and reusable packaging)

- inventory management
- internal (return) transport
- information systems
• Short-term decisions
- inventory, storage and order picking control

- vehicle planning and scheduling
Note that the decisions cannot be taken in isolation, because the many
interdependencies.
Explicit models, to support the return handling decision-making process in
the warehouse, have been largely ignored up to now. For a number of research
areas, we highlighted in Section 4.5 how forward models can be extended, or
how new models can be launched, to include return handling. We identified
several research gaps that can be turned into a research agenda, as described
next.
• Which impact do return flows have respectively on:
- the warehouse layout;

- material handling systems (e.g., whether dedicated return handling
systems do pay off);
- storage and picking procedures (ex. route combination of returns’
storage and pick ups).
Furthermore we suggested the use of the MPEC formulation to e.g. help

retailers deciding upon the return policy.
A new direction to inventory modelling was also put forward as by cate-
gorizing customers in two classes, i.e. less and more profitable depending on
whether, or not, the customer (systematically) returns merchandize.
The environment that surrounds forward and return operations today
should also have a place in academic research. To mention only two: pack-
aging legislation in the European Union and e-commerce. It is a fact that
packaging waste regulation has kindled research interest on material recov-

ery. The impact of packaging legislation, both standardization and reuse,
on present warehousing operations has however not been thoroughly inves-
tigated. Regarding e-commerce, it is well known that fairly tolerant return
policies have impact on return flows (see Piron and Young, 2000). Yet, the
impact on warehouse operations still lacks investigation.
Most likely some of the ideas for future research laid down here are not
straightforwardly carried out. We believe that learning more about the prac-
tice of return handling can subtract some of the modelling difficulties. Not
only quantitative models are needed, but in order to implement it success-
fully, more qualitative, or empirical approaches are necessary. For example,
research towards best practices may help decision makers to come up with
solutions for sorting, buffering, and storage. By conducting simultaneous
desk and field research, quantitative models will plausibly aid on real return
handling decision-making.
Concluding, the handling of return flows is not as the usual handling

of forward flows coming into the warehouse. This was illustrated with an
example in Section 4.3 and further discussed in Section 4.4. It involves other
processes than forward logistics, thus it involves additional decisions that call
for explicit decision-making support tools.
Furthermore, the interactions between forward flows and return flows
should also be taken into account in future research. Chapter 5 takes a first
step in this direction, by analyzing the specific decisions of combining vs.
separating forward and reverse flows during transport, receipt, and storage.
Chapter 5
Return Handling: an
exploratory study
We shall never cease from exploration

And the end of all our exploring
Will be to arrive where we started
And know the place for the first time
T. S. Eliot
5.1 Introduction
According to the Material Handling Industry of America (MHIA, 2002), ma-
terial handling is the movement, storage, control and protection of material,
goods and products throughout the process of manufacturing, warehousing,
consumption and disposal. The focus is on the methods, mechanical equip-
ment, systems and related controls used to achieve these functions, usually
internal, within the company (see, for example, Tompkins et al., 1996). Ac-
cording to research of ELA (1999), 7.7% of the expenses of 500 interviewed
European companies consists of logistics costs, of which 2% is in warehousing
and 3.1% is in transport (1.2% administration and 1.4% inventory costs).
Return handling is even more costly. In some businesses return rates
can be over 20% (for example in the catalogue industry and online sales),
and returns can be especially costly when not handled properly (see Meyer,
1999 and Morphy, 2001). As reported by O’Neill and Chu (2001), Forrester
Research concludes that in 1999 the value of returned goods purchased online
147
148 Chapter 5. Return Handling: an exploratory study
was 600 million US dollars, and the cost of processing them was $468 million.
On top of this, this year, these numbers are likely to augment steeply.
Return flows affect all the actors in the supply chain. Figure 5.1 illustrates
those flows, where the flows with origin at the consumer are represented in
more detail. Used products may enter the reverse chain when collected to be
repaired, refurbished, remanufactured or recycled (see Chapter 2 for more on
recovery options). Next to used products, packaging and waste are sometimes
moved upstream between the chain actors. In addition, take-back agreements
between parties result in reverse flows of non-used products. The focus of this
chapter will be on retail logistics. In this case the reverse flows are mainly
due to commercial agreements or obligations towards suppliers or customers,
i.e, B2B and B2C commercial returns (see the framework in Chapter 2).
Figure 5.1: Reverse flows in the supply chain.
The systematic study of return flows has gained increasing attention, either
with respect to manufacturing or retail logistics (see RevLog, 2003). In spite
of all the literature dedicated to return flows (see Chapter 1), the details of
5.2. Handling returns 149
retail return operations and financial aspects of storage, handling and trans-
portation have been overlooked. Some exceptions are the studies by Beullens
et al. (1998), Dethloff (2001) and Beullens et al. (2004). These contributions
are in the area of vehicle routing when both forward and reverse flows are
taken into account. Yet, in the retail environment, the studies are practically
absent. As put by Tibben-Lembke and Rogers (2002): “Although much has
been written about many specific aspects of reverse logistics, or specific ac-
tivities, the collection and disposition of reverse logistic product, in the retail
context have been largely ignored.” Tibben-Lembke and Rogers (2002) pro-
vide an overview of overall differences between forward and reverse logistics
for retailers. The authors remark that the reverse process is less transparent
and accordingly reverse logistics costs are less visible than in forward logistics.
In this chapter we consider the handling of return flows and its relation
with forward logistics. The factors contributing to the decision whether or
not to combine inbound and outbound flows during the handling process of
product returns are investigated. That is done by comparing nine retailer
warehouses. This leads to the identification of key factors contributing to the
different practices. The analysis brings further insights about the complicat-
ing issues, possible simplifying solutions and practical implications. Finally,
propositions are put forward feeding future studies on return handling effi-
ciency. The analysis and results were reported in De Koster, De Brito and
Van Vendel (2002).
The remainder of the chapter is structured in the following way. Sec-
tion 5.2 reports relevant operational issues regarding the handling of returns.
Subsequently, Section 5.3 presents the cases and the methodology. After that,
Section 5.4 analysis the nine case studies and compares them. The discussion
and implications are put forward in 5.5, and Section 5.6 is dedicated to overall
conclusions, recommendations for practice and research opportunities.
5.2 Handling returns

Not only the product itself but also the product carrier (pallet, roll cage, etc.)
and packaging materials may go back and enter again the supply chain con-
stituting thus reverse flows. Therefore, handling returns may involve dealing
with a variety of products, as well as a variety of packaging materials. Fur-
thermore, products may return before use due to commercial agreements or
after use at the end of the life cycle. The former, also known as B2C com-
mercial returns and reimbursement returns (see the framework in Chapter
2), are not always saleable at the original price, but in many situations these
are new products that can be put again in stock (see De Brito and Dekker,
2003). End-of-life returns, mainly a consequence of environmental legislation

steadily increasing in the European Union, may later be reused, remanufac-
tured, recycled, or properly disposed. In this chapter, end-of-life returns are
not considered since they are currently negligible for the retailers in question.
For further information on end-of-life return handling issues, see De Koster
et al. (2000).
Figure 5.2: The process of handling returns.
In order to handle returns in practice, several decisions about the logistics

process of the returned products have to be taken. The focus here is on
the decision whether or not to combine forward and reverse flows during the
return handling process. In addition, and when appropriate, we comment on
related warehousing decisions (see Chapter 4).
Figure 5.2 exhibits the three stages of this process, as considered in this
chapter:
• Collection & transport to the warehouse
• Receipt & sorting in the warehouse
• Storage in the warehouse
In the first stage companies have to decide if they are going to simultaneously
pick up returns and distribute new goods. The trade-off here is between the
5.3. The case studies: brief description and design 151
exploitation of existing resources (trucks, route planning) and the increase of

complexity (e.g. new and returned products sharing truck capacity). Later,
returned products have to be received and sorted at the warehouse. In this sec-
ond stage, consisting of receipt and sorting, companies have to decide whether
or not these operations are to be done in the same area/facility where pro-
cured products are handled. Finally, in the third stage products in good
condition are put back in inventory to be sold again (a common practice in
several businesses).
The decision on combining vs. splitting reverse and forward flows, is
looked at from two perspectives: i) the physical storage and ii) the registra-
tion of information in the management system. In general terms, firms have
to decide if they will use the same facilities and network structures to handle
both the forward and the return flow of products. This may imply addi-
tional transport, space, or even buildings and specialized labor requirements.
Moreover, all these decisions are subordinated to the overall objectives of the
company such as minimizing costs, providing a high service level and keeping
high controllability of operations. The different stages of handling returns are
the central aspects of the case study analysis. The next section considers the
case study methodology.
5.3 The case studies: brief description and design

The cases comprise the handling of returns in warehouses of three lines of
businesses: food stores, non-food store chains and mail order companies. All
of these are mature industries that deal with returns on a daily basis. In total
nine retailer warehouses were considered, three per industry branch allowing
inter- and intra-industry comparison.
Food retailing is a business with large volumes and small profits. The
margin in the Netherlands is about 1% to 2% per item (see De Koster and
Neuteboom, 2001). The three food retailers studied, identified as FR1, FR2,
FR3, have full line warehouses. This means that the full line of products is
stored in the warehouse, and is supplied via the warehouse to the supermar-
kets.
The non-food store chains are here called SC1, SC2 and SC3. They are all
large non-food store chains in the Netherlands. SC2 and SC3 are the largest
department stores. SC1 is a large speciality retailer in sports apparel and
accessories.
Finally, the mail order companies labelled as MO1, MO2 and MO3 are
the largest mail order companies in the Netherlands. Customers usually order
from catalogues, which are freely distributed. Nevertheless, the Internet is
also increasing its role as a market channel. As case candidates the main
players in the Netherlands were elected.
In the case of food-retailers however, also all national players have full-line
regional warehouses with assortment and sizes comparable to regional players
(see De Koster and Neuteboom, 2001). Thus, for this study of warehouse
operations, it is indifferent to choose national or regional players in the food
sector.
In the first stage of the methodology design, more companies were con-
tacted by phone in order to evaluate the feasibility of the research. Data
availability and willingness to cooperate during the planned research period
were decisive factors on whether, or not, to include a concrete case in this
study. In total nine retailer warehouses became suitable for research. In ad-
dition to detailed interviews at each distribution center with logistic and other
specialist managers, field surveys were carried out on site to gain an appreci-
ation of the issues involved. After that, phone calls followed to clarify aspects
when necessary. Subsequently, interviewees had the opportunity to check the
information collected. The field research took place in 1999/2000. Informa-
tion about the handling process (transport, receipt and storage regarding each
warehouse) was gathered.
In addition and to facilitate the understanding of the conditions in which
decisions are taken, the following additional data was collected: type of prod-
uct and materials being returned (what-type, see Chapter 2), return policy,
and return volume. Moreover, an estimative of the space and manpower as-
signed to return handling, as well as the associated costs are given. This data
will help as a criteria to evaluate the performance of the retailers, with respect
to return handling (see 5.5). Below we list all the elements.
Contextual factors
• Type of returned product or material (goods, carriers, materials)
• Return policy (time and other constraints)
• Return volume (per return flow, and compared with total outgoing flow)
The process
(focus: separation vs. combination of reverse and forward flows)
• Transport (combined vs. separated; also: self-executed or not)
• Receipt (combined vs. separated; also: automated vs. manual handling)

5.4. The case studies: analysis and comparison 153
• Storage (combined vs. separated)
Manpower, space and associated costs

• Manpower during receipt and storage (amount of labor dedicated to
return handling)
• Space for receipt and storage (the area devoted to return handling)
• Personnel and space investment
The costs of stocking and handling returns were estimated on the basis of
the space used for return handling and storage processes. The estimation
was carried out as follows. In the case there were areas dedicated to returns,
these areas were simply measured. The same was done when return stor-
age was combined with outbound storage. Next, the yearly facility costs for
returns were obtained by depreciating the total investment costs of the facil-
ity in 30 years and taking a part of this amount proportional to the space
for returns. In addition, personnel costs were considered, with an average of
approximately 27,300 annual euros per employee (this is an estimated value
based on the salaries in the Netherlands, 2002). These are not the total costs
of handling returns, since transportation costs and inventory holding costs
are not included. However, building/renting the warehouse and manpower
costs represent the large slice of total costs (presumably above 50% and 25%
respectively, see De Koster, 1996). In addition, substantial investments in
information systems for returns usually have to be made (Olsen, 2000; Linton
and Jonhston, 2000), which have a much shorter depreciation period than
30 years. Finally, it is very difficult to include costs of internal transporta-
tion personnel (for instance forklift truck drivers), since they are not solely
dedicated to return handling. In the next section and for clarity purposes,
information on other particularities, retailer specific, or directly related to the
handling of product returns is added.
5.4 The case studies: analysis and comparison

5.4.1 Contextual factors
Food retailers
Among the three investigated food retailers, FR3 is active in the whole coun-
try, while FR1 and FR2 operate at a regional scale. A variety of consumer
goods, from dry groceries, dairy products, beverage and meat, to vegetables,
fruit, and frozen products have to be stored in the warehouse.
FR1 has one central distribution center (DC). The national market share is
about 2%, and the 30 supermarkets are served from the DC. The supermarkets
are totally owned and exploited by the FR1 organization. The majority of
the 8,000- 9,000 stock keeping units (SKUs) are mainly shipped to the 30
supermarkets via the DC. Fresh dairy products, bread and cheese are directly
shipped from the suppliers. The distribution center of FR1 deals with the
following what-type of returns: i ) consumer goods (supermarket stock), ii )
distribution items (various carriers such as roll cages, crates, pallets), iii)
other (waste) materials (such as paper and plastic).
The food retailer FR2 is also a regional chain with a single DC. The na-
tional market share is 1.4% and the 30 supermarkets are served from the
DC, including the bread and the meat that are manufactured in in-house in-
tegrated factories. In total, this supermarket organization deals with about
11,000 SKUs. FR2 deals with return types identical to FR1 i.e. products, dif-
ferent carriers and other (waste) materials. The warehouse has two buildings,
one for cold store, return handling and waste, and the other for the remaining
activities including bread and a meat factory. The food retailer FR2 deals
with the same type of returns as FR1 (see previous paragraph).
FR3 is actually a wholesaler with about 500 franchised supermarkets, for
which the operations of storage and distribution are similar to those of FR1
and FR2. The national market share is 10.6% with an assortment of 12,000
products including tobacco and cosmetics. FR3 has several warehouses. For
this study, the newest warehouse is considered, from which 134 supermarkets
are served. FR3 receives back few products and little waste material because
the supermarkets are responsible for disposing the packaging materials and
waste. If the supermarkets wish the DC to handle it, they have to pay for
it. Therefore, the returns are mainly product carriers. The warehouse has a
two-floor structure, with the top floor dedicated to freeze and cold storage.
Table 5.1. shows quantitative data concerning the return flows (expressed
in roll cages) at the three food retailers. It is clear that the majority of
returns at the food retailers consist of empty carriers, e.g. empty roll cages
and crates (more than 60% of the returns in all cases). Actually, almost all
returns at FR3 are empty carriers because independent entrepreneurs have to
pay if they want the DC to handle their returns. Materials (like paper and
plastic) and waste handling is significant only at FR1 and FR2, both having a
policy to centrally collect and handle returns and waste. Furthermore, at all
food retailers, FR1, FR2 and FR3, return products have to be authorized to
obtain money restitution (see Chapter 4 on integrating operations, medium-
Table 5.1: Outgoing and return flows (in roll cages, per week) at the food
retailers.
Description FR1 FR2 FR3
Outgoing flow 17,000 14,950 28,000
Returned
distribution items 6,500 38% 5,600 37% 8,400 30%
other materials 4,000 24% 3,000 20% 255 0.9%
consumer products 100 1% 100 1% 20 0.1%
Total returns 10,600 63% 8,700 58% 8,675 31%
term decisions).
Non-food store chains

SC1 is part of a European retail organization with five different store chains in
the Benelux in the shoe and sports sector. There are three shoe store chains,
with 87, 40 and 65 stores respectively. The two sports chains have 30 and
16 stores respectively. All the stores are served from the central warehouse
studied in this chapter. The assortment is about 4,000 articles. However,
when size and color are included, there are about 20,000 SKUs of consumer
goods, divided in the categories men’s, women’s, children’s and sports shoes
and clothing, and accessories.
SC2 is a chain of seven large department stores in The Netherlands in
the luxury market segment. The assortment consists of consumer goods such
as shoes, clothing, white and brown goods, books, CDs etc. (about 300,000
SKUs, changing constantly). All stores are delivered 1 to 3 times a day from
the central warehouse. On average, there are 30,000 SKUs in storage, half of
which consist of fashionable clothing, which have a short selling life.
SC3 is a chain of 66 department stores in The Netherlands, with consumer
goods such as clothing, shoes, white and brown goods, books, CDs etc. The
total assortment consists of about 200,000 SKUs (not all on stock). The
stores are replenished at least once a day from one of the four warehouses.
All warehouses carry a different part of the assortment. In this chapter the
central warehouse is studied. The assortment consists of fashionable clothing,
books and media products and is responsible for 50% of total revenues.
The three distribution operations are therefore broadly comparable. They
all serve own stores with a fairly similar assortment and a large number of
SKUs on average on stock, although the total assortment of SC2 and SC3
is much larger than that of SC1. The other difference between SC1 and the
Table 5.2: Outgoing and return flows (in roll cages, per year) at SC1, SC2
and SC3.
Description SC1 SC2 SC3
Outgoing flow 4,500,000 18,000,000 55,000,000
Total returns 630,000 14% 1,700,000 9.4% 4,000,000 7.3%
other two retailers is that SC1 does not have hanging garments. For the
other two retailers this requires special storage and handling systems such as
overhead conveyors (see Chapter 4, on warehousing internal transportation,
medium-term decisions). The main return decision is taken centrally by the
DC, giving the DC more control over the returns (see Chapter 4 on integrating
operations, medium-term decisions). For instance, the DC may decide that
packaging material is returned on Thursdays, product returns on another day
and so on. SC1 tries actively to find the best sales outlet for the products and
reallocates products to different stores. These are special stock adjustments,
which sometimes in practice are referred as to condensation returns. All the
DC’s have considerable amounts of product returns, varying between 7.3% to
14% of outgoing flows (Table 5.2).
Mail order companies

Customers can order by mail, phone, fax, or internet and can return the
merchandize without any obligations up to four weeks after receipt.
MO1 is a US company that trades in exclusive consumer goods for col-
lectors, such as miniature cars, dolls, porcelain and jewels. The European
Distribution Center is investigated, from which all customers in Europe are
delivered, usually by national post organizations (PO). Products are shipped
from the DC to hubs of these PO’s in Europe. The assortment is about 10,000
SKUs, the return time limit is 30 days. About 7,000 products are shipped
per day. This is approximately 7,000 customer orders. Besides the products
themselves, product certificates have to be included, as well as leaflets.
MO2 is a home shopping company, with a wide assortment of consumer
goods. The company has two DCs. The DC studied here carries hanging
garments, small household appliances and small technical products, about
10,000 SKUs in total. Products are shipped via parcel carriers, in total 10
million units per year and 25,000 orders per day (on average, in 1995). The
return time frontier is 1 week.
MO3 is a German mail order company. The Dutch DC ships to customers
in The Netherlands and Belgium. The assortment of this DC consists mainly
of small household appliances and clothing. Registered customers receive a

catalogue twice a year. The maximum allowed time in which customers may
return products is 2 weeks.
Product return flows are about 10%, 30% and 25% of the outgoing flow
respectively for MO1, MO2 and MO3. Apparently, all mail order compa-
nies accept all returns and customers are usually credited (see Chapter 4 on
integrating operations, medium-term decisions).
Table 5.3 summarizes the aforementioned information regarding the return
type and policy of the nine retailers.
5.4.2 The process

Food retailers
FR1 uses small swap bodies with cooling capacity for transport between the
DC and the stores several times per day. Transport is self-executed. After
dropping off a full swap body at the store, the truck picks up the previous
swap body, which is now loaded with returns, and transports it to the DC.
All returns are received for inspection in a separate zone of the shipping area.
After inspection, products are sorted and separated.
A third party will subsequently discard products in poor state, paper,
bread, vegetables and other material waste. Product carriers are sorted man-
ually and dispatched to separate storage areas (there are separate areas for the
distribution items, like empty roll cages, empty crates, and empty pallets).
Products as-good-as-new are manually put back in the inventory together
with the procured products.
All packaging material and waste generated in the supermarkets is cen-
trally collected at the DC, for scale advantage (see Stock, 2001).
Trailers are used to transport to and from the supermarkets in the case of
FR2. Although there are 4 deliveries per day, supermarkets can only return
twice a week. This has to do with the fact that returns are only picked up at
the last supermarket in each route with 2 or 3 supermarkets.
Returned products are received in a separate zone of the second building,
while returned carriers are received at both buildings. Carriers will then be
inspected in separate zones of each of the buildings. The integrated bakery
and butchery handle respectively bread and meat waste. Other waste materi-
als are handled by third parties. Products in an as-good-as-new state go back
to inventory. The collection of packaging material and waste is centralized at
the DC.
FR3 self-executes the transport by means of trailers, many with three
Table 5.3: Return type and policy at nine retailers.

Return type and policy
FR1 FR2 FR3
Consumer products, dif- Consumer products, dif- Few consumer products
ferent carriers, and other ferent carriers, and other and little waste material.
(waste) materials (waste) materials Mainly distribution items
(empty carriers)
Central collection of pack- Central collection of pack- Supermarkets are respon-
aging materials and waste aging materials and waste sible to dispose packaging
at DC at DC materials and waste. They
have to pay if they want DC
to collect their waste
Product returns have to be first authorized so a money restitution can follow
(see Chapter 4 on integrating operations, medium-term decisions)
SC1 SC2 SC3
Consumer goods (sports Consumer goods (shoes, Consumer goods (clothing,
shoes, clothing and acces- clothing, white and brown shoes, white and brown
sories), distribution items, goods, books, etc.), packag- goods, books, etc.), packag-
packaging and other mate- ing and distribution items ing and distribution items
rials (such as paper and (hangers, plastic bins and (hangers, plastic bins and
decorative materials) clothing stands) packaging material)
Special stock adjustments Products can be returned The purchase department
returns are initiated at the when the purchasing de- decides what should be re-
sales department and are partment grants permis- turned. The DC decides the
credited to the stores. All sion. Except for off-season return instant. Returns are
other returns e.g. origi- returns, of which the stores credited to the stores. Ex-
nated by consumer’s com- keep the ownership, stores cept for fashionable cloth-
plaints, are initiated by the are credited for returns ing all returns are sold out
stores, which remain owner at reduced prices
of the products (B2C com-
mercial returns)
MO1 MO2 MO3
Collector items (miniature Clothing, small household Clothing, small household
cars, dolls, porcelain and appliances appliances
jewels, etc.)
Contractual time to return: Contractual time to return: Contractual time to return:
4 weeks 1 week 2 weeks
All products can be returned at no cost within the contractual time to return
(see Chapter 4 on integrating operations, medium-term decisions)
temperature compartments. Usually there are three deliveries per day per
supermarket, giving opportunities to pick up the empty carriers. The re-
turns are received in both floors of the warehouse, while the empty carriers
are received through special dock-doors at the top floor. There is a special
shuttle truck that moves returns between the two storeys, to make sure that
the returns are handled at the proper dock. Crates are sorted and palletized
through an automated system. All transport is daily and self-executed. How-
ever, when excess capacity is needed, transport is outsourced.
Returns collection is included in the route, although this is done differently

at all three retailers. The internal handling differs as well. All retailers
have separate loading docks for empty (beer and soft drink) crates. All have
separate areas where empty product carriers are stored. Waste materials at
FR1 and FR2 are received and handled (sorted) in separate areas, as well.
FR2 also disposes meat and bread waste. FR1 has outsourced this. Returns of
good products, close to the keeping date, form a minority in all organizations
and are added to ordinary stock locations. These are usually sold at selected
outlets at a lower price. At FR3, all waste material is weighed per store, since
the stores have to pay for the disposal. This is not done at FR1 and FR2. All
returns are handled manually, except empty product carriers at FR3, which
are sorted and palletized automatically. As exhibited in Table 5.1, the volume
of such returns is the largest at FR3. Table 5.4 summarizes the return process
for the food and other retailers.

Returns are picked up by the delivery truck and included in the route. SC3
self-executes while SC1 and SC2 outsource the transport. Nevertheless, both
SC1 and SC2 keep control of the transport planning. Receipt and inspection
of returns is rather similar for the three store chains. However, while SC1 and
SC3 have a separate storage area for returns, SC2 only has a separate storage
place for out of season products (see Table 5.4).

After unpacking, inspection, packing and re-labelling, the ‘as-good-as-new’
products are eventually added to regular stock. MO1 has a temporary location
for returns, from where orders are first served. However, when capacity is
reached in this location, product returns are stored together with initial stock.
MO3 also has a hybrid storing strategy. It distinguishes on the one hand bulk
locations for initial stock and on the other hand multi-SKU locations for
Table 5.4: The return process at the nine retailers.

Transport, receipt, sorting and storage
FR1 FR2 FR3
Transport is self-executed, Transport is self-executed, Transport is self-executed,
combined with several deliv- combined with daily deliv- combined with three daily
eries per day, per store eries. However, returns are deliveries
possible only twice a week per
store
S eparated area for checking Similar to FR1). However, Returns are received at both
and separate storage per car- bread and meat waste are han- floors in separated zones,
rier type; products as-good- dled at the integrated facto- while empty carriers are re-
as-new go back to inventory of ries ceived only at top floor. Sort-
new products ing and palletizing empty
crates is automated
SC1 SC2 SC3
Transport is outsourced, but Transport is outsourced, but Transport is self-executed
transport planning is self exe- transport planning is self exe-
cuted cuted
Several deliveries per store per About 1 delivery per sales de- About 1 delivery per store per
week, several stores per truck partment per store per day, day, usually only 1 store per
route. All deliveries are in roll usually only 1 store per truck truck route Returns are picked
cages. Empty roll cages and route. Returns are picked up up at the store by occasion of
returns are picked up at the at the store deliveries
store
Returns are received, checked and sorted at a separated part of the dispatching area
Paper is discarded. Prod- Off-season returns are stored There is a separate storage
ucts as-good-as-new are man- on special pallet locations. area (basement) where prod-
ually stored in separated areas Other good products are ucts are re-priced and stored
respectively for stock adjust- stored on free locations (bin in bulk before they are sold
ments returns, B2C returns storage area) or added to the out at the “budget market”
(e.g. complains) and other regular stock special outlet. Expensive
materials fashion clothing is added to
the regular stock.
MO1 MO2 MO3
Transport is outsourced. Or- Transport is outsourced, but Transport is outsourced to
ders are shipped to hubs of na- transport planning is self exe- P.O. Only sorting per country
tional P.O.s in Europe. Prod- cuted. All shipments are au- is necessary. Transport of for-
ucts are returned via national tomatically sorted by postal ward and return flows is not
P.O.s. Transport of forward code area. Returns are inte- combined
and return flows is not com- grated in the truck route and
bined are picked up at the customer
Returns are checked, reconditioned, repacked, labelled and added to the regular stock.
There is a separate return receipt and handling area
All products are stored in pal- There are three main stor- All regular products that are
let racks. Product returns are age areas: bulk (pallets), pick received in small quantities
at first (and until capacity is stock and hanging clothing are stored on multi-SKU bins.
reached) stored in a separated This also holds for all returns.
location, from where orders There are three main storage
are picked first areas: boxed clothing, hang-
ing clothing and a bin storage
area for other products. Bulk
stock is stored elsewhere
later replenishment (small quantities) and for returns (Table 5.4). In order to
handle the large number of returns, each of the three mail order companies
uses dedicated software.
5.4.3 Manpower, space and associated costs

The (internal) warehousing costs have been estimated. From the subsequent
analysis it appears that the warehousing costs for returns (storage and han-
dling only) form an appreciable amount of the total warehousing costs. In
fact, the total costs of handling returns are much higher when transportation,
and handling in stores are included.
Food retailers
Table 5.5 exhibits quantitative data on the manpower and dedicated space to
handle returns in the warehouse as a percentage of total area and manpower
employed. In matters of space, FR3 needs almost the triple of area for the
returns than FR1, and one third of the space used by FR2 (Table 5.5). One
must recall that on the one hand FR3 deals with little waste and FR1 has
outsourced disposal, while FR2 integrates the treatment of bread and meat
waste. The absolute estimated yearly costs to handle returns are 304, 320 and
394 thousand euros respectively for FR1, FR2 and FR3. In percentages, these
correspond to 6.5%, 12% and 5.8% of the total warehousing costs. Recall that
FR1 has to handle somewhat fewer crates and bins than FR3, but much more
waste. FR2 handles less returns than FR1, no matter what type (see Table
5.1), but the relative costs are much higher. Costs per returned roll cage vary
between 45 and 90 euro cents, with FR1 at the low end and with FR3 closest
to 90 euro cents.
Table 5.5: Manpower (in FTE) and space (m2 ) at the retailers with stores.
Description FR1 FR2 FR3
Manpower in the DC 149 70 198
Manpower for returns 10 6.7% 7 10.0% 12 6.0%
Total DC area 15,000 35,000 43,000
Return area 770 5.1% 6,000 17.1% 2,108 4.9%
SC1 SC2 SC3
Manpower in the DC 50 240 ±50 temporary 170
Manpower for returns 5 10.0% ±8 ±2.8% 10 5.9%
Total DC area 15,000 45,000 33,000
Return area 1,850 12.3% ±4, 500 ±10% 3,000 9.1%

SC1 uses 12.3% of the total warehouse area and 10% of the total employees
to handle returns. The numbers for SC2 are 2.8% and 10%, and for SC3 are
about 6.0% and 9.1% for respectively area and manpower for product and
material returns. (Table 5.5) The estimated yearly costs to handle returns
are in absolute values 192, 324 and 369 thousand euros respectively for SC1,
SC2 and SC3. These values correspond to 10.6%, 3.6% and 6.5% of the total
costs in the respective warehouses. This corresponds to much less than 45
euro cents per return unit, with SC3 and SC1 having respectively the lowest
and the highest costs.

The impression given by the visits to these companies is that they use between
5% to just above 10% of the warehouse space to handle returns, and around
10% to 15% of the working force. However, thorough estimation was not
possible for all mail order companies due to the way some handle returns.
For this reason, return handling’s cost comparison is out of the question. Yet,
it is feasible to provide a rough estimation of the costs. One should keep
in mind the dimensions of mail order companies businesses and that they
have large number of product returns (Meyer, 1999). Given the percentages
provided above, and based on the costs per return unit of the non-food store
chains, one can estimate the warehousing handling cost per return unit as
being approximately 45 euro cents. This embraces the work dedicated to
returns as receipt, inspecting/sorting and bringing them to storage location.
For simplicity one may assume that each individual operation corresponds to
one third of the cost. Goods also have to be transported to the customer, and
back, in the case of returns. In the Netherlands, several mail order companies
charge the customer with about 5 euros for transportation costs. Considering
that customers order 2 or 3 items at a time, the transportation cost per
product is estimated as being 1,50 euros.
Figure 5.3 depicts the processes involved from purchasing to delivering the
goods to the customer, as well as the processes involved in the case of re-
turns. Note that the operation inspection/sorting is much simpler for newly
purchased products than for returned products. In the latter instance, every
individual return has to be checked and re-conditioned if necessary (for in-
stance ironing returned clothes). With Figure 5.3 in mind, Table 5.6 displays
estimated costs for mail order companies, when a return occurs (constructed
as described before).
In Table 5.6 one observes that the costs double (from 1.95 to 3.90 euros)
Figure 5.3: The process of handling returns, with and without returns.
Table 5.6: Simple estimation of the costs for mail order companies, with and
without returns in euros (see Figure 5.3).
with return without return
Forward Process
Receipt 0.15 0.15
Inspection/Sorting 0.15 0.15
Storage 0.15 0.15
Transport 1.50 1.50
Return Process
Transport 1.50
Receipt 0.15
Inspection/Sorting 0.15
Storage 0.15
Transportation paid by the customer −1.50
Total costs 3.90 0.45
when a return occurs, but due to the financial construction (the customer only
pays for transportation costs when he/she keeps merchandize), the “net cost”
increases by more than a factor 7 (from 0.45 to 3.90). These numbers are not
surprising as in many industries, moving products back from the customer to
the producer costs between 5 to 9 times more than moving the same product
forward (Stock and Lambert, 2001).
More costs are involved on top of those discussed above, as all the three
mail order companies had a separate module in the Warehouse Management
System dedicated to return handling. These systems demand substantial
investment costs since it is hardly found as a standard functionality. This
software helps to monitor returns as for instance the number of times an item
has been returned.
5.5 Discussion and implications

In the previous section, we depicted the operational aspects of handling re-
turns associated with nine retailer warehouses in the Netherlands. Next we
put forward three general and three specific propositions, based both on liter-
ature and on our empirical analysis, feeding future studies on return handling
efficiency.
P1: Converting product returns into a sold product more than triples the
costs of immediately selling a product in one-trip.
It is not new that a lot of money is involved in returns. The Reverse Lo-
gistics Council have estimated that handicap return management costs every
year billions of dollars to U.S. firms (Rogers and Tibben-Lembke, 1999).
In the previous section, the estimated handling costs for mail order com-
panies doubled when products are returned. This follows naturally from the
processes involved. All the forward processes have to be performed once again,
but now for returns (transport, receipt, inspection/sorting, and storage). Yet,
the increase in costs is likely to be underestimated since every returned prod-
uct undergoes a thorough check followed by an indispensable reconditioning
procedure. In this respect, return handling is operationally more demand-
ing than forward handling. Actually, return handling in the DC alone has
been estimated as being twice as expensive as forward handling (see Zuidema
and De Koster, 2002b). The point is that returns contribute substantially to
operational costs, emphasizing the need for return handling efficiency.
P2: Reducing uncertainty (not only on quality, quantity, and time, but
also on product diversity) is critical for large-scale return handling.
5.5. Discussion and implications 165
Reverse Logistics has been associated with high uncertainties on time, quan-
tity and quality (see Dekker et al., 2004).
The empirical research allows us to state that not only when, how much
and the state of products, but that also the product diversity is a central
source of uncertainty (see also conjecture P5). Moreover, one can expect
that different combination scenarios of those four factors are more or less
problematic for return handling. For instance more of the same, i.e. large
volume of similar type of products is likely to be handled more efficiently than
a group of products with very different characteristics.
Indeed, we could notice that all retailers had mechanisms to reduce the
aforementioned uncertainty while steering the return scenario as much as
possible. Every retailer with stores centralized the return authorization. Al-
though with limitations, this gives the possibility to determine the instant
(time) for certain types and amounts of returns (quantity, quality and diver-
sity) being collected. For example, a retailer may determine that only one
stream of waste is being returned, or that complaint returns are collected
alone on Thursday, and so on.
Mail order companies have somewhat less chance of reducing uncertainty,
especially the ones that use P.O. services. Yet, MO2 integrates the delivery
route with return pick-ups at customers. In this case, the customer phones
the retailer requiring a pick up, to which the retailer proposes the pick up
date. Although customers are entitled by law with the right of returning
products, the mail order companies formally restrict the return period from
one to four weeks after the customer has received it. Still, there is evidence
that customers do not always respect the time of the return constraint (see
Chapter 6 and Schmidt et al., 1999).
Though mail order companies have limited control over returns, all in-
vested in information systems to register return data. Mail order companies
showed concern in limiting the number of times a returned item goes back to
the market and in recognizing customers that repeatedly return merchandize.
From the interviews it was not possible however to assess which action is
taken in the case of customers that frequently return products. Nonetheless,
each retailer was clearly involved in reducing uncertainty on quality, quantity,
time, and product diversity of returned products.
P3: Return handling has special warehousing requirements that are fun-
damentally different from those of forward flows, being of importance for
efficient handling.
Both academics and practitioners have pointed out that Reverse Logistics
is likely to have different basic requirements than forward logistics either in
respect to infrastructure (Gentry, 1999) or knowledge (Meyer, 1999).

This study gives empirical evidence to this type of assessments, in the
context of product flow handling in the warehouse. Table 5.7 was constructed
with the data of Table 5.1, Table 5.2 and Table 5.5. By observing Table 5.7
(numbers were rounded to not create illusory accuracy), one can notice that
FR3 uses 33% more Full Time Equivalents (FTE) to deliver 65% more prod-
ucts than FR1, per week (FR3 uses 186 FTE to deliver 28,000 roll cages per
week, while FR1 uses 139 to deliver 17,000). In contrast with this good rela-
tive forward performance, FR3 uses more people than FR1 and FR2 to handle
less volume of returns. This happens in spite of automatic crate handling in
FR3 and manual handling at the other two food retailers. The direct impact
on the cost per returned roll cage is the following: FR1 with a cost slightly
above 45 euro cents and FR3 with a cost close to 90 euro cents per roll cage.
One may suspect from these numbers that FR3 is not really being efficient.
However, it is likely that FR3 is in a learning phase, since the correspondent
investigated DC had started operations recently (see Section 5.4).
Table 5.7: Material flow per manpower (in FTE): forward vs. return flow.
Description FR1 FR2 FR3 SC1 SC2 SC3
FORWARD flow
Flow (roll cages/week) 17,000 14,950 28,000 90,000 360,000 1,100,000

Manpower (FTE) 139 63 186 50 (±)265 170
(1) Flow per FTE 122 237 150 1,800 1,358 6,471
RETURN flow
Flow (roll cages/week) 10,600 8,700 8,675 12,600 34,000 80,000

Manpower (FTE) 10 7 12 5 (±)8 10
(2) Flow per FTE 1,060 1,243 723 2,520 3,400 8,000
RETURN relative to FORWARD (flow per FTE)
(2-1)/1 7.69 4.24 3.82 0.40 1.50 0.24
Another aspect is that for almost all retailers with stores an FTE handles
several times the flow, when it is reverse than when it is forward (see Table
5.7). One should keep in mind that in the case of the food retailer, the bulk
of returns are crates, bins, followed by wrap materials and waste. The DC of
each of the food retailers handles less than 1% of products (see Table 5.1).
In the case of the store chains, the return types were not discriminated (see
Table 5.2). Since the goods are mainly clothing, it is not unreasonable to
think that the shops already performed the initial check. Besides this, some
Table 5.8: Material flow per area (in m2 ): forward vs. reverse flow.
Description FR1 FR2 FR3 SC1 SC2 SC3
FORWARD FLOW
Flow (roll cages/week) 17,000 14,950 28,000 86,538 346,154 1,057,692

Area (m2 ) 15,000 35,000 43,000 15,000 45,000 33,000
(1) Flow per m2 1.13 0.43 0.65 5.77 7.69 32.05

RETURN FLOW
Flow (roll cages/week) 10,600 8,700 8,675 12,115 32,692 76,923

Area (m2 ) 770 6,000 2,108 1,850 4,500 3,000
(2) Flow per m2 13.38 1.45 4.12 6.55 7.26 25.64

RETURN relative to FORWARD (flow per m2 )
(2-1)/1 10.80 2.39 5.32 0.14 -0.06 -0.20
of these returns result from stock adjustments, basically not needing thorough
inspection. However, when space is considered, for some retailers products
require more average space for return handling than the space required for
outbound products (see Table 5.8, which was built with data from Table
5.1, Table 5.2 and Table 5.5). This occurs with retailers SC2 and SC3 that
indeed have hybrid storage strategies (see the discussion on the process in the
previous section).
Apart from these mixed facts the point is that a retailer may have a very
good relative performance in forward flow, but the opposite relative perfor-
mance with the reverse flow. As mentioned before, a FTE in FR3 handles
more forward flow than a FTE in FR1, but the opposite happens with the
reverse flow (similarly when comparing SC1 and SC2). All the above fits
in the context of our conjecture: return handling has different warehousing
requirements than those of forward flows. Indeed, return handling encom-
passes operations as inspection, which is not part of forward handling. Yet,
this implies that there is room for research to identify potential specific needs
that matter in efficient warehousing handling. As put by Tibben-Lembke and
Rogers (2002), “... and much future research is needed to discover how to
best structure reverse logistics operations.”
Table 5.9 sums up the decisions of combining (C) vs. separating (S) the
forward from the reverse flows during the three stages of return handling:
transport, receipt, and storage. All the retailers with stores (food and non-
food store chains) combine the transport of product returns with outbound
flows. All the warehouses have a separate area to receive returns. On storage,
Table 5.9: Decision on whether to combine (C) vs. to separate (S) the forward
and the reverse flows.
Food companies Store chains Mail order companies

FR1 FR2 FR3 SC1 SC2 SC3 MO1 MO2 MO3
Transport C C C C C C S C S
Receipt S S S S S S S S S
Storage C C C S C/S S/C C/S C C/S
the majority of the retailers combine the flows, one separates and three retail-
ers have a mixed or hybrid arrangement. Below we discuss three conjectures
with these facts in mind.
P4: For retailers that supply (a sufficient number of) stores it is most
efficient to collect the returned material to the DC with the same truck that
delivers the products.
The rationale behind this is very simple. The truck returns to the warehouse
anyway and it also saves precious space for the stores, which are often placed
in urban areas with little expansion possibility and very high land prices. In
addition, in the case of waste or obsolete products, a number of stores should
be included in the return route to have economies of scale (Stock, 1992). Yet,
one should notice that when more than one store is included per route, the
loading of return material complicates the unloading of purchased products
at the stores that follow in the route. This is the case for FR2 and SC1.
Nevertheless, FR2 was able to simplify the operation by allowing each store
to return material only when the store is the last in the route, which happens
maximally twice a week. This emphasizes the importance of self-executing the
truck route planning. Actually most of the retailers considered in this study
kept control of the route planning. FR1, FR2, FR3 and SC3 self-execute the
transport operations. The store chains SC1 and SC2 outsource the transport
but they keep the route planning in-house. MO2 uses a delivery service but
also keeps the route planning in-house.
P5: For retailers that handle a high volume of returns, it is more efficient
to unload and sort returns in a separate area of the DC.
All the warehouses separate the inbound and outbound flows during the re-
ceipt phase and all the retailers deal daily with large volumes of product and
material returns. Product returns demand operations additional to those of
purchased products, such as inspection. If the volume of returns is not sub-
stantial it could still be reasonable to combine the forward and reverse flows,
but in the converse situation it is not likely. A complicating factor for return
handling is the diversity of products. Yet, handling of large homogeneous
flows may be automated. For instance at FR3 all empty crates are auto-
matically unloaded from roll cages, sorted per type and stacked on pallets. In
practice the design of the warehouse layout has to take into account the return
handling. Comparing the numbers for return costs as part of total warehous-
ing costs enforces this. They vary from 3.6% at SC2 to 12% at FR2. The fact
that FR2 uses two separate buildings to receive product and material returns
demanding transportation between the buildings is likely to contribute to this
relatively higher relative cost.
Separating vs. combining storage is related with the market for returns,
as follows:
P6 a: If the market for returns is different from the original market
then product returns are stored in a different area from those of purchased
products.
P6 b: If the market for returns is the same then storage of product returns
is likely to be combined with purchased products. Exceptions will be found
in the case of high-intended control over returned products.
For food retailers all returned products in good condition are stored together
with new products and they are sold in the same market. The same holds
for mail order company MO2. Department store SC2 combines storage apart
from off-season products, which are sold in another market. SC3 mainly sells
returns in “budget” outlets so they separate storage. This retailer only com-
bines the storage for expensive articles, which are sold in the original market.
SC1 is an exception in the sense that it separates storage but returned prod-
ucts go back to the same market. This is related to the fact that SC1 is very
concerned about keeping track of returns. This comes along with the active
role of SC1 in finding the best store to reallocate product returns. MO1
stores returns at first in a separate location from which orders are picked
first. This goes with MO1’s aim of giving priority to returned products for
shipment to customers. Such an approach provides a quick means to learn
more about the market, the product, and even the process. For instance,
repeated returns may indicate that some aspect is being overlooked during
return check or that customer expectations are not being met at all. When
capacity is reached in the temporary location, product returns are put to-
gether with purchased products. MO3 uses a hybrid solution: small amounts
of received products, including returns, are stored in a single bin (max 10
SKUs) on one location. For MO3 the control over returns is an issue of great
importance. Mail order companies have an additional difficulty because they
have a substantial volume of returns and elevated product diversity (number

of SKU’s) coming back in small quantities (per customer). Before we started
this study we suspected that the combination of volume and product diver-
sity would have an impact on the storage decision. From the analysis though
the complicating factor seems to be the decision of controlling returns. Yet,
both volume and product diversity can be a point of further research since
they might not be a complicating factor by itself but be so when combined
with other factors. Furthermore, to have a relevant degree of return monitor-
ing and control, a separate module for returns in the warehouse management
system is desirable as it is already the case for mail order companies.
5.6 Conclusions
This exploratory study facilitates the development of theory for return han-
dling and supports both recommendations for practice and academic research.
Return handling contributes considerably to operational costs. When prod-
ucts are meant to be put again in inventory, they have to pass a close inspec-
tion and inevitable reconditioning procedures, which are manpower-intensive
and therefore expensive. Not only quantity, quality and timing are relevant
sources of uncertainty in the reverse stream but also product diversity, as this
study showed. Reducing these four fronts of uncertainty streamlines return
handling operations, increasing efficiency. However, to master forward han-
dling does not seem enough for efficient return handling. There is actually
empirical evidence for the existence of different basic requirements in return
handling vs. forward handling, also in the warehouse scene. Among the
studied companies, retailers having a good relative performance in handling
forward flows were not always able to perform as well with respect to reverse
flows.
Regarding the focus on factors contributing to combine vs. to separate
the reverse and forward flows, three conjectures were put forward. In those
we established relations between the following issues (see also Figure 5.5):
1. serving stores & the transport phase decision;
2. return volume & the receipt at the warehouse;
3. the market for returns & the storage decision.
In more detail, we conjuncture that:

5.6. Conclusions 171
Table 5.10: To combine (C) vs. to separate (S) the forward and the reverse
flows: critical factors and implications.
To combine vs. to separate the forward and the reverse flows

Collection and transport Receipt and sorting Storage
Remark All retailers with All retailers sepa- The majority of
stores combine rate receipt retailers combine
transport storage
Influencing To serve (a suf- A substantial vol- The market for
factor ficient number of ume of returns returns
stores)
Complicating More than one store The combination of Intended controlla-
factor per route volume and variety bility over returns
of the return flow
Simplifying To allow stores To automate the Small amounts of
action to return material handling of homo- purchased and re-
only when they are geneous returns turned products are
the last in the route stored at the same
location
Practical im- To keep route To anticipate pack- Information sys-
plication schedule in-house aging (waste) legis- tems for product
lation returns
To design the warehouse (also) with
return handling in mind
Associated P4 P5 P6a, P6b
conjecture
• For retailers that supply a sufficient number of stores it is most efficient

to collect the returned products, carriers and waste to the DC with the
same truck that delivers the products (P4).
• For retailers that handle a high volume of returns, it is more efficient to

unload and sort returns in a separate area of the DC (P5).
• If the market for returns is different from the original market then prod-
uct returns are stored in a different area from those of purchased prod-
ucts (P6a).
• If the market for returns is the same then storage of product returns
is likely to be combined with purchased products. Exceptions will be
found in the case of high-intended control over returned products (P6b).
Table 5.10 refers to these findings during all the three stages of return
handling. One can notice that the retailers do not diverge much in some
of the particular decisions. This finding has been one of the inputs for the
aforementioned conjectures and it constitutes a valuable guideline for future
research. Besides that, some associated recommendations for practice and
research opportunities derived from the detailed information about all the
cases follow below.
5.6.1 Recommendations for practice

Here we make first recommendations for practice regarding the long- and
medium-term warehousing decisions. Figure 5.4 is structured according to
the decision-framework of Chapter 4 and it summarizes the general recom-
mendations. After that, we pay attention to the specific focus of this chapter,
i.e. the decisions on combining vs. separating forward and reverse flows
during transport, receipt and storage.
While designing the warehouse layout, practitioners should take into ac-
count return handling explicitly, i.e. returns receipt location and future stor-
age locations.
For retailers with stores wishing to reduce return uncertainty there are
two key-factors: to centralize return authorization and to keep routing in-
house even when outsourcing transport. These actions give the retailers the
opportunity to restrain return uncertainty by determining the exact moment
of each pick up (taking into account the stores’ needs). Alternatively, they
can set it as a constraint to the third party, when outsourcing transport.
Furthermore, if forward handling is the only concern, the warehouse may
end up with undesirable high costs for return handling due to disregarded
internal transport. Automating the return handling for homogenous returns
may simplify operations. However, this calls for substantial investment and
it relates primarily to returned product carriers. This is a sensitive matter,
especially with the enforcement of the directives on packaging and packaging
waste of the European Union. Therefore, retailers have to think thoroughly
on the consequences of packaging legislation before investing heavily on return
handling automation. For instance, if manufacturers comply to the packag-
ing standards launched by the European Committee for Standardization and
packaging gets lighter, is the automatic handling equipment able to deal with
this?
The last recommendation is to employ a dedicated module within the
warehouse management system for return handling monitoring. This sugges-
tion is related to the indication of fundamental differences in handling the
two types of flows and to deficiencies in standard software regarding return
handling.
Figure 5.4: Decision-scheme on general long- and medium-term warehousing

decisions.
Figure 5.5: Decision-making scheme on combining vs. separating forward and

reverse flows during transport, receipt and storage.
Regarding the focus on factors contributing to combine vs. to separate,

explicit advise is given in Figure 5.5.
5.6.2 Research opportunities

The conjectures presented in this chapter facilitate future studies on return
handling efficiency. For this reason, the perceived relations between factors
contributing to decisions during the handling process have to be further inves-
tigated. Besides this, the absence of disagreement for some decisions should
be a guideline for case selection and for the investigation of additional factors.
One could notice that all the retailers with stores combine the transport of out-
bound and inbound flows. In future research one may seek a priori contrasting
cases, or investigate the role of geographic store’s dispersion. Furthermore,
all the retailers unload and sort returns at a separate area of the DC. A con-
jecture stating that the high volume of returns is responsible for this, was
brought forward. Again, counter cases and new factors can be sought. In this
way, theory on warehouse handling efficiency can be built.
One aspect not taken into consideration here is the overall strategy of
the firm, which may justify different return strategies and handling processes.
This presumably helps us to understand why some retailers do not object so
much to costs when the goal is to keep track or to control returns. Indeed,
research towards mechanisms to cut back uncertainty in its four dimensions
(quantity, quality, time and product diversity) would be of practical value.

Though, there are studies on the impact of return policies on a number of
elements like opportunistic behavior and speed of consumption (see Hess et
al., 1996 and Davis et al., 1998), the impact on operations has not yet been
thoroughly investigated.
Another aspect lacking investigation is on the dispatching tactics when
both returned items and new merchandize are in stock. A typical situation
in a company is having the marketing department believing that new mer-
chandize should be shipped first, while the logistical department may think
otherwise (see Zuidema and De Koster, 2002a). In general terms, there is a
need for research on issues of monitoring and control return handling, bring-
ing efficiency to operations (see also Chapter 4 on quantitative support for
return handling). In addition, the technical literature, as mentioned in the in-
troduction, may benefit from a close look at warehouse running costs. While
modelling warehouse operations and associated costs, researchers would be
more apt to explicitly de-couple forward from reverse costs if robust estimates
were found.
Chapter 6
Inventory Management: a
critical analysis of
assumptions
Man is an animal suspended

in webs of significance
he himself has spun
Clifford Geertz,
paraphrasing Max Weber
6.1 Introduction
As mentioned in the previous chapter, Inventory Management and Control is
one of the key decision-making areas while managing product returns. Fleis-
chmann et al. (2002), van der Laan et al. (1999) and Inderfurth and van der
Laan (2001) are recent examples of scientific literature on inventory control
with product returns. From a modelling perspective, one of the consequences
of reverse flows is the loss of monotonicity of inventory levels between replen-
ishments of new products. That is, the inventory level does not only decrease
because of demand but it may also increase in the case of returns. Since
this makes the analysis much more difficult than traditional inventory con-
trol, authors use simplifying assumptions regarding the return process. These
assumptions typically are 1) the demand flow is a (compound) homogeneous
Poisson process; 2) the return flow is a (compound) homogeneous Poisson
177
178 Chapter 6. Inventory Management: a critical analysis of assumptions
process, and 3) the return process is independent of the demand process (see
Fleischmann et al., 1997; Dekker and van der Laan, 1999). However, there
is nearly no (scientific) literature providing empirical analysis of data on re-
verse flows. Thus, in this chapter we explore the validity of these common
assumptions by means of real data. First of all, we present a methodology
to check the assumptions empirically. We describe actual practice in compa-
nies with respect to information storage on returns and to inventory control.
Moreover, we apply the methodology to real data and we discuss practical
implications of our findings, for instance with respect to information man-
agement on inventory systems with returns. We employ data from three
companies here referred to as CERN, MOC and RF. Both CERN, the Euro-
pean Organization for Nuclear Research (see Cern, 2000) and RF, a refinery
in the Netherlands, deal with internal material returns. MOC, a mail-order
company, handles customer product returns. The results of this study have
appeared in De Brito and Dekker (2003b).
The remainder of the chapter is organized as follows. The next section is
dedicated to a review of the main assumptions in the literature when it comes
to inventory models with return flows. Then, in Section 6.3 the methodology
is put forward and the statistical analysis and tests are described. In Section
6.4 the data of the three companies is described. The data analysis can be
found in Section 6.5. The last section sums up the conclusions, practical and
theoretical implications and research needs.
6.2 Inventory models with returns:

a literature review
Literature on inventory control models with return flows can be divided into
two streams: 1) typical repair models; 2) other models with imperfect corre-
lation between demand and returns. In the former models there is a perfect
correlation between machine failure and the demand for a substitute, where
the number of machines or parts in the system is constant. This chapter
concentrates on product returns that fit in the second group. For recent sur-
veys on repairable inventory theory we refer to Cho and Parlar (1991) and to
Guide and Srivastava (1997b).
This review considers two types of models: deterministic models and
stochastic models.
6.2. Inventory models with returns: a literature review 179
6.2.1 Deterministic models

Schrady (1967) considers a deterministic inventory model in which a certain
percentage of sold products comes back, after a known period of time, to be
repaired. Repaired items are put in inventory to be eventually re-used. Since
the demand and the return processes are assumed to be continuous deter-
ministic flows, the dependency relationship between the demand and return
process is not explicitly modelled. Later, Richter (1996) and Teunter (2001)
extended this model with the option of product disposal. With respect to
the demand and return processes, Schrady’s assumptions remain roughly the
same in both extensions. Richter’s model however considers a discontinuous
return flow as returned products are collected at variable moments in time.
Also in a deterministic context, Minner and Kleber (2001) discuss the op-
timal control for a production and remanufacturing system with continuous
and non-stationary rates for both the demand and return processes.
6.2.2 Stochastic models

The stochastic models can be divided into two groups: periodic review models
and continuous review models.
Periodic review models

This category of models typically focuses on proving the structure of the
optimal policy rather than finding optimal parameter values. For instance,
Simpson (1978) provides the optimal policy structure for an inventory model
with product returns in which product demands and returns can be stochasti-
cally dependent within the same period only. Demand and returns are known
through a joint probability function, which can differ from period to period.
Inderfurth (1997) extends the previous model with non-zero (re)manufacturing
and procurement lead times. All other assumptions equal the ones of Simp-
son. Inderfurth proves that there is a simple optimal control policy structure
as long as the leadtimes for manufacturing and remanufacturing differ at most
one period. Buchanan and Abad (1998) consider a system with partial re-
turns. Each period, a fixed fraction of products is lost while a stochastic
fraction is returned. The authors establish an optimal policy for the case
that the time until return is exponentially distributed. Toktay et al. (2000)
study ordering policies for a business case of single-use Kodak’s cameras. The
model is as follows. After using the camera, customers take it to a shop/ lab-
oratory to develop the film. The laboratories return the used cameras to
Kodak (but sometimes they go to the so-called jobbers). Kodak dismantles
the used cameras and reuses the flash circuit board of every camera in the
manufacturing of new ones. A closed queueing network model is applied to
decide on periodic ordering decisions. Custom demand is treated as a Homo-
geneous Poisson Process (HPP) from which a known percentage is returned.
The time the cameras are with respectively the customer and the lab are mod-
elled by a queueing system with two infinite servers with general processing
times. These two servers together model the time until the camera returns to
Kodak. Another important feature of this paper is the identification of the
information’s value according to different scenarios. Kiesmüller and Van der
Laan (2001) develop a periodic review inventory model where product returns
depend on the demand process. Both the demand and the return streams fol-
low a Poisson distribution. All returns depend on previous demands through
a constant time until return, and two probabilities: the return probability
(when demand occurs, it is assumed to be known whether, or not, an item
returns) and the probability that a returned item is in a sufficiently good
condition to be remanufactured. The authors compare this model with the
situation of independent demands and returns. The outcome supports that it
is worth using information about the dependency structure between demands
and returns. Mahadevan et al. (2002) model a hybrid manufacturing and
remanufacturing system with independent demand and return Poisson pro-
cesses. The authors employ a push policy to combine the decision of when
and how much to (re)manufacture. Teunter and Vlachos (2002) inquire the
necessity of a disposal option for a similar model. The independent demand
and return processes are modelled according to a Poisson or a Normal distri-
bution. The conclusion is that only under certain circumstances the disposal
option brings economic benefits.
A special case of periodic review models is the newsboy model, where only
a single period is considered. Two newsboy models with returns have been
respectively proposed by Vlachos and Dekker (2000) and Mostard and Teunter
(2002). In the first, a constant fraction of the the sold items is returned and
can be re-sold only once. In the second, each item that is sold has a constant
probability of being returned and once returned it has a constant probability
of being recovered. Returned items can be re-sold more than once. Both
articles investigate the optimal order quantity for the single period. These
models are inspired by the order size decision that mail-order-companies and
e-tailers face every season.
6.2. Inventory models with returns: a literature review 181
Table 6.1: The assumptions of the literature on inventory models with returns.
Demand process Return process Processes’ rela-
tions
Deterministic models
Schrady (1967) Continuous Continuous Independent
Teunter (2001) (constant) rate (constant) rate
Richter (1996) Idem (Periodic) Independent
constant rate
Minner and Kleber Continuous Continuous Independent
(2001) (dynamic) rate (dynamic) rate
Stochastic periodic review
Simpson (1978) General General Dependency
Inderfurth (1997) non-stationary non-stationary allowed in same
period only
Buchanan and Random variable A stochastic fraction of demand comes back
Abad (1998) after an exponentially distributed time
Toktay et al. HPP Partial returns and Dependency com-
(2000) general distribu- prised through a
tion queueing system
Kiesmüller and van HPP HPP Returns depend on
der Laan (2001) demand by a con-
stant return proba-
bility and time un-
til return
Mahadevan et al. HPP HPP Independent
(2002)
Teunter and Poisson or Poisson or Independent
Vlachos (2002) Normal Normal
Vlachos and Stochastic A fraction Independent
Dekker (2000) of the demand
Mostard and Idem Partial Each item
Teunter (2002) returns is re-sold with con-
stant probability
Stochastic continuous review
Heyman (1977) Compound re- Compound re- Independent
newal process newal process
(explicit expressions for Poisson distribution)
Muckstadt and HPP HPP Independent
Isaac (1981)
Van der Laan et al. HPP HPP Independent
(1999)
Fleischmann et al. (Compound) HPP Independent
(2002) Poisson Process
Yuan and Cheung HPP A constant fraction of demand comes back
(1998) after an exponentially distributed time
Bayindir et al. HPP HPP Independent
(2003)
Continuous review models

Heyman (1977) analyzes different disposal policies for a single-item inventory
system with returns. He uses a model where demands and returns are inde-
pendent compound renewal processes and all leadtimes are zero. An explicit
expression for the optimal disposal policy is given when the processes are
Poisson. Also Muckstadt and Isaac (1981) investigate the control of a single-
item inventory system with independent demands and returns following a
Poisson Process. The authors derive some approximations. More recently,
Van der Laan et al. (1999) deal with policies in the context of two inventory
facilities, one of new products and the other of remanufactured items. The
model is based on unit demand and unit returns with independent Poisson
processes. Fleischmann et al. (2002) derive an optimal policy and optimal
control parameters for a basic inventory model with returns where demand
and return are independent Poisson processes. Yuan and Cheung (1998) dis-
cuss an (s,S) inventory system with returns and analyze the impact of partial
returns on rental systems. Demands are assumed to follow a homogeneous
Poisson process and there are no lost customers. Customers keep the acquired
item during an exponentially distributed time. When this period ends, the
item is either returned or disposed, with a given probability already known
upon demand. They also assume that the return period is already known at
the time of demand. Bayindir et al. (2003) employ a queueing model to inves-
tigate a hybrid (re)manufacturing system when the return ratio is a decision
variable. Demand and return processes are independent and follow Poisson
distribution. The time in market follows a general distribution. The authors
conclude that the remanufacturing facility is more or less used depending on
whether the production capacity is finite or not.
Although the above does not constitute an exhaustive review of the lit-
erature on inventory models with returns, it serves to identify the common
assumptions of these models. Table 6.1 summarizes the main assumptions of
the demand and return processes, and of the dependency structure between
these two processes. More information on inventory models with returns can
be found in Van der Laan et al. (2003).
The bulk of the literature, especially when uncertainty is modelled, as-
sumes the following: 1) the demand is a homogeneous (compound) Poisson
process; 2) the return process is also a homogeneous (compound) Poisson
process, and 3) the two processes are independent. These choices have to do
with the tractability of the Poisson distribution. This is also the main reason
behind the independence assumption. In fact, less restrictive assumptions
complicate the analysis significantly. Besides, as the review of the literature
6.3. Methodology 183
shows, in order to pursue an exact analysis or to give explicit expressions for

optimal policies one has to make these common assumptions. The indepen-
dence assumption has also been motivated by the scarcity of individual data
on product returns (see Fleischmann et al., 1997). This motivates empirical
analysis of real data.
6.3 Methodology
The statistical analysis encompasses the demand process, the return process,
and the relation between these two processes. Before developing a testing
framework for the demand and return processes, first the notation is intro-
duced. Let P (t) be a counting process that at time t ≥ 0 equals
N (t)
X
P (t) = Q(k) (6.1)
k=1
where N (t) ∈ Z is the number of customer arrivals in the time interval (0, t),
and Q(k) is the amount of items ordered (or returned) by client k. P (t) is a
Homogeneous Compound Poisson Process (HCPP) if (see Tijms, 1994):
• N (t), t ≥ 0 is a Poisson Process;
• Q(k) is identically and independently distributed;
• N (t) and Q(k) are independent processes;
Let the subscript D refer to a demand process and the subscript R to a return
process and let T be the random variable indicating the time to return. It
is clear that if the return probability p = 1 and T is negative exponentially
distributed, then PD (t) being an HCPP implies that PR (t) is an HCPP which
is independent of PD (t). However, in the case p < 1, PR (t) is not a homo-
geneous compound Poisson process given PD (t). To understand the latter
consider a demand y time units ago. The probability that this demand will
trigger a return is equal to
p · e−λ·y
(6.2)
p · e−λ·y + (1 − p)
where λ is the scale parameter of the exponential distribution of the random
variable T , which represents the time to return. Since Equation 6.2 depends
on y, we are not in the presence of an HCPP.
6.3.1 Testing HCPP

One may still opt for approximating PR (t) with an HCPP if there are demands
occurring regularly. To evaluate whether a process P (t) behaves as an HCPP
consists of checking the three sufficient and necessary conditions mentioned
before. In order to first check whether it is acceptable that N (t) is a Poisson
Process, one can then check if
A) N (t) follows the Poisson distribution, and if
B) N (t) has independent increments.
The last item implies that the occurrence of events after any time t is in-
dependent of the previous occurrences. For the first part, A), a chi-squared
goodness-of-fit test can be used (see Bain and Engelhardt, 1967). To employ
this test the potential outcome space has to be split into intervals. For a good
test performance, the number of intervals should be as many as possible as
long as the expected number of observations is at least about 5 per interval.
The second part, i.e. B), can be handled through a Ljung-Box Q test which
directly checks if the autocorrelations between increments are significant or
not. This test can be found in standard statistical packages, and can be
carried out without sophisticated knowledge of time series theory (see Box
et al., 1994). The investigation over Q(k) can be carried out by applying a
Wald-Wolfowitz runs test for randomness (see Walpole et al., 1998) and again
a Ljung-Box Q test. For the former test one finds in the literature tables of
critical values from sample size 6 onwards. Finally, the test of independence
based on Pearson’s coefficient can be used to trace whether N (t) and Q(k)
are independent or not (Bain and Engelhardt, 1967). This test gives the
best results when the sample is larger than 10 data observations. Next sub-
section describes the technical details of the aforementioned statistical tests,
which are summarized in Figure 6.1. The choice of those tests was made
targeting the following objectives: simplicity, parsimony, usefulness and test
performance (against small sample sizes).
6.3.2 Testing exponentiality of the time to return

Besides the assumptions in the literature regarding demand and return each
being a HCPP, these two processes are commonly assumed to be independent.
In the proposed methodology there is no direct investigation about whether
the demand and return processes can be dealt with as independent processes,
because the data to be analyzed is likely to have only a small amount of
returns. Instead, it is investigated whether it is reasonable to model the
time until return with an exponential distribution. The advantage of this is
1. To test if N (t) is a Poisson Process

A) N (t) follows a Poisson distribution → Chi-squared test
B) N (t) has independent increments → Ljung-Box Q test
↓
2. To test if Q(k) is
A) independent → Ljung-Box Q test
B) identically distributed → Wald-Wolfowitz test
↓
3. To test if N (t) and Q(k) are independent
→ test of independence
Figure 6.1: Statistical analysis scheme to test if a process is an HCPP.
that one may pool over multiple products, when the return lag is likely to
be the outcome of a logistical process rather than being a product specific
process. Inherently, one can expect that the process affects the outcome of
the tests. The exponential assumption also appears in the literature (see
Section 6.2). So, indirectly it is tested how dependent returns are of previous
demand occurrences. The outcome has important consequences for inventory
management, which are discussed later in this chapter.
The time until return is the period from the moment a product is acquired
until it is returned. A goodness-of-fit statistical test, k 2 , constructed by Gan
and Koehler (1990), is used for this investigation. According to the authors,
this test fairly detects the non-exponentiality in the presence of data from
a Weibull, a LogNormal, or Normal truncated, among two dozens of other
distributions.
6.3.3 Statistical tests: more details

Here, more technical details on how to carry out the aforementioned statistic
tests are given (see Figure 6.3.2).
Chi-squared goodness-of-fit test
Test hypothesis:
1A) N (t), t ≤ 0 follows a Poisson distribution;
The procedure is as follows (see Bain and Engelhardt, 1994). First, one has
to divide the whole potential outcome space into c intervals A1 , A2 ..., Ac ; Let
oj be the number of observations within interval j and pj the probability of
an observation falling in Aj and ej = n ∗ pj the expected number in the j th
interval with n being the sample size. For an error Type I equal to α (i.e. with
the probability of rejecting the hypothesis given that it is valid being equal
to α) we reject the hypothesis if the value of test statistic is larger than the
(1 − α) quantile of a Chi-squared distribution with c − 2 degrees of freedom,
i.e. if
c
X (oj − ej )2
≥ χ1−α (c − 2) (6.3)
ej
j=1
To have an accurate test ej should be larger or equal than 5, or so, for each
interval Aj . To have as many cells as possible is desired because it increases
the degrees of freedom. For the test on the demand process we observed
the number of arrivals per month, and for the return process the number of
returns per year. After that, the chi-squared test was employed.
Ljung-Box Q-test
Test hypotheses:
1B) N (t) has independent increments;
2A) Q(k) is independently distributed;
The test is based on a chi-squared test statistic (Box et al., 1994), which is an
approximated distribution of the autocorrelations. The MINITAB statistical
package was employed to carry out this test.
Wald-Wolfowitz test
Test hypothesis:
2B) Q(k) is identically distributed.
To apply this test (see Walpole et al., 1998), one has to divide the sample
into two mutually exclusive categories, e.g. above and below the median. The
test statistic is the number of runs, i.e. groups of one or more observations
of the same category. The test detects deviation in randomness and it may
suggest non-randomness due to trends or periodicity. To test the hypothesis,
one may rather divide the sample into two and then test if the distributions
are identical. However, these tests are only suitable for large enough samples
(Kanji, 1999).
Test of independence
Test hypothesis:
3) N (t) and Q(k) are independent.
The test is based on Pearson’s correlation coefficient (see Bain and Engel-
hardt, 1994). The time between arrival of clients and the amount ordered
(or returned) constitute the samples. In this way, it can be tested whether
the demand process is independent of the ordered quantities (and similarly to
the return process). For very small sample sizes, less than 10 observations, a
non-parametric rank test is preferable, e.g. the Spearman’s rank correlation
test.
k 2 goodness-of-fit test
Test hypothesis:
Time until return has an exponential distribution
The test is based on the P − P probability plot (see Gan and Koehler, 1990).
This plot is built drawing Zi = F (Xi /λ̂) versus pi , where
• F is the cumulative distribution function,
• Xi is the ith lower ordered observation, i = 1, ..., n with n being the

sample dimension (i.e. the sample X is put in ascending order, X1
being the lowest observed number and Xn the highest),
• λ̂ is the estimated scale parameter,
• pi is an appropriate plotting position.
The degree of fit is decided based upon the linearity featured in the graph.
The statistic proposed is based on measures of linearity:
Pn
2 ((Zi − Z̄)(pi − p̄))2
k = Pi=1n 2 2
(6.4)
i=1 (Zi − Z̄) (pi − p̄)
Approximations for the lower pth percentile are obtained through the formula
below:
1
1 − kp 2 = (6.5)
λp + n · βp
A table with the coefficients λp and βp can be found in Gan and Koehler
(1990).
6.4 The data

Case 1 - CERN
CERN is a very large research center with more than 7 000 scientists. These
scientists have several warehouses at their disposal with more than 15 000
stock keeping units. Scientists have two ways of acquiring the products they
need for their experiments: by going to the self-service stores or, by order-
ing via the web from the main warehouse. The first is meant for low-value
regularly used items, i.e. the ‘consumer’ type of goods, and the latter for all
other items (industrial goods. In both situations, scientists provide the budget
code of the project they are working on. This code is therefore registered and
the project is charged for the delivery (CERN, 1999). For already 10 years,
CERN registers inventory transactions through the LIMS package which has
been used on top of Baan’s TRITON system. Dr. B. Schorr and Dr. M. Kr-
ever developed the LIMS package (CERN, 1997). The database encompasses
the following general information for each transaction occurrence:
• product identification by means of a code and a name;
• product quantity and its monetary value;
• store identification (main vs. self-service store);
• transaction types, e.g. replenishment from supplier, delivery to client,

and returned by client;
In the case of products issued to scientists, the date of order placement, date
of planned and actual delivery, budget number and client code are in principle
registered.
In the case of a return, the date of transaction, the budget number and
client code can also be registered. Originally, an attribute was created in
CERN’s database to register the reason for returning the product, but in
practice it is not used. The main return reason is actually reimbursement
(see Chapter 2). In addition, the database had reason’s codes for wrong
deliveries, technical faults, delivered too late or even never ordered by the
client.
If the transaction is a replenishment, the supplier code and name is recorded
in the database. CERN’s data is not only extraordinary by its dimension and
life span, but also due to the individual registration of transaction occurrences,
including the identification of returns, client codes and budget numbers. In
most other commercial packages, transaction data are stored only for a short
6.4. The data 189
time and are aggregated to e.g. monthly or yearly demand figures. When
a product is returned at the CERN, it is possible to trace back its date of
purchase.
The procedure we used to trace back returns is as follows. If the client
number is registered, the return is associated with the most recent purchase
by this client whose amount is larger than the quantity returned. If not,
but the budget number is known, the inspection is now done through the
issues purchased under this budget number. The rationale behind this is that
sometimes the client who receives the product is not the one who returns it.
Still, it is most likely that it is a client within the same project, i.e. with the
same budget number.
In several cases of return occurrences it is not possible to make a link
with any previous delivery, because the client and budget identification are
missing. In other cases, there is no registration of previous deliveries to these
clients. Possible explanations are that the products are being returned under
another project, or that they were purchased before the implementation of
the database.
Case 2 - Mail-order Company MOC

The second set of data comes from a mail-order company active in the Dutch
market, here referred to as MOC. The main product line is fashion, but house-
hold appliances and furniture are also found in MOC’s catalogues (consumer
goods). Customers can order by mail, phone, fax or internet. Merchandize
can be returned at no cost up to a limited amount of time after delivery, yet
the company sometimes also accepts later returns. The returns correspond
to the reimbursement returns type of the framework in Chapter 2.
MOC sends out two catalogues per year: spring/ summer and autumn/
winter. Before the season starts, the catalogue is sent to a subset of recurrent
customers, which can pose their orders even though the products are not
delivered before more than one month has passed. These orders help the
expert’s committee to point out an estimation of how many of each single
item will be sold over the whole season. This estimated value is periodically
updated together with the return percentage, as more information becomes
available on both sold and returned items.
MOC is very defensive towards making data on product demand and re-
turns open to the public. The data made available by MOC for this research
is as follows: estimated values for the average return percentage per time unit
both for fashion and hardware products and for the years 2000 and 2001.
Detailed information on the time unit and the return percentages remains
confidential throughout the chapter. Based on the provided information, 4

samples of time until return are going to be constructed and used for the
statistical testing: fashion-2000; fashion-2001; hardware-2000; hardware-2001
(more information can be found in Section 6.5). Then, for each of them is
tested whether, or not, it is reasonable that the underlining distribution con-
sists of a negative exponential. Due to the character of the data, the demand
and return processes are not investigated in isolation.
Case 3 - Refinery RF
The third case concerns a refinery in the Netherlands. The warehouse keeps
thousands of materials in stock. The warehouse owns the majority of these
materials while “customers” own the rest. In the latter situation, though
materials are kept in stock in the warehouse of the refinery, there are no
monetary exchanges directly associated with a particular demand or material
return. Nevertheless, the transactions are registered in the database. We
consider the spare parts inventory, which mainly consist of slow movers with
consumption in a maximum of four months per year. Product returns consist
of spares that maintenance personnel did not need for a given maintenance
job or rotable spare parts previously repaired to an as-good-as-new condition.
Thus, the returns reason correspond to the service returns category of the
framework of Chapter 2.
The company uses the SAP R/3 material management module for their in-
ventory control. In this system information on demand is stored on a monthly
basis. Accordingly, the immediate returns are netted. Only returns that occur
in future months are visible (unless they have been netted with new demands).
The return lag is determined by identifying for each return the latest month
with a positive (netted) demand. Unlike the CERN case, we did not have
client or budget codes for tracing. So, the tracing cannot be done with the
certainty of the CERN’s case. Yet, since in this case we are dealing with
slow-movers, the extracted data is credibly very close to the real process.
The inventory control policy of SAP employs basically the net demand and
ignores returns. The database consists of data from the last 4 years. We have
analyzed data from some 800 materials.
6.5 Data analysis

The scarcity of data and, in many cases, its confidentiality explain the lack
of publications on empirical analysis of return flows. Yet, this chapter tests
real data on demand and product returns. As mentioned previously, three
6.5. Data analysis 191
sets of data are considered: data on demand and in-company returns of the
warehouse of CERN in Switzerland and of two companies in the Netherlands,
a mail-order company and a refinery. Other products than the ones discussed
here were investigated. However, it was not always possible to trace back all
the return occurrences (see Section 6.4), and sometimes there were insufficient
observations to pursue satisfactory testing. In spite of these difficulties, the
process of data analysis and the findings have both academic and practical
implications, as discussed later.
Case 1 - CERN data analysis

For each product, the results of the analysis state whether, or not, it is rea-
sonable to accept that:
Hypothesis 1 : The demand process is an HCPP.

Hypothesis 2 : The return process is an HCPP.
Hypothesis 3 : The time elapsing between return and demand occur-
rences follows an exponential distribution.
Table 6.2: Statistical analysis of two products of CERN’s database.

Demand process Return process Time until return
No. Period Decision No. Period Decision No. Period Decision
obs. obs. obs.
Stabilizer gas
283 88-98 Rej. 12 88-97 Not rej. 12 88-97 Not rej.
36 96-98 Rej.
26 97-98 Not rej.
Large agenda
1043 88-99 Rej. 14 89-98 Not rej. 17 89-99 Not rej.
1009 88-98 Rej.
362 96-98 Rej.
253 97-98 Rej.
The decision to reject, or not, in each of the tests is based on a Type I error,
α = 0.05, i.e. the probability of rejecting the hypothesis given that it is
valid, equals 0.05. The results of the analysis of two products can be found in
Table 6.2. One of the products is a stabilizer gas, and the other is catalogued
as a large agenda. Both of these products have return rates around 5% of
the issued quantities and a large number of issues. One of the reasons these
Figure 6.2: Stabilizer gas: date of purchase vs. purchased quantity.
Figure 6.3: Large agenda: date of purchase vs. purchased quantity.

products were chosen for analysis is the possibility of tracking a sufficient

number of observations to pursue the required tests. In the case of unit
purchases and by assuming no lost or backlogged demand, one may opt for
carrying out the tests only for the demand and for the time until return. If the
first two hypotheses are not rejected, and making an analogy with queueing
systems, one can consider the demand-return process as an M/M/∞, i.e.
both the time between demand occurrences and the time until return follow
an exponential distribution. For such a queueing system the output flow will
eventually behave as a Poisson Process (Ross, 1996), dispensing in this way
the need for testing it.
In the case of the stabilizer, it is rejected that the demand process follows
a homogeneous Poisson Process for the periods 88-98 and 96-98, but not
when only data from 97 and 98 is taken into account. Thus, demand does not
remain stable for a period of 10 or even 3 years, but the opposite is acceptable
for a couple of years (see Figure 6.2). Regarding the return process, a HCPP
is not rejected. It is also statistically not rejected that the time until return
is exponentially distributed. begincenter
The large agenda is a seasonal product, as observed in Figure 6.3. There-
fore, for all the periods considered, the assumption of a HCPP demand is
rejected. Actually the demand occurrences do not even behave as a Poisson
Process. However, it is accepted that the time in market is exponentially
distributed and that the return process is a HCPP. One can explain this by
the fact that there are always enough products in the market. Making an
analogy with queueing systems this means that the server is always busy, so
the output flow does not depend so much on the demand process, but on the
service time.
Figure 6.4 is a dotplot of the ordered quantities of the product agenda.
One can notice that there are on the one hand many orders of low quantities
and on the other hand a few orders of large quantities. This puts forward
a division in two groups, namely small and large orders. Other graphical
representations, among which the histogram of Figure 6.5 helped us to make
the division at the order quantity 20: large orders are the ones of quantities
larger or equal than 20 (100 out of 1043). We proceed testing whether, or
not, the smaller orders follow an HCPP. The hypothesis is rejected. The
seasonality is likely to be the reason for this outcome. Regarding the large
order, a tailor-made procedure to integrate these orders in a possible ordering
policy is advisable.
Due to insufficient data of individual products, three groups of products
(aggregated data) are investigated next. The existence of many, but infre-
quent individual orders, justifies that the demand follows a Poisson Process
(Tijms, 1994). Therefore, the statistical analysis deals with the time until
return alone. Group 1 consists of several gases, Group 2 of products with
traceable returns, and Group 3 includes Group 2, as well as the returns of the
product ‘large agenda’ (Table 6.3).
Table 6.3: Outcome of the investigation whether, or not, time until return is
negative exponentially distributed.
Group 1 Group 2 Group 3
A B B C
No. obs. 18 40 37 53 35
Decision Not Rej. Rej. Not Rej. Rej. Not Rej.

Note: A= total number of observations; B = A minus some very large observations; C = B
minus the products which returned within 1 week.
It is accepted that the sample of 18 return inter-occurrence times in the

first group of products follows an exponential distribution. For the total of
40 return inter-occurrence times of the second group, the exponentiality is,
however, rejected.
To investigate whether the data could reasonably follow another distri-
bution, four probability-plots of the exponential and other distributions were
traced (Figure 6.6). The plot for the exponential points out that the three
large observations are disturbing a good adjustment of this distribution. In-
deed, it is not rejected that the data is exponentially distributed if the three
large return times are removed. These return times fall between 6 months and
one year after purchase, while all the other returns occur within less than 3
months after acquisition. The 3 “odd” return observations may be the result
of the termination of projects and a separated treatment is recommendable.
Yet, to decide how to deal with these situations in practice, it would be help-
ful to really have registered the return reason, as it was conceived by CERN’s
database, but regrettably not used.
Figure 6.4: Large agenda: dot-plot of the purchased quantity.
Figure 6.5: Large agenda: histogram of the purchased quantity.

For Group 3 of products, the very large observations were put aside in
analogy to the analysis of Group 2. Still, it is not statistically acceptable that
these products come back after an exponentially distributed lag. There is one
aspect that has not been considered until now. From an inventory perspective,
when products come back almost immediately, it is like the demand did not
occur in the first place. Therefore, one can distinguish between short-lag
returns and longer lag returns. The longer lag returns are the ones likely to
have more impact on inventory control, since they may suggest annulment of
order replenishments by the warehouse. The short-lag returns can be simply
subtracted from the demand process.
Figure 6.6: Group 2 of products: four probability plots.
The issues discussed in this paragraph give emphasis to the importance of

recording return reasons. Moreover, one would then be able to identify the
returns which feed demand, and those which not (e.g. wrong delivery, no
longer needed). This information could later be used in the forecast of returns
and it could help to design improved inventory policies. Following the above
arguments, it was investigated for Group 3 whether, or not, longer lag returns
follow an exponential distribution. Short-lag returns were those of one week
or less return time. In that case, the hypothesis was not rejected.
Case 2 - MOC data analysis

The demand in distance-selling environments (and especially on fashion) is
known to fluctuate considerably through time, dispensing the test to evalu-
ate whether, or not, the demand behaves as an HCPP. MOC provided the
estimated values for the average return percentage per period both for fash-
ion and hardware for the years 2000 and 2001. For example, the estimated
returns for fashion products in the year 2000 are spread over the first five pe-
riods after sales as follows: 5%, 70%, 22%, 2% and 1%. The absolute values
are confidential.
Because MOC protects detailed information on product demands and re-
turns, the testing samples are constructed based on the information provided
by MOC, i.e. the estimated return rates per time unit. Four samples were
put together, namely: fashion- 2000; fashion- 2001; hardware- 2000; hardware-
2001. The size of the sample is 100, which is not an unreasonable number of
returns in this setting. In other words, 100 return-lag occurrences were put
together according to the estimated return rates per period. It was tested
whether the time elapsed between return and demand occurrences followed
an exponential distribution.
Figure 6.7: MOC Fashion 2000: histogram of the return lag.
For all the four cases tested (fashion-2000; fashion-2001; hardware-2000; hardware-
2001) the hypothesis has been rejected. Figure 6.7 is a histogram of the
return lag for Fashion in the year 2000. A tentative explanation for non-
exponentiality is the presence of the short-lags. However, when the long-lags
are tested the hypothesis is still rejected. This is because of the thick tail
spread in time as manifested in Figure 6.7, even without the short-lags.
Case 3 - RF data analysis

In the case of the refinery, it was not tested if demand follows an HCPP
because of the lack of demand occurrences per product and because the data
is aggregated in monthly-netted demand. We scanned about 800 materials
and found 137 return occurrences. From those, about 8% could not be traced
back, which we therefore left out of consideration. We tested whether the
time elapsing between return and demand occurrences follows an exponential
distribution, when the lags are pooled over all the products. The assumption
is rejected for α ≥ 0.005, so it is rejected for the usual values of α. We also
have tested the hypothesis for non-owned products only (corresponding to
80% of the traced returns). The rationale behind this test has to do with
the non-existence of monetary exchanges by occasion of returns of materials
that are in any case owned by the customer. The customer is not going to
be refunded when he or she returns any products or materials. Thus, there is
no monetary incentive to return as soon as possible. Therefore, we suspected
that this would lead to a spread of a thick tail, conflicting with the exponential
distribution. Figure 6.8 illustrates the existence of a tail. However, the tail
effect still remains when only non-owned materials are considered.
From the data we noticed that for some products the stock level was above
the desired maximum level (as specified by RF) for a substantial amount of
time. The fact that returns are not taken into account in the inventory control
contributes to this overstock.
Figure 6.8: RF: histogram of the return lag.


In this chapter we presented a methodology for the statistical testing of the
common assumptions in the literature on inventory modelling with product
returns. We proposed statistical tests (see Figure 6.1) to test the following
hypotheses:
• The demand process is a Homogeneous Compound Poisson Process.
• The return process is a Homogeneous Compound Poisson Process.
• The time elapsing between return and demand occurrences follows a

negative exponential distribution.
With the developed methodology it is possible to carry out the desired

testing even with small sample sizes (below 20). Obviously, for extremely
small sample sizes no statistical test will be able to guarantee much rigor.
We applied the developed methodology to the following three sets of real
data.
Case 1: data from an in-company warehouse with internal returns in

Switzerland (CERN),
Case 2: data from a mail-order company with commercial returns in

the Netherlands (referred to as MOC), and
Case 3: data from an in-company refinery warehouse in the Netherlands

(referred to as RF).
We found that for some products the common assumptions on the theo-
retical models fit reasonably well. However, we also found products for which
they do not. Regarding the (non-)validity of the common assumptions, we
empirically observed the following.
A) Demand did not always behave as an HCPP due to seasonality

and/or non-stationarity (e.g. agenda and stabilizer gas- CERN).
B) In spite of non-stationary demand, there are products for which we

did not reject that the return process behaves as an HCPP (e.g. large
agenda- CERN). One can explain this by the fact that there are al-
ways enough products in the market. Making an analogy with queueing
systems this means that the server is always busy, so the output flow
depends on the service process rather than the demand process.
C) In some cases we did not reject that the time to return behaves
according to a negative exponentially distribution (e.g. stabilizer gas
and Group 1- CERN). So, for these cases the assumption that the time
to return is negative exponentially distributed appears to hold.
D) There are products for which the time to return did not seem to
behave as an exponential distribution. This is likely caused by a thick
tail (e.g. see RF analysis) or due to many short-lag returns (e.g. Fashion
2000- MOC).
We employed data on products of in-company warehouses (CERN and

RF) with internal returns and a mail-order company with commercial returns.
However, we believe that the findings and its implications are transferable to
a much larger group of demand-return processes. In the case of equipment
leasing, we conjecture that although the tail of the time to return distribution
may behave as an exponential, the same is unlikely to hold for the short-lag
returns. In maintenance settings, we believe that the validity of assumptions
depends on whether the maintenance is corrective or preventive. In the first
case the assumptions are likely to be valid.
6.6.1 Theoretical insights

There are 2 situations to consider.
1. When modelling, we assume independence between demands and re-

turns.
- If demand follows a Poisson distribution we have, using standard

results for a M/G/∞ queuing system, that the return process is
also Poisson. If, when modelling, independence between demands
and returns is assumed this means that it is appropriate to assume
Poisson process for both the demand and return streams.
- If demand is not a Poisson process, but the product is a fast-
mover and the time to return is a negative exponential distribution,
then we can use standard results from a G/M/∞ queuing system.
Using the high-traffic argument (‘servers’ are always busy), we have
that the return process is Poisson. This implies that the Poisson
distribution is a good approximation for the return process, as long
as the product is a fast-mover and the time to return is negative

exponentially distributed.
2. When modelling, we explicitly take into account dependence between

demands and returns.
In this case, and when the time to return is negative exponentially dis-
tributed, one can exploit the memoryless property. If you know at the
moment of sale if a product will return or not, the total number of prod-
ucts in the market that will return is sufficient to forecast the number of
returns. Note, however, that in practice this only occurs in very specific
situations. For example, if you not only sell to costumers, but also offer
service contracts to some of them, you know which products will return.
Another example for which you know in advance which products return
is when some products are sold while others are leased. A drawback of
this situation, however, is that the lease time is not likely to be negative
exponentially distributed.
If you do not know in advance if a sold product will return or not, you
cannot make use of the memoryless property (see the introduction of
Section 6.3).
6.6.2 Managerial implications

Returns can be said to be a “necessary evil” as they complicate inventory
control. Generally, it is important to know how much will come back and
when. This knowledge can be used to avoid superfluous replenishment orders
and the outdating of products. Therefore, it is important that companies
monitor returns, investigate return reasons, and estimate (the distribution
of) the return lag. Companies may then try to influence the return lag, for
instance to shorten it.
In the context of inventory management and control, we put forward the
following suggestions:
• First separate short-lag returns from long-lag returns. Then, proceed

0
to netting of the short-lag returns, i.e. (D = D − R), so it is as
returns have never occurred. Especially for long-lag returns: monitor
the return distribution, the return rates and register return reasons.
The latter helps to find return patterns that may be useful to find ways
of shortening the lag and/or to fine-tune the inventory control policy
(e.g. cancellation of replenishments when many products are expected
to return).
• When investing in product return information (e.g. tracking and trac-

ing) and its storage, the following should be taken into account.
If you know in advance which products are going to return, then, accord-
ing to the theoretical insights above, one should investigate whether, or
not, the return lag is likely to behave as a negative exponential distribu-
tion. If it is, one only has to keep track of the total number of products
in the market that will return to forecast returns.
If you do not know in advance which products are going to return, or if
the time to return is not negative exponential, you may wish to collect
advanced information on returns.
In a stationary environment investments on tracking and tracing may
pay off. However, in a non-stationary environment or if information is
not reliable, an advanced use of information may have lower performance
than a more parsimonious use. Chapter 7 investigates this issue in
depth.
It is important to note, however, that in practice there are other consid-
erations regarding the investment in information systems then inventory
management alone. We mention customer relationship, product respon-
sibility and control, and visibility.
From the process of acquiring data and from the data available for this re-
search we conclude that detailed data on product and material returns is
extremely scarce. When data on returns is found, it is most likely to be
aggregated with demand, limiting statistical analysis. Besides this, we antici-
pate that many companies are losing opportunities to improve their inventory
management due to lack of data on returns (see also Chapter 7).
6.6.3 Research needs

This chapter presented the first empirical analysis regarding the validity of
common assumptions in the literature on inventory management with product
returns. The data analysis only covered some specific situations and more
analysis is needed to generalize the results.
We have used several real examples to verify whether the time to return
behaves as a negative exponential. For some instances the hypothesis was
accepted, for others rejected. In some situations the impact of using the
wrong assumptions may be small, for others they may be large. Research is
needed to assess this impact. Also, for the cases for which the impact is large,
models need to break with the traditional assumptions in the literature on
inventory modelling with returns.
Inventories for seasonal products with returns are very difficult to manage.
Especially at the end of the season it is important to consider the product
returns, as the product’s value drops rapidly at the end of the season. This
should be considered in research on inventory modelling with product returns.
Anyway, researchers can be more explicit about the situations in which their
assumptions hold or not.
Chapter 7
Inventory Management with

product returns: the value of
information
In any decision situation,

the amount of relevant information available
is inversely proportional to the importance of the decision.
(Cooke’s Law), Paul Dickson
7.1 Introduction
Products and packaging return into the supply chain for a diversity of rea-
sons (see Chapter 2). For instance, beverage containers are returned by the
consumer to the retailer against reimbursement, single-use photo cameras are
turned in for the film to be developed, and now millions of products pur-
chased through mail-order-companies, e-tailers and other distant sellers, are
being returned every day.
There is a lot of money involved with the handling of product returns
(Rogers and Tibben-Lembke, 1999). One of the difficulties in handling re-
turns efficiently is that return flows are often characterized by a considerable
uncertainty essentially regarding time and quantity. If one could exactly know
how much is going to be returned and when, one would certainly benefit from
incorporating this perfect information a priori in the management of produc-
tion, inventory, and distribution. In many cases this is far from feasible (see
205
206Chapter 7. Inventory Management with product returns: the value of information
Trebilcock, 2002). Nevertheless, one may attempt to forecast the timing and
the amount of product returns. To do so, one has to hypothesize about the
return flow properties based on historic demand and return data.
There are not many papers that investigate the impact of information on
inventory management with product returns. The ones that do so assume
known return probabilities or consider specific cases where the most-informed
method leads to the best forecast. However, in environments where data is
scarce or unreliable, or in environments that are volatile, information may
be misleading. This study reports on the impact that such (mis)information
has on inventory management. This chapter identifies situations in which
the most informed method does not necessarily lead to the best performance.
Furthermore, the impact on inventory related costs of having inaccurate es-
timates of the time-to-return distribution is investigated for a wide range of
parameters. In addition, the implications for the practice of inventory man-
agement with returns will be discussed. This work has been reported in De
Brito and Van der Laan (2003).
The remainder of the chapter is structured as follows. First, Section 7.2
illustrates that estimates of the return rate can be massively erroneous in
practice. Next Section 7.3 reviews the literature on the topic. Subsequently
Section 7.4 elaborates on procedures to estimate the net demand. By means of
an exact analysis, Section 7.5 identifies situations in which more information
does not necessarily lead to better performance. Section 7.6 then presents
a simulation study to quantify the impact of misinformation with respect
to inventory related costs. In Section 7.7, the managerial implications are
discussed and Section 7.8 concludes by giving recommendations for future
research.
7.2 (Mis)information in practice

This chapter reports on the impact of (mis)information in an inventory system
with product returns. Misinformation means here that the properties we
assume for the future return flow do not correspond to the actual properties of
the future return flow. Further attention is paid next on how misinformation
can occur in a given system.
Figure 7.1 illustrates how the “information” arrives to the decision-maker.
A tool transforms data into “information” (e.g. system parameters), with
which the decision-maker controls the system. On the one hand, data may be
insufficient, abundant, ambiguous or conflicting (see Zimmerman, 1999) caus-
ing misinformation as input. On the other hand, the tool may be imperfect
creating some sort of distortion. The impact of misinformation on inventory
7.2. (Mis)information in practice 207
systems has for long been studied. Bodt and Van Wassenhove (1983) ex-
plained how firms could see the advantages of using Material Requirements
Planning (MRP) being drastically reduced due to misinformation. Gardner
(1999) showed how important is to choose an adequate forecasting tool while
investing in customer service levels. Some misinformation mechanisms are
well-known in inventory management, like the bullwhip effect, as introduced
by Lee et al. (1997). Misinformation is a general phenomenon, not specific for
systems with product returns. Yet, academics and practitioners agree that
managing systems with product returns is very demanding with respect to
neat information, which is sparsely available (see Gooley, 1998; Guide, 2000b;
De Brito et al.,2003; Kokkinaki et al., 2004).
Figure 7.1: Relation between data and decision-making.
In this chapter, the source or reason of misinformation is not a matter

of concern. The objective of this study is to evaluate the impact of misin-
formation on the system, given that misinformation occurs. Next a real-life
example illustrates that misinformation on product returns does happen in
practice. The company’s name is here disguised as MOC.
MOC is a large Western European mail-order-company (see also Chapter
5). Well before the start of the season, the company has to decide on what
to offer for sale in the upcoming catalogue. A great part of the total stock
to cover all the season has to be ordered many weeks in advance because
production lead times are long. This is many times the case in Europe or the
U.S. where manufacturing comes from Asia (see Mostard and Teunter, 2002).
The order decisions take into account the expected future demand and returns
for each line of products. Since there is however no data available on the sales
of the coming season, the forecast is not much more than a rough estimate.
In particular, a mail-order-company in the Netherlands uses estimates that
on average are more than 20% off the real return rate and sometimes even
more than 80%. The above illustrates that large mis-estimation of the return
rate does occur in practice. This chapter analyzes its impact on inventory
management.
7.3 Literature review

The literature dealing with product returns has been growing fast in the
last years. This literature falls in the general umbrella of closed-loop supply
chain management. For contextual contents, see Rogers and Tibben-Lembke
(1999); Guide and Van Wassenhove (2003); Dekker et al. (2004). Inventory
Management has received chief attention from the beginning until now, which
has brought about a richness of studies (see e.g. the following recent ones:
Van der Laan et al., 1999; Inderfurth et al., 2001; Fleischmann et al., 2002).
In spite of these and many other contributions, there are few articles that
simultaneously consider the forecasting of product returns, inventory man-
agement and information issues. Below follows a review on specific literature
contributing to close the aforementioned gap.
Goh & Varaprasad (1986) developed a methodology to compute life-cycle
parameters of returned containers by employing data on demand and returns.
The authors claim that a careful estimation of these parameters aids to effec-
tive inventory management. They apply their method to data on soft drinks
as Coca-Cola and Fanta from the Malaysia and Singapore markets. The
approach requires a time series of aggregated demand and another one of ag-
gregated return data. They used a 50-point time series of monthly data. The
precision of the method depends on how accurate the estimation of the return
distribution is. This estimation has the Box-Jenkins time series techniques
as basis. The authors call the attention of the reader to the fact that a data
set with less than 50 points is too short to employ the methodology. Large
time series should also be avoided. Therefore, they recommend a time series
of 50 points, or a 4-year period of data coming from a stable market environ-
ment. Our research contributes to contrasting situations: with imperfect or
misleading data.
Kelle & Silver (1989) proposed four forecasting procedures of net demand
during lead time in the case of reusable containers. Every procedure has a
different level of information requirement. The least-informed method uses
the expectation and the variance of the net demand together with the prob-
7.4. Forecasting methods 209
ability of return. The most-informed one calls for individual tracking and
tracing of containers. The authors evaluate the forecasting methods taking
the most-informed method as a benchmark. The analysis, however, applies
only to the case of perfect information on the return parameters. This chapter
employs the methods proposed by Kelle and Silver (1989) but it elaborates
on the potential impact of (mis)information.
Toktay et al. (2000) consider the real case of new circuit boards for Ko-
dak’s single use remanufactured camera. The goal is to have an ordering pol-
icy that minimizes the procurement, inventory holding and lost sales costs.
A six-node closed queueing network is employed to represent Kodak’s sup-
ply chain. Accordingly, the returns of cameras depend on past sales by a
return probability and an exponential time lag distribution. Some of the pro-
cedures of Kelle and Silver are used to predict the unobservable inventory
at the customer-use network node. The authors compare several forecasting
methodologies with different levels of information. This chapter provides an
exact and more general analysis of the impact of (mis)information. Further-
more, the four methods of Kelle and Silver are compared numerically for a
wide range of parameter values with respect to misinformation. Therefore,
besides an intra-method comparison, here an inter-method analysis is also
presented.
Marx-Gómez et al. (2002) consider a case of photocopiers that may re-
turn to the producer after being used. The authors put forward a method
to forecast the number and time of returned photocopiers. Firstly, data is
generated according to two scenarios: successful vs. not so successful return
incentives. These scenarios, together with expert knowledge, constitute the
basis for developing a set of forecasting rules. The expert evaluates factors,
such as demand and life cycle parameters. An extended approach of the
model is suggested by allowing a follow-up period to self-learn the rules. This
neuro-fuzzy process calls for data on demands and returns. Besides this, one
should notice that the method of Marx-Gómez et al. depends on a priori
knowledge on product returns, which have been acquired by the producer.
Overall one can conclude that the proposed procedures are very demand-
ing with respect to reliable data. This chapter takes the more realistic per-
spective of misinformation and analyzes its impact.
7.4 Forecasting methods

The main notation used in this chapter is as in Table 7.1.
Consider a single product, single echelon, periodic review inventory sys-

Table 7.1: Notation.

t, the current time period; t = 0, 1, 2, . . .
L, order leadtime
DL (t), r.v. representing the demand during the interval
[t + 1, t + 2, . . . , t + L]
RL (t), r.v. representing the returns during the interval
[t + 1, t + 2, . . . , t + L]
N DL (t), the net demand during the leadtime (equal to DL (t) − RL (t)).
pj , the probability of an item returning after exactly
j = 1, . . . , n periods, n being the largest j for which
the return probability is larger than zero.
the probability of an item ever being returned, i.e. nj=1 pj .
P
p,
ui , demanded amount during period i and i ≤ t.
yi , the total amount returned in each previous period i ≤ t.
S, base-stock level.
k, safety factor
tem. Each individual demand returns with probability p according to some

return distribution. We assume that this process does not change in time
(stationary case). Returns are immediately serviceable. The order lead time
is a fixed constant L. Demands that cannot be satisfied immediately are fully
backordered. Following the methodology of Kelle and Silver (1989), for each
time period t a base-stock policy is applied. If the net lead time demand,
N DL (t) = DL (t) − RL (t), follows a normal distribution the optimal base
stock level is given as
p
S = E[N DL (t)] + k · Var[N DL (t)] (7.1)
with E[N DL (t)] and Var[N DL (t)] being the expectation and variance of the
net demand during lead time. The safety factor k is determined according to
some desired performance level (see e.g. Silver et al., 1998).
In order to estimate E(N DL (t)) and Var[N DL (t)], the four methods first
put forward by Kelle and Silver (1989) are used. Apart from the expecta-
tion and variance of the demand during lead time (E[DL (t)] and Var[DL (t)]
respectively), each method requires a different level of information for esti-
mating the lead time returns. Below, in increasing order of information needs,
the four methods, denoted A–D, are listed (see Appendix 1 for details).
Method A - Average behavior

This method requires the following information:
• p, the overall return probability, i.e., the probability that a product

is being returned eventually.
Figure 7.2: Method A (it uses the overall return probability to compute the
expected returns coming from demand during the lead time.
This method is an approximation in the sense that all the returns during
the lead time are assumed to be perfectly correlated with the demand during
that same lead time and independent of previous demands (see Figure 7.2).
Returns during the lead time are estimated through the overall return prob-
ability (it is like if a product is to return, it returns instantaneously). Then,
demand and returns during the lead time are simply netted (see also Chapter
6). No historical information is used with respect to demands and returns, so
in a static environment the resulting base stock level is constant in time.
Method B - Return distribution

Suppose that we are at the end of period t. This method requires information
on previous demand per period and the knowledge of the return distribution
as follows:
• ui , demanded amount during period i ≤ t.
• pj , the probability of an item being returned exactly after j periods,
j = 1, ..., n, with n being the largest j for which the return proba-
bility is non-zero.
This method makes use of the return probabilities to determine the num-
ber and moment of the returns during the order lead time (see Figure 7.3).
Figure 7.3: Method B (it uses past demand and return probabilities p0j s to
compute returns coming from past demand and it employees expected values
for returns coming from demand during the lead time).
Method C - Return distribution & return information per period

Suppose that we are at the end of period t. In addition to the requirements
of method B, this method makes use of observed data on aggregated returns:
• ui , demanded amount during period i, i ≤ t.

bility is non-zero.
• yi , the total amount of returned products in each period i, i ≤ t.
This method updates the number of items that have returned, which are
observed, and takes this into account to compute future returns (see Figure
7.4). Thus, Method C aims at improving Method B by taking into account the
Figure 7.4: Method C (it uses information on past demand and past returns to
compute the returns coming from past demand; it uses information on current
demand to compute returns coming from period t; and it uses expected value
to compute returns coming from demand during the lead time).
correlations between the observed aggregated returns in recent periods and

the future lead time returns. An analytical method, however, is not available,
so Kelle and Silver (1989) developed an approximation (see Appendix 1) that
is accurate as long as the purchased amount is relatively large and the return
probabilities are positive for several periods (in practice n ≥ 4).
Method D - Return distribution & tracked individual returns

Let t be the last observed period. Besides the requirements of method B this
method requires to track back in what period each individual return has been
sold:
• ui , demanded amount during period i, i ≤ t.

bility is non-zero.
• Zit , the observed total number of product returns from each
past purchase ui , i < t.
To employ this method, one has to be able to trace back the returns, i.e. to
know from which order period each specific return is coming from (see Figure
7.5).
Given perfect information this method makes optimal use of all relevant
information.
Figure 7.5: Method D (it uses information on past demand and observed
returns from each past demand to compute the returns coming from past
demand; it uses information on current demand to compute returns coming
from period t; and it uses expected value to compute returns coming from
demand during the lead time).
Summarizing, all four methods for estimating the net demand during the
order lead time, make use of the expectation and variance of demand. Ad-
ditionally, each method has different requirements with respect to product
return information. Method A is the least demanding: only an estimate of
the return rate is needed. Apart from the return rate, Method B requires the
return time distribution. On top of that, Method C also needs a record of
the aggregated returns per period. Finally, to employ Method D one needs
to invest in a system that allows to scan individual returns and track them
back to the period in which they were originally sold.
Given perfect information one expects that the method that uses more
information outperforms the methods that use less. The remainder of the
chapter investigates how the various methods perform in presence of misin-
formation. First, Section 7.5, identifies situations in which Method B may
outperform the most informed method, Method D. Then Section 7.6 com-
pares Methods A–D with respect to inventory related costs and confirm the
findings of Section 7.5 by means of a simulation study.
7.5 Forecasting performance

This section analyzes the relative performance of Methods B and D given
misinformation. Methods A and C contain approximations which makes them
less interesting for an exact analysis. Besides that, Method A is a rather naive
forecasting method, which is not expected to perform very well in general (this
will be confirmed in Section 7.6) and the performance of Method C tends to
be very close to that of Method D (Kelle & Silver, 1989).
Since, given perfect estimation, Method D is expected to lead to the best
forecast of the expected lead time net demand, we use Method D given perfect
information as a benchmark for our study. Given imperfect estimation, both
Methods B and D will do worse than the benchmark, but it is important to
know whether there are situations in which Method B outperforms Method
D. Therefore, with respect to the expected lead time net demand we would
like to compare

EB [N DL (t)] − ED [N DL (t)] and E [N D (t)] − E [N D (t)]
b b
D L D L
Here ‘b’ denotes a forecast, based on forecasts {b

pi } of {pi }. However, the above
expression depends on observations Zi of past product returns coming from
demand ui . Accordingly, we take the conditional expectation with respect to
the return process given the history of the demand process, ER|D :
7.5. Forecasting performance 215
n o
F{E} = ER|D E [N D (t)] − E [N D (t)] − E [N D (t)] − E [N D (t)]
b b
B L D L D L D L
(7.2)
Note that if F{E} < 0 then, on average, Method B outperforms Method D
with respect to the expected net leadtime demand and vice versa if F{E} > 0.
Similarly, the performance measure with respect to the variance of the lead
time net demand is
n o
F{V } = ER|D V (< t)B − V (< t)D − V (< t)D − V (< t)D (7.3)
b b
To make the analysis more readable we define πi = t−i

P
j=1 pj
P Pt−1
and we write ‘ ’ for ‘ i=t−n+1 ’. In section 7.5.1 we analyze the expectation
of the net demand during leadtime and in section 7.5.2 its variance.
7.5.1 Analysis regarding the expectation of lead time

net demand
Define
b B [N DL (t)] − ED [N DL (t)],
E{BD} = E
b D [N DL (t)] − ED [N DL (t)]
E{DD} = E
Obviously, E{BD} = 0 implies that, given misinformation, Method B performs

at least as good as Method D and E{DD} = 0 implies that, given misinforma-
tion, Method D performs as good as the benchmark and therefore performs at
least Method B will not have a better performance. Conditioning on the signs
of E{BD} and E{DD} we have 4 remaining cases, EXP1–EXP4, to analyze:
Case EXP1: E{BD} > 0 and E{DD} > 0

Case EXP2: E{BD} < 0 and E{DD} < 0
Case EXP3: E{BD} < 0 and E{DD} > 0
Case EXP4: E{BD} > 0 and E{DD} < 0
From the analysis of EXP1–EXP4 in Appendix 2 we conclude that, on av-

erage, Method B results in a better forecast of the expected lead time net
demand in the case we consistently under- or overestimate the return proba-
bilities pj 0 s.
From cases EXP1–EXP4 it is also clear that the difference between the
methods increases as the Ri get bigger. In other words, for higher return
rates the differences between the methods are also larger (with respect to the
forecasts of the expected lead time net demand).
7.5.2 Analysis regarding the variance of lead time net demand

Using expressions (7.13) and (7.23) in Appendix 1 we define
d B [N DL (t)] − VarD [N DL (t)]
V{BD} = Var
Ph b i
= ui Ri (1 − R
bi ) − (ui − Zi )Qi (1 − Qi )
d D [N DL (t)] − VarD [N DL (t)]

V{DD} = Var
P h i
= (ui − Zi ) Q b i (1 − Q
b i ) − Qi (1 − Qi )
We attempted an analysis of the performance measure with respect to the

variance, as defined in (7.3), similarly to the analysis pursued for the expected
value. However, we did not obtain straightforward inequalities (see De Brito
and van der Laan, 2003). It is very likely that with respect to the variance
there are situations for which Method B outperforms D, but also situations
for which D outperforms B. We enhance the analysis by looking again at the
special case that we misestimate the overall return probability, i.e. pb = a · p
for some a : 0 ≤ a ≤ 1/p, while the shape of the return distribution remains
intact. We can write R bi = aRi and Q b i = aRi /(1 − aπi ). Still it is difficult
to analyze F{V } directly because of the summations. So, instead we analyze
the individual coefficients of ui in V{BD} , V{DD} ,and F{V } . Denote these by
ν{BD},i , ν{DD},i , and φi . If we define Zi = βi ui , 0 ≤ β ≤ 1 we can write
h i
Ri 1−βi
ν{BD},i = Ri a(1 − aRi ) − 1 − 1−π i 1−πi
h i
a(1−aRi /(1−aπi )) 1−Ri /(1−πi )
ν{DD},i = (1 − βi )Ri 1−aπi − 1−πi
 h i
aRi 1−πi


 φ1,i = aRi (1 − aRi ) − 1 − 1−aπi 1−aπi (VAR1)

 φ2,i = −φh1,i (VAR2)


i
φi = φ3,i = R i 2 1 − Ri
− a 1 − aRi 1−πi
− a(1 − aR i )

 1−πi 1−aπi 1−aπi
(VAR3)




φ4,i = −φ3,i (VAR4)

Figure 7.6 gives an example of how the preference regions are constructed
through ν{BD},i , ν{DD},i , and φi in terms of a and Ri . Note that the boundary
7.5. Forecasting performance 217
between the preference regions of Method B and D only depends on φi (the

coefficient of F{V } ) and not on the coefficients of V{BD} and V{DD} . Since φi
does not depend on Zi the analysis is independent of a particular realization
of observed returns.
Figure 7.6: Preference regions for F{V } in terms of a and Ri ; πi = βi = 0.6,

p = 0.7.
Not all the combinations of Ri and a in Figure 7.6 are feasible, since
Ri ≤ p − πi and a ≤ 1/p. Figures 7.7a-f depict the preference regions for the
feasible area only and for various values of p and πi . From these figures it
appears that Method B performs better as Ri gets smaller, πi gets larger or
p gets smaller, particularly if a < 1. Note that the Ri tend to be small if the
base of the time-to-return distribution, n, is large compared to the lead time
L. At this time we would like to stress that the analysis of this section does
not depend on time t, since equations (7.12)–(7.3) do not depend on t, nor
did we make any assumptions on realizations of the demand process {ui } and
observed returns {Zi }.
In summary, we conclude that it is not at all obvious that Method D,
which is the most informed method, performs better than Method B. In fact,
we have identified situations in which Method B performs better, on aver-
Figure 7.7: Preference regions in terms of a and Ri for various values of πi

and p.
(a) πi = 0.3; p = 0.40 (b) πi = 0.3; p = 0.70
(c) πi = 0.3; p = 0.95 (d) πi = 0.9; p = 0.95
(e) πi = 0.6; p = 0.70 (f) πi = 0.6; p = 0.95

7.6. Cost performance 219
age, with respect to the expectation and variance of lead time net demand.
In particular, when all return probabilities are underestimated or all return
probabilities are overestimated Method B has opportunities to outperform
Method D. In the next section we will quantify the impact of misinformation
with respect to costs.
7.6 Cost performance

7.6.1 Experimental design
In order to quantify the impact of misinformation on inventory related cost
performance we conducted a simulation study. We consider holding and back-
order costs as described later. The experiments are based on the inventory
system that was introduced in Section 7.4 and are conducted in the follow-
ing manner. Each period t we draw the cumulative demand D(t) from a
normal distribution with mean µD , variance σD 2 , and coefficient of variation
σD
cvD = µD (values are rounded to integers; negative numbers are treated as
zero). For each individual item of this cumulative demand we determine the
time to return based on the pre-specified return probabilities, {pj }. Each
period, estimates of the expectation and variance of the net demand during
lead time are computed according to one of methods A–D. These estimates
are subsequently used to compute the base stock level, S. At the end of each
period, overstocks are charged with a holding cost $ h per item, per period,
whereas stockouts are penalized with $ b per occurrence. At the end of each
simulation experiment we calculate the total average cost per period as the
total average holding plus backorder costs per period. Note that all methods
use estimates, {bpj }, of the real return probabilities, {pj }, since the latter are
not known. The same holds for the overall return probability, p, which is
estimated as pb.
Each simulation experiment consists of at least ten simulation runs of 5.000
periods, preceded by a warm-up run of the same length. The simulation stops
as soon as the relative error in the total average costs is less than 1%. In order
to make a better comparison among simulation experiments we make use of
common random numbers. Please note that in this study all parameters are
assumed to be constant over time.
Based on the estimates E[N DL (t)] and Var[N DL (t)] of the mean and
variance of net lead time demand the base stock level S is computed as in
(7.1). Assuming that the net demand during lead time is normally distributed,
the cost optimal value of the safety factor, k, satisfies G(k) = 1 − hb , where
G(.) is the standard normal distribution (see Silver et al., 1998).
The time-to-return distribution {pj } consists of two components: The

overall return probability p and the conditional time-to-return probabilities
{pj } given that the item returns. The (unconditional) time-to-return prob-
abilities then are defined as pj = p · pj , j ∈ [0, 1, . . . , n]. In the simulation
experiments we use two conditional time-to-return distributions. The first one
is a geometric distribution with conditional expected return time T = 1/q,
i.e. pj = q(1 − q)j−1 , j = 1, 2, . . . , ∞. The second is a discrete uniform dis-
tribution with conditional expected return time T = (n + 1)/2, i.e. pj = 1/n,
j = 1, 2, . . . , n. Please note that in this study all parameters are constant in
time.
7.6.2 Numerical study

In the case of perfect information, i.e. p̂ = p and p̂j = pj , method D will
outperform all other methods since it is using all of the available information in
a correct way. In order to investigate the effect of misinformation, we consider
two types of errors in the parameter estimates. The first is a misspecification
of the overall return probability, p, while the expected return time is preserved.
The second is a misspecification of the conditional expected time-to-return, T .
This affects the shape of the time-to-return distribution, but the estimated
overall return probability is preserved. For example, suppose that the real
time to return distribution is given by {p1 , p2 , p3 } = {0.2, 0.1, 0.2}. Then
an estimate of {b p1 , pb2 , pb3 } = {0.1, 0.05, 0.1} would have the same conditional
expected time-to-return (2.0), but a 50% lower estimated return probability.
An estimate of {b p1 , pb2 , pb3 } = {0.3, 0.1, 0.1} would have the same overall return
probability (0.5), but a lower conditional expected time-to-return.
For numerical comparisons we define the relative difference of some method
m with respect to method D as follows:
costs method m − costs method D
relative difference = × 100%
costs method D
Perfect information
Comparing the four methods in the case of perfect information (Tables 7.2–
7.3) we observe here that method D is indeed superior to the other methods,
although the differences with respect to methods B and C are not significant
for p ≤ 0.8 (less than 1 percent). The performance of Method A is extremely
poor. It uses the assumption that all lead time returns are correlated with the
lead time demands. This causes a systematic underestimation of the variance
in the lead time net demand, especially for high return rates and large lead
Table 7.2: Comparison of Methods A–D in the case of perfect information;

Geometric time-to-return distribution; µD = 30. ∗ ) results below 1% are not
significant.
Methods
A B C D
relative difference∗ costs
Base case:
p = 0.5, T = 1.67, cvD = 0.2
L = 4, h = 1, , b = 50 24.9% 0.2% 0.0% 26.07
p = 0.8 111.8% 0.9% 0.4% 22.85
p = 0.9 129.3% 3.1% 0.6% 21.24
T = 2.50 35.7% 0.3% 0.1% 27.98
T = 5.00 56.7% 0.4% 0.0% 30.70
cvD = 0.4 36.3% 0.0% 0.0% 44.80
cvD = 0.8 24.8% 0.0% 0.0% 83.00
L=8 10.6% 0.2% 0.1% 32.94
L = 16 4.1% 0.1% 0.0% 43.37
b = 10 11.0% 0.1% 0.2% 19.18
b = 100 36.7% 0.1% -0.2% 28.69
times. Because of this poor performance we will not consider Method A in

the remainder of the numerical study.
Misinformation on the overall return probability

We define the relative error in the estimated return probability as follows
p̂ − p
relative error = × 100%
p
Tables 7.4–7.5 show that Method B structurally outperforms the more
information intensive methods C and D in the case of misinformation of plus
or minus 10% or more. Under an error of −20%, the relative difference can
be as large as -5%. Under an error of +20%, the relative difference can be
as large as -20% for a return probability of 0.5 and as large as −60% for a
return probability of 0.8. The performance of Method C is fairly close to the
benchmark, although it usually performs worse.
The relative cost differences become bigger as the return rate goes up. This
is not surprising as with increasing returns also the impact of (mis)information
increases. Note that the cost improvement of Method B with respect to the
benchmark can be as large as 60% under 20% overestimation and a return
probability of 0.8.
Table 7.3: Comparison of Methods A–D in the case of perfect information;

Uniform time-to-return distribution; µD = 30. ∗ ) results below 1% are not
significant.
Methods
A B C D
relative difference∗ costs
Base case:
p = 0.5, T = 2.50, cvD = 0.2
L = 4, h = 1, b = 50 44.9% 0.3% 0.2% 28.59
p = 0.8 171.4% 2.4% 1.1% 26.23
p = 0.9 181.4% 4.3% 1.8% 24.83
n = 4.50 59.6% 0.7% 0.3% 31.14
n = 8.50 65.0% 0.3% 0.1% 32.68
cvD = 0.4 65.0% 0.2% 0.1% 50.22
cvD = 0.8 48.3% 0.2% 0.1% 93.61
L=8 18.2% 0.3% 0.2% 34.93
L = 16 8.6% 0.2% 0.0% 44.80
b = 10 24.1% 0.6% 0.2% 21.07
b = 100 68.2% 0.6% 0.2% 31.33
An increase of the lead time makes the relative differences smaller. A

longer lead time results in a larger portion of the lead time returns that come
from expected issues during the lead time itself. Methods B, C, and D treat
this category of returns in exactly the same way, so the relative differences
become smaller.
Using a similar argument we expect that an increase in the expected time-
to-return, T , results in larger relative differences. As the expected time-to-
return increases the portion of lead time returns that come from expected
issues during the lead time decreases. The relative differences therefore in-
crease, as Tables 7.4–7.5 also show.
An increase in demand variation also has a negative impact on the relative
differences. This probably is because more demand uncertainty also leads to
more return uncertainty. More information on the return distribution than
has less impact on cost performance. The difference between the methods
thus becomes smaller as the uncertainty in returns increases.
The analysis of Section 5.5 showed that in the case of misinformation of
the return probability Method B performs better than Method D with respect
to the expected lead time net demand, while the difference between the meth-
ods increases as the return rate increases. With respect to the variance the
analysis was less straightforward. Sometimes Method B outperforms Method
Table 7.4: Comparison of Methods B–D in the case of misestimation of the

return probability; Geometric time-to-return distribution; µD = 30. ∗ ) results
below 1% are not significant.
Methods
B C D
p̂ error rel. difference∗ costs
Base case: 0.40 -20% -1.9% 0.3% 35.15
p = 0.5, T = 1.67, 0.45 -10% -0.9% 0.2% 29.07
cvD = 0.2, L = 4, 0.55 +10% -5.2% 0.7% 35.54
h = 1, b = 50, k = 2.054 0.60 +20% -17.9% 2.4% 84.33
0.64 -20% -5.3% 1.2% 42.45
p = 0.8 0.72 -10% -3.9% 1.3% 31.18
0.88 +10% -43.1% 8.5% 92.26
0.96 +20% -67.1% 6.6% 692.04
0.40 -20% -3.1% 0.5% 37.10
T = 2.50 0.45 -10% -1.5% 0.3% 31.06
0.55 +10% -7.7% 1.2% 37.13
0.60 +20% -26.7% 4.4% 86.12
0.40 -20% -0.6% 0.0% 90.71
cvD = 0.8 0.45 -10% -0.2% 0.0% 85.39
0.55 +10% -0.8% 0.0% 87.21
0.60 +20% -3.5% 0.2% 101.41
0.40 -20% -0.8% 0.1% 87.60
L = 16 0.45 -10% -0.7% 0.1% 61.95
0.55 +10% -5.1% 0.5% 170.99
0.60 +20% - 6.9% 0.6% 902.68
0.40 -20% -1.7% 0.3% 38.11
b = 100 0.45 -10% -0.8% 0.2% 31.79
(k = 2 .326 ) 0.55 +10% -5.7% 0.8% 40.11
0.60 +20% -19.8% 2.9% 105.85

return probability; Uniform time-to-return distribution; µD = 30. ∗ ) results
below 1% are not significant.
Methods
B C D
p̂ error rel. difference∗ costs
Base case: 0.40 -20% -2.1% 0.3% 37.23
p = 0.5, T = 2.50, 0.45 -10% -0.7% 0.3% 31.45
cvD = 0.2, L = 4, 0.55 +10% -4.9% 0.4% 36.63
h = 1, b = 50, k = 2.054 0.60 +20% -18.5% 1.1% 76.85
0.64 -20% -4.9% 0.9% 45.24
p = 0.8 0.72 -10% -2.4% 1.2% 34.08
0.88 +10% -34.0% -1.0% 72.54
0.96 +20% -56.5% -3.8% 394.72
0.40 -20% -3.6% 0.6% 39.97
n = 4.50 0.45 -10% -1.5% 0.4% 34.21
0.55 +10% -7.4% 0.9% 39.01
0.60 +20% -28.6% 2.8% 79.98
0.40 -20% -0.6% 0.0% 100.03
cvD = 0.8 0.45 -10% -0.1% 0.0% 95.47
0.55 +10% -0.4% 0.0% 97.04
0.60 +20% -2.8% 0.1% 107.60
0.40 -20% -1.0% 0.1% 88.86
L = 16 0.45 -10% -0.7% 0.1% 63.27
0.55 +10% -5.7% 0.3% 164.27
0.60 +20% -7.9% 0.3% 857.01
0.40 -20% -1.9% 0.3% 40.39
b = 100 0.45 -10% -0.6% 0.4% 34.40
(k = 2 .326 ) 0.55 +10% -5.3% 0.4% 41.05
0.60 +20% -20.5% 1.4% 93.51
D and sometimes it is the other way around. From the numerical results we
conclude that the effect of the expected lead time net demand dominates the
effect of the variance.
Misinformation on the conditional expected time-to-return

We define the relative error in the estimated conditional time to return as
T̂ −T
T × 100%. For the geometric distribution this leads to
1/q̂ − 1/q q
relative error = × 100% = ( − 1) × 100%
1/q q̂
and for the uniform distribution we have
(n̂ + 1)/2 − (n + 1)/2 n̂ − n
relative error = × 100% = × 100% .
(n + 1)/2 n+1
According to Tables 7.6–7.7, misinformation of the conditional expected
time-to-return has little effect if the return probability is small. For p = 0.8
though, both Methods B and C perform much better than the benchmark and
are far more robust with respect to misinformation (Figure 7.4). Again the
relative differences with respect to the benchmark is positively correlated with
the return probability and expected time-to-return, and negatively correlated
with the lead time and demand variation.
Figure 7.8: Misinformation (-30%,+30%) of the expected time-to-return (Ge-

ometric time-to-return distribution, p = 0.8, q = 0.6, µD = 30, cvD = 0.2, L =
4, h = 1, b = 50).

expected time-to-return; Geometric time-to-return distribution; µD = 30. ∗ )
results below 1% are not significant.
Methods
B C D
T̂ error rel. difference∗ costs
Base case: 1.33 -20% -6.5% -7.0% 25.34
p = 0.8, T = 1.67, 1.50 -10% -0.9% -2.0% 23.56
cvD = 0.2, L = 4, 1.83 +10% -0.2% -1.1% 23.18
h = 1, b = 50, k = 2.054 2.00 +20% -3.8% -5.1% 24.33
1.33 -20% -0.4% -0.6% 26.46
p = 0.5 1.50 -10% -0.1% -0.3% 26.20
1.83 +10% 0.1% 0.0% 26.05
2.00 +20% -0.5% -0.2% 26.13
2.00 -20% -12.6% -13.3% 30.69
T = 2.50 2.25 -10% -2.3% -3.8% 26.96
2.75 +10% 0.0% -2.0% 26.41
3.00 +20% -4.8% -6.2% 27.88
1.33 -20% -2.9% -3.0% 39.95
cvD = 0.8 1.50 -10% -0.4% -0.8% 38.50
1.83 +10% -0.2% -0.7% 38.30
2.00 +20% -2.3% -2.9% 39.46
1.33 -20% -3.0% -3.6% 32.63
L = 16 1.50 -10% -0.5% -1.2% 31.65
1.83 +10% 0.1% -0.7% 31.35
2.00 +20% -3.4% -4.3% 32.63
1.33 -20% -7.8% -8.6% 28.30
b = 100 1.50 -10% -0.8% -2.3% 25.86
(k = 2 .326 ) 1.83 +10% 0.4% -0.8% 25.37
2.00 +20% -4.1% -5.4% 26.75

expected time-to-return; Uniform time-to-return distribution; µD = 30. ∗ )
results below 1% are not significant.
Methods
B C D
T̂ error rel. difference∗ costs
Base case: 3.50 -22% -32.4% 5.3% 48.03
p = 0.8, T = 4.50, 4.00 -11% -7.8% 2.7% 34.58
cvD = 0.2, L = 4, 5.00 +11% -4.6% -6.6% 33.40
h = 1, b = 50, k = 2.054 5.50 +22% -16.5% -18.0% 38.45
3.50 -22% -3.4% 0.6% 32.75
p = 0.5 4.00 -11% -0.4% 0.4% 31.57
5.00 +11% -0.3% -0.7% 31.49
5.50 +22% -2.6% -3.0% 32.30
3.50 -22% -12.7% 2.0% 63.38
cvD = 0.8 4.00 -11% -1.9% 0.8% 55.29
5.00 +11% -3.1% -3.9% 56.38
5.50 +22% -9.8% -10.7% 61.25
3.50 -22% -24.4% 4.5% 54.86
L = 16 4.00 -11% -6.7% 2.3% 43.69
5.00 +11% -4.6% -6.4% 42.66
5.50 +22% -16.5% -18.4% 49.24
3.50 -22% -3.4% 0.7% 36.21
b = 100 4.00 -11% -0.7% 0.4% 34.84
(k = 2 .326 ) 5.00 +11% -0.2% -0.6% 34.56
5.50 +22% -2.6% -2.9% 35.48
Extensions
Here we present a new adaptation of Method A and an extension of Method
B. The disappointing performance of Method A is mainly due to the assump-
tion that all returns during the lead time are correlated with the demands
during the lead time. The other extreme, which we propose here, is to as-
sume that all lead time returns are independent of the lead time demand. The
expectation of lead time demand does not change, but the variance is then
given as VarA [N DL (t)] = (1 + p2 )Var[DL (t)]. This method, denoted Method
A0 - Average behavior with independence - will overestimate the variance in
the lead time net demand, but it will generate less (costly) backorders (see
Figure 7.5). Only if p is close to 1, Method A outperforms Method A0 .
Figure 7.9: Performance of Method A both under perfect correlation and zero
correlation (Geometric time-to-return distribution, p = 0.5, q = 0.6, µD =
30, cvD = 0.2, L = 4, h = 1, b = 50).
It was seen that Method B was rather robust under imperfect information
of the time-to-return distribution. This suggests that Method B could be
applied as a standard with a simple uniform distribution. The information
requirement is then reduced to parameters p and n and there is no need to
forecast the pi ’s individually. This has significant managerial implications (as
we stress below).
7.7 Discussion and managerial implications

Given perfect information, Method A performs in general very poorly and is
not recommended for practical implementation. Thus, to only have knowledge
7.7. Discussion and managerial implications 229
on average behavior does not seem to be cost effective. Including information

on the return distribution, however, does seem to provide a sufficient level
of sophistication, as the performance of Method B shows. With respect to
inventory related costs, the differences between Methods B, C and D are not
such that they justify investments in recording detailed return data.
Both for misinformation on the return rate and on the return distribution,
the differences between the methods become smaller with the decrease of the
return probability and the increase of the lead time. Obviously, if there are
few returns forecasting of the lead time returns is not really an issue. To
understand the latter one should note that when L is large, most of the items
that return during the lead time were also purchased during the lead time.
The forecast of this type of returns is not based on historical data, so all
Methods B–D give exactly the same forecast.
With respect to misestimating the return rate, in general it is better to
underestimate the return rate than to overestimate, since stockouts are usually
much more costly than overstocks. Therefore, if an interval estimate of the
return rate is available, one may opt to use a value that is closer to the
lowerbound rather than the upperbound.
The most robust method given misestimation of the return rate is Method
B. Method B systematically outperforms Methods C and D if the return rate
is misestimated by merely 10%,or more. The cost differences are particularly
high if return rates are overestimated. With respect to misestimating the
conditional time to return distribution, Method C and again Method B are
much more robust than Method D.
The above results strongly suggest that Method B has a sufficient level
of sophistication both in the case of perfect estimation and imperfect estima-
tion. In the latter case Method B is far more robust than the most informed
method, Method D. For the inventory management of the mail order com-
pany of Section 7.2 this means that orders only have to be based on the return
distribution and realized demand per period, but there is no need to track
individual returns.
Finally we observed that Method B was fairly robust given misespecifi-
cation of the time-to-return distribution. Therefore the following practical
adaptation of the use of Method B was proposed. Companies may opt to
disregard the shape of the return distribution and to simply use Method B
with a flat shape, i.e. a uniform distribution. The advantages are huge in
terms of information spare. The company no longer has to estimate the re-
turn probability per period, i.e. the individual pi ’s. It suffices to estimate the
overall return probability and the max return period, i.e. p and n. These are
actual the parameters for which companies more comfortably give estimates.
7.8 Summary of conclusions and further research

This chapter has reported on the impact of (mis)information on forecasting
performance and performance with respect to inventory costs by analyzing
four forecasting methods as proposed by Kelle and Silver (1989). All methods
make use of the expectation and variance of the demand, but different levels
of information with respect to returns. The least demanding method, Method
A only uses the return rate. Method B also requires the return distribution.
Method C additionally uses a periodic record of returns. Finally, Method D
needs to track back the period in which each individual product return was
sold.
Given perfect information, forecasting performance increases as the level
of information increases, Method D naturally being the best method. Yet,
from the analysis we have concluded that Method B presents a reasonable
level of sophistication given perfect information and is exceptionally robust
given misinformation in comparison with the other methods. Furthermore
since the Method B is quite insensitive to the shape of the return distribution,
the companies can simply employ the uniform return distribution, i.e. using
Method B with a flat shape, that is by fitting an uniform distribution.
Method D does not appear to be very robust given misestimation. This
leads to the conclusion that companies are not likely to recover investments
on advanced return data with inventory savings, especially in volatile en-
vironments. Naturally, it is worth further investigating the value of return
information, with respect to multiple criteria, like for instance production
scheduling and human resource management (see Toktay et al., 2004 for a
comparison of the methods with respect to order variability, “the bullwhip
effect”).
The huge gap in performance between Methods A and B suggests that
refinements of Method A are possible. In this chapter an enhanced Method
0
A, i.e. Method A was put forward, where the variance during the leadtime
is underestimated. In addition, there is an opportunity to look further for
similar simple methods with low information requirements but a reasonable
performance.
In the analysis of misinformation, there was a static situation where a
systematic error was introduced. This is natural for small errors, since it is
difficult in general to reason whether observations indicate trend changes or
mere stochastic behavior. However, large variations could be dealt with by
an adaptive method. This is a topic for further research.
7.8. Summary of conclusions and further research 231
Appendix 1: Forecasting methods

Proofs can also be found in Kelle and Silver (1989).
We define
• µD , the expected value of Demand D

2 , the variance of Demand D
• σD
which we use sometimes to simplify notation.

Method A - Average behavior
This method requires the following information:
• p, the overall return probability, i.e., the probability that a product

is being returned eventually.
The expectation and variance of lead time net demand according to Method
A are equal to
EA [N DL (t)] = (1 − p)E[DL (t)] (7.4)
and its variance is
VarA [N DL (t)] = (1 − p)2 VarA [DL (t)] + p(1 − p)E[DL (t)] (7.5)
Proof:
By definition, we have that,
EA [N DL (t)] = EA [DL (t)] − EA [RL (t)] (7.6)

VarA [N DL (t)] = VarA [DL (t)] + VarA [RL (t)]
− 2cov[DL (t), RL (t)] (7.7)
Let us first compute EA [RL (t)] and VarA [RL (t)] (Part A of the proof) and
later the covariance, i.e. cov[DL (t), RL (t)] (Part B of the proof).
Part A
RL (t) is a mixed binomial random variable, with a random number DL (t) of
trials and known success probability p. Let each trial be called Xm , m =
1, ..., DL (t) . Each trial has a probability of success p, independently of any

other trials. Thus,
P[Xm = 1] = p
P[Xm = 0] = 1 − p
and
var(Xm ) = p(1 − p)
Since
DP
L (t)
RL (t) = Xm
m=1
it follows that (see Tijms, 2003, pg. 435)
EA [RL (t)] = E[Xm ]E[DL (t)]

VarA [RL (t)] = E2 [Xm ]Var[DL (t)] + Var[Xm ]E[DL (t)]
EA [RL (t)] = pE[DL (t)] (7.8)

2
VarA [RL (t)] = p Var[DL (t)] + p(1 − p)E[DL (t)] (7.9)
Part B
Let us now compute the covariance, cov[DL (t), RL (t)].
By definition, we have that
cov[DL (t), RL (t)] = EA [DL (t)RL (t)] − EA [DL (t)]EA [RL (t)] (7.10)
with
∞
i P
i!
pk (1 − p)i−k
P
EA [DL (t)RL (t)] = iGDL (t) k (i−k)!(k−1)!
k=1 i=1
∞ i
i!
pk (1 − p)i−k
P P
= iGDL (t) k (i−k)!(k−1)!
i=1 k=1
P∞
= [iGDL (t) ][ip]
i=1
∞
i2 GDL (t)
P
= p
i=1
2
= p [E
A [DL (t)]]
= p VarA [DL (t)] + [EA [DL (t)]]2

where we used that, E(Y)=i · p when Y is a random variable binomially

distributed with parameters (i,p).
Substituting, the above and 7.8, in 7.10, we obtain the following
cov[DL (t), RL (t)] = p [VarA [DL (t)]] (7.11)
To finalize the proof, let us substitute 7.8, 7.9, and 7.11 in 7.6 and 7.7 as
adequate. We get that
EA [N DL (t)] = (1 − p)EA [DL (t)]

VarA [N DL (t)] = (1 − p)2 VarA [DL (t)] + p(1 − p)EA [DL (t)]
Method B - Return distribution

Suppose that we are at the end of period t. This method requires information
on previous demand per period and the knowledge of the return distribution
as follows:
• ui , purchased amount during period i ≤ t.
bility is non-zero.
The expectation and variance of lead time net demand according to Method
B are
t t+L−1
" #
X X
EB [N DL (t)] = L · µD − ui Ri + µD Ri (7.12)
i=im i=t+1
t
X
2
VarB [N DL (t)] = σD + ui Ri (1 − Ri )
i=im
t+L−1
X
2
(1 − Ri )2 + µD Ri (1 − Ri )

+ σD (7.13)
i=t+1
with im = max {1, t − n + 1} and Ri the success probability of having prod-

ucts from past demand ui returning during any of the the lead time periods
t + 1, t + 2, · · · , t + L:

 0,
 for i < t − n + 1
Pjm
Ri = p
j=1 t−i+j , for im ≤ i ≤ t
 Pjn p ,

for t < i < t + L
j=1 j
where jm = min {L, n + i − t} and jn = min {n, t − i + L}
Proof:
Let zi,j be the number of products sold in period i and returned in period j,
j = 1, ..., n. The zi,j , j = 1, ..., n are a sequence of multinomial trials with the
respective distribution as follows:
n
P
P (zi,1 = k1 , ..., zi,n = kn , zi,∞ = ui − kj )
j=1
n
P
ui − kj
ui !
pk11 ...pknn p∞
j=1
= n
P
k1 !...kn !(ui − kj )!
j=1
where zi,∞ represents the number of products sold in period i that will never
return.
We are interested in computing the total number of returns during the lead
time. Let Wi be the total number of returns during the lead time periods, i.e.
t + 1, ...t + L, coming from ui . We have that
Wi = zi,t+1−i + ... + zi,t+L−i , i5t

Wi = zi,1 + ... + zi,t+L−i , t<i<t+L
Being Wi a sum of multinomial trials, Wi follows a Binomial distribution, with

ui number of trials, and with a success probability given by Ri as previously
defined.
Let now ‫ג‬t,L be the total number of returns during the lead time periods,
t + 1, ..., t + L. ‫ג‬t,L is the sum of independent Wi with i ≤ t + L − 1.
Let us distinguish between returns that come from periods until t, and returns
that come from lead time demand.
First part: i ≤ t
The ui , for i ≤ t, are observed values. Thus the expected value and
variance of the correspondent total number of returns, that come back during
the lead time, are as follows:
t t
EB [RL (t)](1)
P P
= E[Wi ] = u i Ri
i=1 i=im
t t
VarB [RL (t)](1) =
P P
Var[Wi ] = ui Ri (1 − Ri )
i=1 i=im
with im = max{1, t − n + 1}
Second part: i > t

Here, ui , for i > t, have not been observed yet. Therefore, we use E[DL (t)] =
µD and Var[DL (t)] = σD 2 . We obtain
t+L−1 t+L−1
EB [RL (t)](2)
P P
= E[Wi ] = µD Ri
i=t+1 i=t+1
t+L−1 t+L−1
VarB [RL (t)](2) = Ri2 σD
2 + µ R (1 − R )
P P
E[Wi ] = D i i
i=t+1 i=t+1
Finally, the expected value of the net demand during the lead time is given
as follows:
EB [N DL (t)] = E[DL (t)](2) "− EB [RL (t)](1) + E[RL (t)] (2)

#
t+L
P Pt t+L−1
P
= µD − u i Ri + µD Ri
i=t+1 i=im i=t+1
" #
Pt t+L−1
P
= L · µD − ui Ri + µD Ri
i=im i=t+1
and the variance is given by

t
2 +
P
VarB [N DL (t)] = σD ui Ri (1 − Ri )
i=im
t+L−1
2 (1 − R )2 + µ R (1 − R )
P
+ σD i D i i
i=t+1
where equality 7.5 was used and with im = max{1, t − n + 1}

Method C - Return distribution & return information per period

Suppose that we are at the end of period t. In addition to the requirements
of method B, this method makes use of observed data on aggregated returns:
• ui , purchased amount during period i, i ≤ t.

bility is non-zero.
• yi , the total amount of returned products in each period i, i ≤ t.
The expected lead time net demand according to method C is
EC [N DL (t)] = EB [N DL (t)] − c T −1 (y − E(y)) (7.14)
and the variance is
VarC [N DL (t)] = VarB [N DL (t)] − c T −1 c† (7.15)
with y = (yt , yt−1 , ..., yt−n+2 ) the vector of recent aggregated returns, T the
covariance matrix of vector y and T −1 its inverse matrix; and c a vector of
covariances defined, for j = 1, 2, . . . , n − 2, by
P
t−1
cj = Cov i=im W i y t−j+1
Pt−j
ut−n+i pn−j+1−i jm=1
Pm
= − i=i m
pn−i+m
with c† being the transpose of vector c and jm = min {i, L}

And, the elements of matrix T , Tj,k = cov(yt−j+1 , yt−k+1 ), are defined as
follows.

t−k
P
T = − pt−j+1−i pt−k+1−i , for j = 1, ..., n − 2, j ≤ k ≤ n − 2

j,k




 j=ij
Tk,j = Tj,k , for k < j

 t−k
P
pt−j+1−i (1 − pt−j+1−i ), for j = 1, ..., n − 2

 Tj,j =


j=ij
with ij = max {1, t − j + 1 − n}
Proof: This method makes use of the observed total amount returned in
each period up to t, i.e. yt , yt−1 , ..., yt−n+2 . Recent aggregated returns are
correlated with the returns during the lead time. Thus, this method makes
use of the conditional expectation and variance of these returns as follows:
 
t−1
X
EC [RL (t)(1) ] = E  Wi | yt , yt−1 , ..., yt−n+2  (7.16)
j=im
 
t−1
X
VarC [RL (t)(1) ] = V ar  Wi | yt , yt−1 , ..., yt−n+2  (7.17)
j=im
The above expressions cannot be expressed in an exact analytical form. How-

ever, a multidimensional normal vector gives a good approximation, if demand
is reasonably large and if n ≥ 4. The approximation allow us to write the
expression 7.16 and 7.17 as follows (see Kelle and Silver, 1989):
t−1
X
EC [RL (t)(1) ] = E( Wi ) + c T −1 (y − E(y)) (7.18)
j=im
t−1
X
VarC [RL (t)(1) ] = V ar( Wi ) − c T −1 c† (7.19)
j=im
with y = (yt , yt−1 , ..., yt−n+2 ), T , and c as defined before, and T −1 being the
inverse matrix of T, and c† being the transpose of vector c.
Thus, the expected value and variance of the total returns during the lead
time corresponds to
EC [N DL (t)] = EC [RL (t)(1) ] + ut Rt + EB [RL (t)(2) ]

= EB [RL (t)] − c T −1 (y − E(y)) (7.20)
VarC [RL (t)] = VarC [RL (t)(1) ] + ut Rt (1 − Rt ) + VarB [RL (t)](2)

= VarB [RL (t)] − c T −1 c† (7.21)
Please note that EC [RL (t)] is as EB [RL (t)] but corrected with the term
c T −1 (y − E(y)). Similarly, VarC [RL (t)] is as VarB [RL (t)] but corrected
with the term c T −1 c† . Therefore, we have that the expected value and the
variance of the net demand during the lead time is as given next.
EC [N DL (t)] = EB [N DL (t)] − c T −1 (y − E(y))

VarC [N DL (t)] = VarB [N DL (t)] − c T −1 c†
Method D - Return distribution & tracked individual returns

Let t be the last observed period. Besides the requirements of Method B this
method requires to track back in what period each individual return has been
sold:
• ui , purchased amount during period i, i ≤ t.
bility is non-zero.
• Zit , the observed total number of product returns from each
past purchase ui , i < t.
To simplify notation we use next Zi instead of Zit .

The expected value and variance of the lead time net demand according
to method D are respectively
ED [N DL (t)] = L · µD
" t−1 t+L−1
#
X X
− (ui − Zi )Qi + ut Rt + µD Ri (7.22)
i=im i=t+1
t−1
X
2
VarD [N DL (t)] = σD + (ui − Zi )Qi [1 − Qi ] + ut Rt [1 − Rt ]
i=im
t+L−1
X
2

+ σD [1 − Rt ] + µD Ri [1 − Ri ] (7.23)
i=t+1
Ri
with Qi = Pt−i , the success probability associated with the binomial
1− j=1 pj
conditional random variable Wi given Zi , the returned amount from past
demand ui , i ≤ t.
t−1
X
2
i=im
t+L−1
X
2

+ σD [1 − Rt ] + µD Ri [1 − Ri ] (7.24)
i=t+1
Proof:
As before, we are interested in computing the expected value and the variance
of the net demand during the lead time. For returns coming from the demand
during the lead time itself, the expected value and variance of these returns is
given as in Method B, i.e. by EB [N DL (t)(2) ] and VarB [N DL (t)(2) ] (also used
by Method C, for these returns). For returns coming from previous demand,
Method D makes use of conditional return probabilities, as it observes the
returns of any particular period up to time t.
The Wi , total return during the lead time, is conditioned by Zi , the ob-
served total number of product returns from each ui . We have, for i < t,
E[Wi | Zi ] = ui − Zi
Ri
Var[Wi | Zi ] = Pt−i
1 − j=1 pj
Thus, we have that
t−i
X
ED [RL (t)] = (Ui − Vi )Qi + ut Rt + EB [N DL (t)(2) ] (7.25)
i=im
t−1
X
VarD [RL (t)] = [(Ui − Vi )Qi (1 − Qi )] +
i=im
= ut Rt (1 − Rt ) + EB [N DL (t)(2) ] (7.26)
with
Ri
Qi = Pt−i
1− j=1 pj
Finally, we obtain
" t−1 t+L−1

#
X X
ED [N DL (t)] = L · µD − (ui − Zi )Qi + ut Rt + µD Ri
i=im i=t+1
t−1
X
2
i=im
t+L−1
X
2

+ σD [1 − Rt ] + µD Ri [1 − Ri ]
i=t+1
Appendix 2: Forecasting performance

In the following please note that
πi = t−i
P
j=1 p j , Qi = Ri /(1 − πi ),
and
ER|D {Zi } = ui πi .
For simplicity write ‘ ’ for ‘ t−1
P P
i=t−n+1 ’.
Analysis regarding the expectation of lead time net demand

Using 7.12, i.e.
t t+L−1
" #
X X
EB [N DL (t)] = L · µD − ui Ri + µD Ri
i=im i=t+1
and 7.22, i.e.
ED [N DL (t)] = L · µD
" t−1 t+L−1
#
X X
− (ui − Zi ) Qi + ut Rt + µD Ri
i=im i=t+1
we have
b B [N DL (t)] − ED [N DL (t)]
E{BD} = E
Xh i
= (ui − Zi ) Qi − ui R
bi (7.27)
b D [N DL (t)] − ED [N DL (t)]
E{DD} = E
X
= (ui − Zi ) Qi − Qbi (7.28)
Now we analyze the cases EXP1–EXP4 as defined in section 7.5.1 with respect
to the performance measure F{E} as defined in relation (7.2). Let X % Y
denote that method X performs at least as good as method Y.
Case EXP1: E{BD} > 0 and E{DD} > 0
Proof:
We have that
Xh i
F{E} = (ui − Zi ) Qi − ui R bi − (ui − Zi )(Qi − Q
bi )
Xh i
= (ui − Zi )Q b i − ui Rbi
Xh i
= (ui − Zi )Q b i − ui Rbi
" #
X R
bi
= (ui − Zi ) − ui R bi
(1 − π bi )
" #
X R
bi
= (ui − ui πi ) − ui Rbi
(1 − π bi )
X (1 − π)

= (ui Rbi − ui R bi (7.29)
(1 − πbi )
X
(1 − π)

= (ui Ri
b −1 (7.30)
(1 − π bi )
and since E{DD} > 0, follows from 7.28 that
Qi > Q
bi (7.31)
because Zi , the observed total number of product returns from past demand
ui , is by definition smaller or equal than ui (so ui − Zi is always larger or
equal than 0).
It follows immediately from 7.31 that it is infeasible to consistently overes-
timate the return probabilities pj 0 s. In the opposite case (to consistently
underestimate the return probabilities pj 0 s) it follows that F{E} ≤ 0 meaning
that B % D.

Case EXP2: E{BD} < 0 and E{DD} < 0
Proof:
With similar steps as used in 7.30, we have that
Ph
bi 1 − (1−π)
i
F{E} = (ui R (1−b
πi ) (7.32)
and since E{DD} < 0, follows from 7.28 that
Qi < Q
bi (7.33)
It follows immediately from 7.33 that it is infeasible to consistently under-

estimate the return probabilities pj 0 s. In the opposite case (to consistently
overestimate the return probabilities pj 0 s) it follows that F{E} ≤ 0 meaning
that B % D.

Case EXP3: E{BD} < 0 and E{DD} > 0 =⇒ B % D
Proof:
We have that

bi (1 + (1 − π) ) − 2Ri
X
F{E} = ui R (7.34)
(1 − π
bi )
and since E{DD} > 0, follows from 7.28 that
Qi > Q
bi
It follows immediately from 7.35 that it is infeasible to consistently overes-

timate the return probabilities pj 0 s. In the opposite case (to consistently
underestimate the return probabilities pj 0 s) it follows that F{E} ≤ 0 meaning
that B % D.

Case EXP4: E{BD} > 0 and E{DD} < 0 =⇒ B % D
Proof:
We have that

P (1−π) b
F{E} = ui 2Ri − (1 − πi ) )Ri
(1−b (7.35)
and since E{DD} < 0, follows from 7.28 that
Qi < Q
bi
It follows immediately from 7.36 that it is infeasible to consistently under-

estimate the return probabilities pj 0 s. In the opposite case (to consistently
overestimate the return probabilities pj 0 s) it follows that F{E} ≤ 0 meaning
that B % D.

In summary, if we consistently overestimate or underestimate the return prob-
abilities pj’s, then B % D.
From cases EXP1–EXP4 it is also clear that the difference between the
methods increases as the Ri get bigger. In other words, for higher return
rates the differences between the methods are also larger (with respect to the
forecasts of the expected lead time net demand).
Chapter 8
Reverse Logistics research:

what does the future bring?
Change is the law of life.

And those who look only
to the past or present
are certain to miss the future.
John F. Kennedy
8.1 Introduction
The field of Reverse Logistics has been expanding fast with many technical
contributions, case studies and some theoretic literature arising from research
(see previous chapters). Yet, it remains an academic field in formation, where
concepts and theories are not as vivid as in better established research fields
like supply chain management (see Melissen and De Ron, 1999).
In this chapter, we propose a reflection on the future development of
reverse logistics, mainly as an academic field. In this way, we assist in giving
priorities to research in the field. At the end we evaluate reverse logistics
issues with respect to research impact and complexity.
We carried out a Delphi Study with an international panel of academics
working on reverse logistics issues as a means to reflect on the future of reverse
logistics as an academic field. The Delphi study consisted of two rounds of
anonymous questionnaires with an intermediary feedback phase. In short,
245
246 Chapter 8. Reverse Logistics research: what does the future bring?
every member received information on the answers to the first questionnaire,

which they could rank and rate.
The overall analysis allows us to make recommendations concerning re-
search and teaching about reverse logistics. Furthermore, we find evidence
of different contributions to the field, depending on whether the researcher
is a senior, junior or a researchers working “accidentally” in the area. In
addition, we explore the impact of the differences the between European and
North American researchers.
This chapter is organized as follows. First, Section 8.2 describes the em-
ployed methodology. Section 8.3 reports on an exploratory study that served
as input for the Delphi study. Section 8.4 presents the specific design of this
Delphi study. In Section 8.5 we display the profile of reverse logistics re-
searchers and the profile of the participants of the study. Then in Section 8.6
we discuss the insights that this study brings about reverse logistics. Section
8.7 put into evidence the specific added-value to the field of senior, junior
and accidental researchers. In Section 8.8 we split the analysis according to
the European versus North American participants. Before the end, we glance
over the rest of the world. The last section puts together the conclusions. An
earlier report on this study has appeared in De Brito and van Wassenhove
(2003).
8.2 Methodology
The employed methodology for the main investigation was a Delphi study,
which is described in more detail below. Prior to the Delphi study, we carried
out an exploratory study, which brought awareness to the potential pitfalls of
the main study. For the exploratory study, we employed the Nominal Group
Technique (NGT).
Both NGT and Delphi are techniques to structure communication among
a group of individuals. The NGT is appropriate for the exploratory phase
as it allows us to follow the process in loco at the same time that helps
communication to flow in a structured way. By following the process, insights
are gained on how to structure communication in follow-up studies. In our
case, we obtained insights on how to build up the Delphi questionnaire. The
Delphi itself is a suitable method to reflect on the future of a field, as it calls
upon the views of experts.
8.2.1 An exploratory study

For the exploratory study, we employed the Nominal Group Technique (NGT)
to structure the interactions of a group of reverse logistics researchers. Their
task was to come up with ideas on “What are promising areas for future
research on reverse logistics?” The study took place during the annual work-
shop of the European working group on Reverse Logistics, called RevLog (see
RevLog, online), in the end of the summer in 2002. The outcome of this
exploratory study is reported in the Section 8.3.
NGT is a technique to structure communication among a group of indi-
viduals. To summarize, after having a question/task, the discussion process
follows the following steps(with some variations): Step 1- Silent generation
of ideas; Step 2 - Round-robin record of ideas; Step 3 - Clarification of each
idea and discussion within each group; Step 4- Individual voting (ranking or
rating) and wrap-up. The process has been designed 1) to assure creativity;
and 2) to balance participation among members (see Delbecq et al., 1975).
The NGT is a technique to structure group meetings, and therefore, it
will achieve structured output. Nonetheless, this is still raw material and
very likely not rich in detail, since participants are stimulated to be brief in
stating their ideas. Any analysis of the results should take into account the
group formation and how the discussion proceeded. Special caution should go
in taking overall conclusions. For instance, to pool rankings/ratings is very
dangerous. Note that groups vote from their own set of ideas.
8.2.2 The Delphi study

Delphi is another method to structure communication among a group of indi-
viduals, also called panel. Delphi studies are well fit for purposes like problem
solving, policy-making and proposal review among others (Linstone and Tur-
off, 1975). Experts or stakeholders on the subject in study constitute the
panel. Applying the Delphi method basically involves the following. Sev-
eral rounds of questions and feedbacks are sent to the panel. Each feedback
includes both the mild and the extreme panel standpoints relatively to the
previous round of questions. Individual comments/answers remain anony-
mous to the group. Through the process, each member of the panel has the
chance to refine his/her own answers or comments.
Since the first published Delphi study in the sixties (Dalkey and Helmer,
1963), Delphi has been employed in many fields of research like technology,
marketing, supply chain management and so on (see Boks and Tempelman,
1997; Smit and Popma, 2002; Akkermans et al., 2003). For a review on Delphi
studies, we refer to Linstone and Turoff (1975).
In the literature, many have pointed out the advantages over face-to-face
meetings. For instance, in face-to-face meetings, winning an argument is
sometimes more important than to look for a better understanding of the
problem. With the Delphi, since anonymity is guaranteed, a participant can
quietly change his/her mind without losing face (see Martino, 1972). Besides
this, other advantages are (see Kenis, 1995): to balance participation (avoid-
ing the follow-the-leader effect); to avoid a reactive chain of ideas (avoiding
stagnation around the first welcomed idea); and to make participants free
of the burden of interpersonal relationships (being free from social-emotional
concerns and accordingly behavior). Sceptics of the Delphi study have also
exposed the disadvantages, especially directed to the original Delphi format.
The critics accuse the method of being illusorily over simplistic and an easy
temptation for being sloppy (see Slackman, 1975). Two pillars of the Del-
phi study are 1) the concept of expert; 2) the questionnaires and respective
summarized feedbacks. Thus the researcher(s) carrying out Delphi studies
has to be extremely careful with respect to the sampling process and the
constructing of the questionnaires and follow-up summaries.
For this study, experts on reverse logistics and related areas were chosen,
based on their relevant scientific research and other experts’ judgement. The
exploratory study, carried out a priori, delivered us helpful input on potential
traps, and guided the questionnaire as well (see Section 8.3). Furthermore,
it helps to check the questionnaire by posing for instance the following ques-
tions: are the questions too vague?; are the questions too precise?; or, perhaps
biased?; is the question by any chance too demanding, or is it within reach
for an expert? (see Dillman, 1978). It is imperative to fine-tune the question-
naire with the expert panel. Besides this, the validity of Delphi cannot be
guaranteed if the feedbacks do not represent the true answers of the panel.
8.3 An exploratory study

8.3.1 Implementation and discussion
The study took place during the annual workshop of the European working
group on Reverse Logistics, called RevLog (see RevLog, online), in the end of
the summer in 2002. Fifteen participants were involved in this study, 13 be-
longing to RevLog. The two guests, outside the RevLog group, were a senior
and a more junior researcher. All the participants of this study, either RevLog
members or not, were familiar with the work of the RevLog group (see Dekker
et al., 2004). The participants were less-junior (e.g. Assistant Professors) and
more-junior researchers (PhD candidates), who use to publish in internation-
8.3. An exploratory study 249
ally renowned scientific journals. The senior researchers of the RevLog group
did not participate. This was thought to be more adequate. RevLog senior
researchers are experts in Reverse Logistics, who would dominate the results,
even by only causing unintentional inhibition in the more junior researchers.
Since this phase had an exploratory character, we favored to give freedom to
the more junior researchers. An earlier report on this study has appeared in
De Brito (2003).
Three groups of 5 people each, were put together. In each group, a mod-
erator was nominated. In making the groups we aimed to have a diversity of
research backgrounds in each group. During the workshop, each group met
at the same time, but in separate rooms. The moderators welcomed the par-
ticipants stating that the meeting would serve to gain insights on the future
of reverse logistics research. The discussion followed the general four steps of
the NGT. The moderators did also participate in all the steps; they count as
participants as well.
The moderators met before the workshop to discuss the formulation of
the question. The questions were brought down to the following two:
• What are promising areas for future research on reverse logistics?
• What are promising research areas to include in a follow-up RevLog

project?
The moderators auto-tested the question. The test showed that the second
question would lead to highly individually-centered research subjects while
the first would give more room for other research subjects than the ones each
respondent is individually involved with. Since this was important to the
objective of discussing promising areas for future research on reverse logistics,
and some were invited participants (not members of RevLog) the first question
was chosen.
Figure 8.1 represents the sheet of paper given to the participants. Partic-
ipants were instructed to list any idea on promising ideas for future research
on reverse logistics, on a sheet of paper. They were instructed to synthesize
their ideas in a few words (one short statement). The participants were given
5 minutes to generate ideas individually and in silence, which was respected.
After that, the ideas were recorded in a round-robin fashion. This means
going around the table asking for one idea from one member, recording it
on a flip-chart, and then another idea from the next participant, and so on.
Arguments for/against a given idea, or the motivation for it, should not be
given at this point in time. It is also up to the participant that gives the idea,
the wording of his/her own idea.
Figure 8.1: Representation of the sheet of paper given to the participants.
Every group generated between 12 to 25 ideas for promising areas for future
research on reverse logistics. After the round robin, there was a clarification of
the ideas listed in the flip chart. Everyone should understand what is meant
by it. Arguments in favour/against could be expressed (not necessarily by
who has suggested it). The participants used diverse arguments against/in
favor of each item. Table 8.1 lists just a few
Table 8.1: Employed arguments during NGT.
“This area is growing” - “This area publishes”

“That issue is really important for practice”
“I personally would not do such research” / ”I want to do something I like”
“I wouldn’t want to work with psychologists”
“We should have competence to do it” - “That area is outside our scope”
“We can include one item a bit out of our competence, but not more than one.”
“I do not see why the military needs special focus” - ”Too broad” - “Too detailed”
“Many problems in this area have not been solved or even stated until now”
“Important issue, even more important in the future”
“Already much research has been done in this area”
During the voting phase, participants were first asked to write down the
five most important items and rank them from 1 to 5 (1 being the most
important). A At this point, an opportunity was given to re-discuss some
of the items that are perceived as receiving “too many” or “too few” votes.
Finally participants were once again asked to pick the five most important
items and to rate them in a 0 to 10 scale (being 10 the most important).
When two or three participants considered an item as of rank 1 or 2,
8.3. An exploratory study 251
later more participants rated this item. To this may have contributed, on
the one hand, valid arguments during the second round of discussion, and, on
the other hand, an inherent search for group consensus. In contrast, we also
observed that some items were highly but single-rated. This seems to indicate
that some participants do not give up from the subjects that are most dear
to them.
8.3.2 Observations and input for the Delphi Study

It is not our purpose here to report on the ideas listed as being the promising
areas for future research on reverse logistics. It is to be kept in mind that
this was an exploratory study to thwart the way for the main study, i.e. the
Delphi. The objective was to gain insights on possible pitfalls or on helpful
hints to conduct the Delphi study.
We observed that many of the items did not have to do with logistics
alone. Words like integration, coordination, acquisition, marketing, and ac-
counting were among the ideas of the participants as promising areas for
future research on Reverse Logistics. This announces that researchers in the
field are conscious of the multidisciplinary character of reverse logistics. Be-
sides this, participants anticipate the development of the field in the direction
of Integrated Reverse Logistics (in words like co-ordination, strategic) or even
Extended Reverse Logistics (in words like acquisition and marketing). This
would demand co-ordination with other research fields and therefore alliances
with researchers with other core expertise than reverse logistics. With this in
mind, the invited panel for the Delphi included not only specialized reverse
logistics researchers but also other researchers in related areas.
From some of the arguments used during the NGT discussion (see Table
8.1), one notices that the ideas are biased by personal opinions and preferences
(e.g. “I personally would not do such research”). Therefore, the Delphi
questionnaire was designed with a group of questions on the participant’s
view of the state of the art of reverse logistics research. On the one hand,
participants were asked about what they considered to be 1) important past
research contributions, and 2) major limitations in reverse logistics research
(see Table 8.2). Furthermore, when asking on promising areas for future
research, participants were asked about areas of 1) their own expertise, but
also 2) outside their own expertise.
8.4 Design and implementation of the Delphi

A team of four researchers put together this research project. The general
structure of the Delphi study was as follows: a first questionnaire was sent
to a panel of experts; after receiving the answers, those were processed and
a second communication was sent to the panel. The second communication
served to inform the panel on the obtained answers. At the same time, the
panel was asked to rank and rate the answers.
Before selecting the panel of experts, we discussed the range of the ques-
tionnaire. A group of international academics active in reverse logistic issues
was chosen since the object of the study is primarily the development of re-
verse logistics as a research field. All the 35 invited academics accepted to
participate. The participants are associated either with European or North
American universities.
The first questionnaire included three groups of six questions about re-
search, theory/modelling, and pedagogy. The aspects covered by the research
and modelling/theoretical questions were: major research contributions so
far, limitations of current research, topics for future research, research trends
and breakthroughs. The questions about pedagogy were about the main chal-
lenges in teaching and on what could ease the task of teaching (see Table 8.2).
Many of the questions were related. For instance, some of the questions about
research (in general) and theory/modelling issues were very much the same,
stimulating answers enriched by further details. The questionnaire was tested
for wording and understanding PhD students on the logistics and other areas.
The improved questionnaire was given to all the participants.
Table 8.2: Topics covered by the Delphi questionnaire.

Research Theory/Modelling Pedagogy
· major research contributions so far;

· limitations of past research; · challenges;
· topics for future research (own expertise);
· topics for future research (others’ expertise); · utilities to ease
· research trends; the teaching tasks;
· research breakthroughs;
During the analysis of the answers, we proceeded to grouping in some

degree (e.g. when the same was said by means of different words), but several
answers were unique. Both were reported to the panellists in the 2nd ques-
tionnaire, which was in itself a feedback on the first. The answers for each
8.5. Reverse Logistics researchers and the panel 253
question were handed over, varying between 8 and 22 items depending on the
degree of concordance. At this stage, the participants were generally asked
to rank and rate, the top answers, with respect to importance. Regarding
trends, the panel was asked to rank them first and then rating them with
respect to impact. A similar procedure was asked to breakthroughs, which
were rated with respect to likelihood. Concerning pedagogy, the panel was
asked to rate the answers according to the priority that they should have.
Out of the initial 35 panellists, 26 filled-out the 2nd questionnaire.
8.5 Reverse Logistics researchers and the panel

To be aware of the profile of reverse logistics researchers is very relevant
because the researchers are in fact the resources that have to carry out the
future research in the field. Concordantly, any suggestions coming out of the
Delphi study have to be within their reach. In the reverse logistics field, we
roughly recognize three standing kinds of researchers as follows:
• the junior
• the senior
• the accidental
Table 8.3 describes the main strengths and the foremost weakness of these
three research-types. Please note that this is rather a caricature-approach
than a rigorous characterization.
The panel was constituted by 35 invited academics, 16 associated with
European universities and 19 with North American universities. From the two
continents, both senior and more junior researchers participated in the Delphi
study. These researchers are experienced in teaching and they publish in
internationally renowned journals like Management Science, Harvard Business
Review, California Management Review, Operations Research and so forth.
Most of the researchers have a strong background in operations research and
operations management. Among the panel there were also other researchers of
which much of their work is focused on supply chain management, industrial
ecology, systems dynamics and environmental management, among others. In
sum, the panel incorporated the three types researchers (Table 8.3).
Table 8.3: Reverse Logistics: describing three kinds of researchers.

The Junior The Senior The Accidental
· Sound knowledge · Sound knowledge · Sound knowledge

on RL issues; on RL issues; on related fields;
Strengths · Solid modelling · Broad business

skills; perspective of RL;
· Highly motivated; · Solid background · Innovative ways

on modelling; of looking to RL as
a research field;
Weakness · Too focused on · Highly tied-up · Shortage of back-

their own topic; with various com- ground knowledge
promises; on RL issues;
8.6 Reverse Logistics: past, present and future

8.6.1 Research
Table 8.4 summarizes the panel’s perspective over the following issues:
• past research contributions,
• current limitations, and
• future opportunities
This is discussed below. During this discussion, we present more details

on the contents of the participants’ answers.
Past research contributions

The panel considered the following issues as being the most important past
research contributions in the field of reverse logistics, up to now:
- modelling,
- recognition of the field as a new body of knowledge, and
- a better understanding of reverse logistics practices

8.6. Reverse Logistics: past, present and future 255
Table 8.4: Reverse Logistics research: past, present and future.

Past Contributions Present Limitations Future Opportunities
1) Modelling; 1) Ties with practice; 1) Links with other fields;

2) Recognition of the field; 2) Employed methodology; 2) Modelling;
3) Understanding the 3) Development of theory; 3) Strengthening the field;
practice; 4) Lack of resources (e.g. 4) Empirical work;
data);
This actually reflects the way the field of reverse logistics has germinated.
The bulk of academic literature on reverse logistics is in quantitative modelling
(see Dekker et al., 2004 for a recent compilation of quantitative approaches for
reverse logistics). Network design, acquisition, inventory control, production
planning and transportation were the most mentioned modelling items by
the panel. This is not at all surprising since most quantitative work concerns
these items (see e.g. Fleischman et al, 2004; Guide and Van Wassenhove.,
2001; Van der Laan et al, 2004; Inderfurth et al., 2004; Beullens et al., 2004).
The initial quantitative approach has put forward the complexity of re-
verse logistics problems and the way they differ from the ones in traditional
logistics. This has helped to the recognition of the importance of the field.
Besides that, researchers have backed up their insights with realism by car-
rying out empirical work (see Chapter 3). Researchers working in reverse
logistics are comfortably familiar with the practice and the panel gives top
weight to that.
Other items like coordination, environmental aspects of reverse logistics,
value of information, advanced planning systems, information technology (IT)
were also mentioned by the panel. This acknowledges that :
• the field has a multidisciplinary character,
• the fields is gaining a reasonable degree of maturity.
Current Research Limitations

With respect to the current research limitations, on top of the list are the
following items:
- Ties with practice
- Methodology
- The development of theory
- The lack of resources (e.g. data).
Though researchers are aware of how reverse logistics functions in practice,

modelling has not been sufficiently linked with that understanding. The panel
recognized this by mentioning that, on the one hand, the models have many
times unrealistic assumptions, and, on the other hand, no major policy im-
plications come out of the models. These two comments are indeed related,
but the essence of it is that the panel faces any short of realism or relevance
as being a limitation to reverse logistics research. Similarly, the panel called
for more research rigor by bringing to charge unsatisfactory methodology, like
the dominance of mono-disciplinary and case-by-case approaches with over-
emphasis on details.
Besides this, reverse logistics as a new body of research suffers from the
ordinary entanglements of any new field of research. To start with, there is
lack of theoretical thickness in reverse logistics. The panel touched upon two
items: insufficient differentiation with respect to forward logistics and the lack
of an integrated theory. Besides that, there is a lack of resources, namely data
and problems in accessing the research of other research groups (low visibility
among research groups).
Both the researchers’ criticism and the more inherent consequences of a
novel research field raise many questions that have to be dealt with. Rather
than a denial of the field, this is a sign that the reverse logistics academic
community is ready to strive after more.
Research Opportunities
According to the panel the top opportunities for future research on reverse
logistics are:
- links with other fields,
- modelling,
- strengthening the field, and finally
- empirical work.
The message is a sort of

“let’s keep doing what we have been doing, but let’s improve it.”
The way to do it, was also pointed out by the panel. On the one side,
by stopping being so narrow and therefore making links with other disci-
plines and with practice. The highlighted disciplines/theories were market-
ing, design, and the extended product concept. This hints that the field is
stretching itself, accompanying developments like the extended supply chain
management concept (already noticed in the exploratory study, see Section
8.3).
Regarding modelling, the panel pointed at more holistic optimization tak-
ing into account ecological, informational, technological, and public matters.
The call for links with other disciplines, in particular through holistic mod-
els is consistent with the previously identified limitation of mono-disciplinary
procedures. On the other side, researchers have to work in the inner-self of
the field to strengthen it. Some directions to do that as referred by the panel
include clearly identifying the drivers for reverse logistics, the advantages over
traditional forms of logistics, and urging on theoretical build-up for reverse
logistics.
8.6.2 Trends and breakthroughs

The panel was also asked about trends and what would be found as a research
breakthrough. Table 8.5 synthesizes top trends with high impacts and, the
top breakthroughs with high likelihood.
By “top” and “high” we refer to items that fall in the 50% upper-tail of
the total rankings/ratings. Thus, Table 8.5 resumes all the research trends
that simultaneously respect this criterion for ranking and then for rating
(according to impact). Analogously for breakthroughs, but in this case the
rating was with respect to likelihood. The reader should keep in mind that
similar approaches were used to construct coming tables in this and following
section.
The panel identifies the following two items as the top research trends
with high impact.:
- the shift from operational to strategic issues, and
- the intensification of tactic-operational issues
At first, the two may appear contradictory but not if one bears in mind
that the shift into strategic issues is likely to be relative. After all, the majority
of researchers in the panel has a strong background on operations (see Section
8.5). IT tools is also considered to be a top trend with high impact.
The field seems moving towards integration with forward logistics, as cou-
pling reverse and forward logistics is considered to be a top trend with high
Table 8.5: Top research trends and breakthroughs with high impact and
likelihood.
Top Trends Top Breakthroughs
· More on behavior · Integrating most

of parties and co- relevant disciplines;
ordination;
· Coupling reverse · A set of simple mod-

and forward logistics; els explaining behav-
ior of parties;
High High
· Theoretical devel- · A coherent theory;
opment;
· IT tools; · Showing competi-

tiveness;
Impact · Shift from opera- ·Critical factors for Likelihood

tional to strategic is- reverse logistics;
sues;
Intensification of
tactic-operational
issues;
impact. However, the field by no means can aim at such integration with-
out proper consideration of behavior of parties and respective co-ordination,
which are as well identified as top trends with high impact. Indeed, the panel
considers a set of models explaining behavior of parties as a high research
breakthrough, and attainable. Moreover the panel expects that bringing a
holistic perspective to reverse logistics by integrating most relevant disciplines
is plausible.
The development of a coherent theory is contemplated as a top trend,
with high impact. It is interesting to note that the panel trusts that to put
together a coherent theory is only a matter of time, and it will be a valuable
breakthrough for the field. Other weighty breakthroughs and within reach are
to identify the critical factors for reverse logistics and showing competitive-
ness. The panel is therefore confident that academics will be able to point out
and announce how to manage the barriers and the stepping stones of reverse
logistics.
8.6.3 Pedagogy
Table 8.6 condenses the top challenges in teaching reverse logistics, for which
the panel considers that high priority should be given. In addition, it exhibits
top utilities that would extenuate the difficulties of this task.
Table 8.6: Top pedagogic challenges and utilities with high priority.
Top Challenges Top utilities
· Good materials · Text book + ex-

(few); ercises (basic/ ad-
vanced models)
High · Identifying clear · Overview Books High

lessons without going (e.g. Carnegie
into much details; Mellon & RevLog)
Impact · Finding interesting · Material demon- Impact

cases; strating strategic im-
pact on firms;
· Establish signif- · Case material, with

icance - Motivate possible use of multi-
(business) students media.
Since reverse logistics is a relatively recent field of research, teaching re-

verse logistics has been very much at the ad hoc level with teachers having to
start from scratch while preparing their own material for classes. Textbooks
on logistics begin to present a few pages dedicated to reverse logistics, but
exercises and in-depth cases are nearly non-existent. Finding interesting cases
and few good materials are two challenges deserving high priority. Accord-
ingly, the panel suggests that top priority is given to instruments like textbooks
with exercises and case material. Two overview books to be published this
year are specifically mentioned (see Guide and van Wassehove, 2003; Dekker
et al., 2004). Reverse logistics issues are many times taught as an appendix
of another course like logistics, supply chain, operations, production, and en-
vironmental management. Time is thus a pressure factor, contributing to
difficulties, as identifying clear lessons without going into much details and
to establish the significance/motivating the students. The panel believes that
material demonstrating strategic impact on firms would relieve some of the
pressure.
In the questionnaire, we also included a question on which case studies
or real examples researchers use when they teach about reverse logistics. In
another part of the questionnaire we also asked the panel to name 2 or 3 suc-
cessful examples of reverse logistic practices. For those that answered both
questions, around 60% mentioned examples from car industry and the reman-
ufacturing of single-use cameras (mostly associated with Kodak). Around
40% also mentioned the remanufacturing of photocopiers (associated with
Xerox and Océ). However, when compared with the cases that they use in
class, there was some discrepancy with an average 20% decrease in teach-
ing the 3 aforementioned successful reverse logistics practices. Partially the
panel explained this discrepancy by affirming that they use in class “their
own cases.” Possibly some of these cases are carried out by Master students
for their final thesis. That researchers want to use them in class is under-
standable because they will be far more familiar with their own cases, than
someone else’cases. And, since time is a pressure factor, as mentioned before,
the “successful cases” are likely to be sacrificed. But, when this happens, it
is not surprising that business students will be easily motivates. Ironically,
“motivate (business) students” is considered by the panel as a top challenge
in teaching reverse logistics and “case material” as a top utility. These issues
are among those that should have top priority, according with the panel.
How can we then give top priority to these issues? An approach is to
increase the knowledge of teachers on “successful cases.” As one researcher
stated “I would use Kodak camera and Xerox asset management cases if
they were available in detail.” To facilitate the accessibility to “successful
cases,” we suggest the organization of educator’s days/sessions in workshops
8.7. The senior, the junior and the accidental 261
or conferences on the topic and with special attention for dissemination of

successful reverse logistics practices.
8.7 The senior, the junior and the accidental

This section looks at the differences in ratings (2nd round), with respect to
the future research opportunities.
According to the panel the top opportunities for future research on reverse
logistics are (see Section 8.6):
- strengthening the field and empirical work (Realism),
- modelling (Rigor ), and
- links with other fields,
Table 8.7 exhibits the average ratings of the items, by type of researcher,
i.e. senior, junior and accidental.
Table 8.7: Realism, rigor and links: average ratings.

Realism Rigor (Modelling) Links
Senior 14.0 11.9 22.0
Junior 9.8 13.3 17.8
Accidental 7.8 8.75 20.5
One can observe that there are no perfect clusters of researcher type and
rated research opportunities. Yet, we observe the following:
• the highest average rating for realism is by the senior group
• the highest average rating for rigor is by the junior
• the highest average rating for links is the accidental
Concluding, the senior researcher is likely to bring more realism than any
other type, the junior will put more rigor into the research, and what the
accidental type has most to offer are the links with other disciplines.
8.8 European vs. North American researchers

Since the early nineties several books about Reverse Logistics have been sup-
ported by the Council for Logistics Management (Stock, 1992, Kopicky, 1993,
Stock, 1998, Rogers and Tibben-Lembke, 1999). Yet, the academic commu-
nity in Europe “gave in” sooner to reverse logistics as a new field of research
(see Thierry et al., 1995, Fleischmann et al., 1997, RevLog, 1998-). In Eu-
rope, environmental issues brought early attention to the potential benefits of
recovering products (Corbett and van Wassenhove, 1993). Recovery quotas
and packaging directives by the European Union deepened the academic and
company interest on reverse logistics.
Actually, by comparing the practice of reverse logistics in North American
and European, one observes that the prime driver in most North American
cases has been profit while in Europe legislation has also been important (see
Chapter 3). Thus, the evolution of reverse logistics has been different in the
two continents (for instance U.S. environmental practices still rely heavily on
policies from the eighties, see Murphy and Poist, 2003).
It is therefore not eccentric to conceive that reverse logistics research is
also at a different stage in either of them. If so, the two communities may
account dissimilarly to the future of reverse logistics research. We elaborate
more on this by distinguishing between the answers of researchers associated
with European (E) universities versus North American (NA) ones.
The distance between the relative ratings of recognition of the field, de-
scription of practice, and modelling is much smaller for the North American
panel than for the Europeans (see Figure 8.2). This is not surprising. The
North American research community is beginning to be convinced about the
importance of the field, especially since the publication of the book of Rogers
and Tibben-Lembke (1999). The book is an extensive description of the Amer-
ican business practices of reverse logistics. As stated there, according to the
Logistics Council, bad management of reverse logistics was then costing bil-
lions of dollars to American companies. In Europe, the recognition of the
field of reverse logistics is for long a settled matter. E.g., already since years
a specialized research group on reverse logistics has been active in Europe (see
RevLog, 1998-). The core of European research has been dedicated to intense
quantitative modelling. Although much case study research has taken place
in Europe (see De Brito et al., 2003), it is natural that European researchers
remain weighting to a high degree their bulk of research, i.e. modelling.
8.8. European vs. North American researchers 263
Figure 8.2: Past research contributions: relative ratings; 100% corresponds

to the highest sum of ratings
For the North American participants the current top limitation of reverse
logistics research is the lack of ties with practice followed by methodology
and the other way around for the European panel. The lack of theory is
more emphasized by the Europeans, who have been for long investigating
reverse logistics issues. Therefore, they presumably have waited longer for
the theoretical development of the field giving more weight to it (see Figure
8.3).
Figure 8.3: Current research limitations: relative ratings; 100% corresponds

to the highest sum of ratings.
The central opportunities according to the North American panel are the links
with other fields of research (actually, also for Europeans, see Figure 8.4). By
doing that, North American researchers increase the chance of drawing atten-
tion to reverse logistics, granting further recognition (a matter not entirely
resolved in North America). The favored discipline to establish a link was
marketing, followed by the extended product concept. This goes along with
important values in North America like market, and service. The questions
here associated are: What about the marketing of recovery products? Is the
consumer receptive to recovered products? How recovered products can com-
pete with brand new ones? Can we serve customers better if we offer them
recovered products?; and so on.
The links with other fields is also looked upon by the European panel as
the upmost topic for future research, where design, extended product concept
and strategy were often mentioned topics. For the European panel modelling
is still a prominent opportunity (very much attached to their core compe-
tencies). Within modelling, they mentioned IT tools to a high degree. This
indicates that Europeans believe to be prepared to support reverse logistics
with advanced tools. Such confidence is certainly lifted up by the prolonged
contacts that European researchers have with the practice of reverse logistics.
Empirical work is considered to be more of a research opportunity for the
North American panel, than for the European (relatively to other items).
In the past, North American research on reverse logistics has been very
much theoretically driven. For instance, as put by Drejer et al. (2002), U.S.
research has sacrificed realism to be more focused on precision and general-
izability. However, North American researchers are willing to deviate from
this traditional research paradigm, as they reasonably rate field research as a
research opportunity.
Table 8.8: Top research trends: North America vs. Europe.

North America Europe
· Theoretical development;
· Theoretical development; High · More on behavior of parties ;
· Shift from operational to strate- and co-ordination
gic
issues; Impact · Shift from operational to strate-
gic issues;
·Multidisciplinary approaches; · IT tools;
· Coupling reverse and
· Intensification of tactic- forward logistics
operational issues;
· Intensification of tactic-
operational issues;
8.8. European vs. North American researchers 265
Table 8.9: Top research breakthroughs with high likelihood: North America
vs. Europe.
North America Europe
· IT support for RL;

· A set of simple models explaining High · Integrating OR models
behavior of parties;
with RL’s strategy
· A coherent theory; Likelihood · A set of simple models explaining
behavior of parties;
· A clear distinction between;
· Showing competitiveness; forward and reverse logistics
· Integrating most relevant
· Critical factors for reverse logis- disciplines
tics;
· A coherent theory;
· Showing competitiveness;
· Critical factors for reverse logis-
tics;
Figure 8.4: Future research opportunities: relative ratings; 100% corresponds

to the highest sum of ratings.
Table 8.8 presents the top trends with high impacts and Table 8.9 presents
top breakthroughs with high likelihood. The items that are mentioned alone
by the European or by the North American researchers are underlined. As
mentioned before (Section 5), more on behavior of parties and co-ordination
is a natural consequence of coupling reverse with forward logistics. The North
America participants also agree that these two items are a top research trend
but they attribute it a lower relative impact. The North American panel
accredits to multidisciplinary approaches more impact than to e.g. theoretical
development of the field. IT Tools is thought to be a top trend with high
impact for Europeans but not for North American researchers.
The European panel is more condescending with respect to the plausibility
of top research breakthroughs. The long experience of the European research
community is likely to contribute to this.
8.9 A glimpse on the rest of the world

In this chapter we proposed a reflection on the future of reverse logistics as an
academic field (see section 8.1). To do so, we carried out a Delphi study with
experts in research on the field of reverse logistics, and related fields. These
experts happen to be affiliated with either with North American or European
universities, as most of the work in the area has been done there. We would
like to remark that we do not imply that the future of reverse logistics depends
exclusively on the developments in these two continents. Throughout these
thesis, some research contributions coming from other geographic areas have
been as well referred to (Yuan and Cheung, 1998, see Chapter 4); (Díaz and
Fu (1997); Chang and Wei, 2000; Gupta and Chakraborty, 1984, see Chapter
3).
Some of the aforementioned references reported on real life examples. Díaz
and Fu (1997) applied an inventory model to a case of spare parts management
at the subway system Caracas, Venezuela. Chang and Wei (2000) discussed
the recycling network for household waste of the city of Kaohsiung in Taiwan.
Gupta and Chakraborty (1984) reported on the rework taking place in a glass
plant in India.
Though Europe is leading in pro-environmental legislation, the environ-
mental awareness (and ergo behavior) by governmental institutions is becom-
ing a wider phenomena. In Asia, Japan and Taiwan are leading regions on
this matter (see De Koster et al, 2003), though other regions in Asia have a
long history on recycling.
As remarked by Breukering (2001) India has been using waste paper as raw
material in the glass industry as from the 1950’s and China (including Hong
Kong) imports more than 90% of the total plastic waste traded internationally,
which is recycled. The author discussed recycling from the perspective of
international trade and how governmental entities (re)act on it. For instance,
during the nineties, India increased the import tax on paper waste putting
the paper recycling industry in trouble. Similarly and around the same time,
China prohibited the import of waste plastics. Both actions were later revised
as they were not sustainable for the recycling industry. The conclusion by
the author is that there is a lack of understanding on international trade
of secondary materials, leading to a overemphasis of negative aspects rather
than looking to the positive ones. The same applies to other regions in the
world.
Wamiti (2001) defends the great potential that reverse logistics has in
Kenya. The author goes over several examples: 1) the collection of barley
spent grains in the beer industry for further use in farming ; 2) the imports
of second-hand shoes and clothing; 3) replacement of engines in used cars; 4)
rethreading of tyres. Only the first example is considered really successful.
The other cases suffer from governmental policy to protect the (inefficient)
local industry.

We observed that the view of the panel on past research contributions reflected
the way the field has been growing up. The panel pointed out current limi-
tations of reverse logistics research. Some of those are inherited from reverse
logistics being a young field of research. In addition, the panel exposed that
more research rigor, more realism and more relevance need to be brought
to the field; otherwise it will remain very restricted. By the opportunities
pointed out, the field seems to be stretching itself and becoming rather linked
with other fields of research like marketing. The panel showed confidence in
the plausibility of strengthening the field by e.g. identifying critical factors
for reverse logistics and putting together a coherent theory (see Section 8.6.1.
With respect to teaching reverse logistics, the main challenges are, ac-
cording to the panel: i) finding interesting cases, as well as ii) good materials.
Accordingly, the panel suggests that top priority is given to instruments like
textbooks with exercises and case material. Time is a pressure factor as reverse
logistics issues are many times taught as an appendix of another course. The
panel believes that material demonstrating strategic impact on firms would
relieve some of the pressure of identifying clear lessons without going into
much details. We also noticed that there was some discrepancy between the
successful reverse logistics cases and the examples discussed in class. A plau-
sible reason is that teachers lack of in-depth familiarity with those successful
practices.
Aligned with this, we suggest the organization of educators’ days/sessions
in workshops or conferences on the topic with special attention for the dis-
semination of successful reverse logistics practices. These sessions should aim
to support reverse logistics teaching. Therefore, it is expected that the format
will be somewhat different from the traditional workshop/conference presen-
tations.
In addition we indicated how close the views of the European and North
American participants are. The North American research community is start-
ing to accept the importance of the field, giving appreciable importance to
recognition of the field, description of practice, and modelling as past contri-
butions. The Europeans, who have been for long investigating reverse logistics
issues, have waited longer for the theoretical development of the field giving
more weight to it as a current limitation. As the top research opportunity,
both North American and European agree that it is to link reverse logistics
with other fields. The favored discipline to establish a link for the NA re-
searchers is marketing, followed by the extended product concept, reflecting
how important values like market, and service are in North America. Euro-
peans give also substantial weight to future modelling as an opportunity, and
within it they often mentioned IT. This indicates that Europeans believe to
be prepared to support reverse logistics with advanced tools. Such confidence
is certainly a result of their a long history of contacts with practice.
In the past, North American research on reverse logistics has been very
much theoretically driven. However, North American researchers are willing
to deviate from this traditional research paradigm, as they reasonably rate
field research as a research opportunity. Finally, the European panel was more
condescending with respect to the plausibility of top research breakthroughs.
Again, the long experience of the European research community with reverse
logistics is likely to contribute to this.
We have assessed which future academicians expect for reverse logistics
as a research field. In essence the panel claimed for
• more realism,
• more rigor, and
• more links with other disciplines.
These are undoubtedly ambitious aspirations. The question that remains
is: how can we work so this becomes reality? First of all, the fact that there is
willingness and belief is very positive. This especially because the panel had
the representation of the three types of researchers: the senior, the more junior
and the accidental (see Section 8.5). We would like to advance that as a matter
of fact only a coordinated exploitation of the three research types can bring
through the desired future for reverse logistics research. Putting it plane,
the senior researcher will bring realism, the junior rigor, and the accidental
researcher will further facilitate the links with other disciplines. Especially if
this takes place in a inter-continental fashion, we foresee relevance (see Figure
8.5).
Figure 8.5: Achieving relevance for Reverse Logistics.
Besides this, the development of reverse logistics research also depends

on how it develops in practice, which will be the stressed aspects and where.
As we commented (see Section 8.9), environmental awareness is not an Euro-
pean exclusive phenomenon, and reverse logistics has growing opportunities
throughout the world.
Chapter 9
Concluding remarks
Now this is not the end.

It is not even the beginning of the end.
But it is, perhaps, the end of the beginning.
Sir Winston Churchill
The research presented in this thesis was designed to provide a better under-
standing of reverse logistics. We brought insights to reverse logistics decision-
making and we laid down a theoretical ground for the development of reverse
logistics as a research field.
This thesis aimed at
• structuring reverse logistics as a research field,
• a better understanding of reverse logistics practices,
• structuring and supporting reverse logistics decision-making, and
• conjecturing about the future development of reverse logistics as a re-
search field.
In the remainder of this chapter, we summarize the main findings. Besides
that, we go over the limitations and put forward opportunities for further
research.

Here we recapitulate the findings of this thesis. To do so, we consider the
chapters one by one.
271
272 Chapter 9. Concluding remarks
Chapter 2: To structure the field and to provide an overall under-

standing of reverse logistics.
This chapter provided an overall understanding of reverse logistics by struc-
turing the field. In fact, we put together a framework for reverse logistics, i.e.
a basic conceptional structure for reverse logistics, as follows: we identified the
elementary ingredients (dimensions) of reverse logistics, we structured them,
and we described their relations to each other.
To build the framework we employed three sources of input: selective liter-
ature, a review of case studies (Chapter 3 of this thesis), and the knowledge on
reverse logistics that was accumulated throughout the whole PhD trajectory.
To build up the framework for reverse logistics five key dimensions were
considered:
• Why products go back in the supply chain i.e. Why- returning;
• Why companies become active with reverse logistics, i.e. Why-receiving;
• What is being returned? The analysis comprised, on the one hand

crucial product characteristics for reverse logistics, and on the other a
classification of products types;
• How are returns processed? In this respect we described the overall

processes in a reverse logistics process, and we gave special attention to
recovery options;
• Who is executing reverse logistic activities? Here, the enquiry was on

the actors and their roles in implementing reverse logistics;
More specifically, we have differentiated three driving forces for reverse

logistics:
• economics,
• legislation, and
• corporate citizenship.
We have also stressed that the economic gains can either be direct or
indirect. Furthermore, the three drivers are also interlinked and boundaries
are sometimes blurred, as reverse logistics is often carried out for a mix of
motives.
The return reasons were organized according to three supply chain stages:
• manufacturing returns, which embrace raw material surplus, quality-

control returns, and production leftovers or by-products.
• distribution returns, which include product recalls, returns coming back

due to commercial agreements (B2B commercial returns), internal stock
adjustments and functional returns.
• customer returns, which comprise reimbursement guarantees, warranty

returns, service returns, end-of-use and end-of-life returns.
The recovery options can be organized in great concordance with Thierry

et al, 1995, that is product level (repair); module level (refurbishing); com-
ponent level (remanufacturing); selective part level (retrieval); material level
(recycling); energy level(incineration).
On product characteristics, we qualified three main product features af-
fecting reverse logistics:
• the composition of the product,
• the deterioration process, and
• the use-pattern.
On product types, the typology presented was inspired by the work of

Fleischmann et al. (1997) and the UN product’s classification, and the fol-
lowing types were differentiated:
• civil objects,
• consumer goods,
• industrial goods,
• ores, oils and chemicals,
• packaging and distribution items,
• spare-parts, and
• other materials (like pulp, glass and scraps).
Regarding actors in reverse logistic systems, we divided them in three

groups:
• typical forward supply chain actors,

• specialized reverse chain players,
• governmental institutions, and
• opportunistic players.
With respect to their roles, we detected 1) actors that are actually re-
sponsible for operations in the reverse logistics chain; 2) others, that set up
or combine operations, they have a role of an organizer ; 3) many parties are
busy carrying out the processes, like the collectors, so they play an executer
role; 4) there is the sender/giver that facilitates the product for recovery and
the future client that will acquire recovered products, without which recovery
would not make practical sense, so they have an accommodator role.
We stressed that the combination and characterization of the five dimen-
sions determine to a large extent the kind of issues that arise in implementing,
monitoring and managing such a system. This was illustrated with a collec-
tion of typical cases.
Chapter 3: To provide a diversity of reverse logistics activities (and

to help constructing the framework).
One cannot provide proper insights into reverse logistics without being famil-
iar with how firms are dealing with it in practice, which are the trade-offs
during decision-making, how decisions are being supported, and so on.
We have provided a content analysis of more than 60 cases reported in
the literature dealing with reverse logistics aspects. The study was organized
according to the following topics 1) Network Structures, 2) Relationships, 3)
Inventory Management, 4) Planning and Control, and 5) Information Tech-
nology.
We remarked that Reverse Logistics Network Structures are well organized
around the typologies on the how and the who. The cases found included the
following:
• Networks for re-distribution/re-sale,
• Networks for remanufacturing,
• Networks for Recycling - Public Networks, and

The cases showed that there are quite different incentive tools to enforce
certain desired behavior: 1) deposit fees; 2) buy back options; 3) trade-in’; 4)
acquisition price; 5) timely and clear information; 6) power; 7) appealing for
environmental responsibility; 8) appealing for social responsibility.
Some incentives are also part of sales contracts, while others require the
customer to buy another product in exchange. There are also tools that are
not directly coupled to a selling activity like a gift to a non-profit organization.
Actually, it seems that only deposit fees are specific for product recovery. The
other mentioned tools are also used to attract customers in general.
Concerning Inventory Management, the why-returning typology seemed a
natural way of grouping and discriminating the reverse logistics issues raising
from the cases. We found cases for
• End-of-use, and
• End-of-life.
We found case studies on planning and control of product recovery in

the following categories: 1) the separate collection of (parts of) products for
recovery 2) the separate processing of (parts of) products for reuse or disposal
3) the combined planning and control of collection of products for recovery
and distribution of new products 4) the combined planning and control of
processing products for recovery and production of new products
For each of the above decision-making areas, we concluded that some of the
dimensions of the framework (see Chapter 2) are predominantly important,
as follows
• Networks ⇔ Recovery option (How ) and Actor (Who)
• Relationships ⇔ Actor (Who)

type (What)
• Production Planning ⇔ Recovery option (How ) and Actor (Who)

Besides the analysis per decision-making area, we also presented the cases
organized by return reason (why-returning) vs. recovery option (how ), and
the cases organized by driver vs. recovery option. This analysis revealed some
relations and issues. For instance, end-of-life returns appear to be highly cor-
related with recycling. If products or their modules would be designed for
remanufacturing, however, remanufacturing would likely be more attractive,
even at the end of life. The main driver for recycling is legislation. Actually,
current legislation does not seem to be able to stimulate other forms of recov-
ery. However, if legislation would also pro-actively focus on product design
issues, rather than dealing with recovery in a reactive way, other recovery
options than recycling would likely be more stimulated.
In addition, we reported on cases describing the use of IT to support
reverse logistics. We found cases reporting evidence for IT support in all
the stages of the life-path of a product: product development, the supply
chain loop, and at the customer. For instance the data logger technology (see
EUREKA, 2003) is very promising regarding the need of a quick access of
the state of the product. In this way uncertainty on quality can be reduced
and savings can be made on activities, as e.g. disassembling. The cases
also show that though the technology to process and to transmit information
with promising benefits for reverse logistics seems to be available, the lack
of appropriate data is in still a bottleneck in the implementation of decision
support systems.
We observed a lack of cases on reimbursement returns. However, with the
growth of the catalogue industry (also in the Internet), posing new challenges.
Ergo, field research in this area is likely to growth as well.
The majority of the case studies dealt with one aspect of a real reverse
logistics situation but they did not give the overall business environment,
which made insights rather one-dimensional. Thus, we pointed out the need
for conducting more integral case study research, by mapping the business
context together with more broad information on critical factors, trade-offs
and implications.
Besides that, the lack of theory for reverse logistics (see Dowlatshahi,
2000) or even for supply chain management (Croom et al., 2000) adds to
companies’ inability of knowing what matters in reverse logistics. Therefore
the development of theory should be on top of the research agenda to support
reverse logistics decisions.
Chapter 4: To put forward the state-of-the-art of quantitative mod-

els for return handling and to identify opportunities for future re-
search
Though available quantitative models can be adapted to support warehousing
return handling, this has been largely ignored up to now. For a number of
research areas, we highlighted how forward models can be extended or new
models can be launched to include return handling in areas such as:
• facility layout & design,

• outsourcing,
• integrating operations (return policy and reusable packaging),
• inventory management,
• internal (return) transport,
• information systems,
• inventory, storage and order picking control, and
• vehicle planning and scheduling.
We brought to the attention several research opportunities, which can

be turned in a research agenda, as follows: What is the impact that return
flows have on 1) the warehouse layout; 2) material handling systems (e.g.,
whether dedicated return handling systems do pay off); 3) vehicle planning
and scheduling (mind sequencing restrictions); 4) storage and picking proce-
dures (e. g. route combination of returns’ storage and pick ups). Further-
more we suggested the use of the Mathematical Programming problem with
Equilibrium Constraints (MPEC) formulation to e.g. help retailers deciding
upon the return policy. A new direction to inventory modelling was also put
forward by categorizing customers in two classes, i.e. less- and more- prof-
itable depending on whether the customer (systematically) returns, or not,
merchandize.
Chapter 5: To identify the main factors contributing whether, or

not, to combine forward and reverse flows during return handling.
The identified factors during transport to the warehouse, the receipt at the
warehouse, and, storage are respectively as follows: to have, or not, stores;
the return volume; and, the future market for returns.
In more detail, we conjectured that:
• For retailer organizations that supply (a sufficient number of) stores it

is more efficient to collect the returned products, carriers and waste to
the distribution center with the same truck that delivers the products.
• For retailers that handle a high volume of returns, it is more efficient to

unload and sort returns in a separate area of the distribution center.
• If the market for returns is different from the original market then prod-
uct to store returns in a different area from those of purchased products
is preferable.
• If the market for returns is the same then storage of product returns
is likely to be combined with purchased products. Exceptions will be
found in the case of high-desired control over returned products
Before we started this specific research project we suspected that the com-
bination of volume and product diversity would have an impact on the storage
decision. From the analysis though it came out that the complicating factor
was the decision of controlling returns. Yet, both volume and product diver-
sity can be a point of further research since they might not be a complicating
factor alone but be so when combined with other factors. Furthermore, to
have a relevant degree of return monitoring and control, a separate module
for returns in the warehouse management system is desirable as it is already
the case for mail order companies.
Chapter 6: To inquire the validity of common assumptions in the

inventory management literature with product returns.
We established a framework for the statistical testing of common assumptions
in the inventory management literature with product returns, i.e.
• the demand process is a Homogeneous Compound Poisson Process (HCPP).
• the return process is a Homogeneous Compound Poisson Process (HCPP).
• the time elapsing between return and demand occurrences follows a

negative exponential distribution;
With the developed methodology it was possible to carry out the desired
testing including small (below 20) sample sizes. Only for extremely small
sample sizes, the methodology does not work. It is unambiguous that in
the last case, either the proposed or any other statistical framework cannot
guarantee rigor. We employed the framework to three sets of real data, as

follows
Case 1: data from an in-company warehouse with internal returns in
Switzerland (CERN).
Case 2: data from a mail-order company with commercial returns in the
Netherlands (referred to as MOC).
Case 3: data from an in-company warehouse in the Netherlands (referred
to as RF).
We found products for which the common assumptions on the theoretical
models fit reasonably, but also products for which they do not. Regarding
the (non-)validity of the common assumptions, we empirically observed that:
• Demand may not behave as an HCPP due to seasonality or over long

periods of time;
• In non-stationary environments, the return process may behave as an
HCPP;
• Time to return may not behave as an exponential distribution due to a
long tail or on account of the short-lag returns;
Concerning inventory management and control, we suggested that compa-

nies would separate short-lag returns from long-lag returns. Then, to proceed
to netting of the short-lag returns and to monitor the long-lag returns (return
distribution, the return rates and register return reasons). The most gains of
feeding inventory policies with detailed information on product returns are
likely to be in medium well-behaved environments.
Chapter 7: To assess the impact of (mis)information on inventory

management
In the literature there are not many references related to research on the im-
pact of information on inventory management with product returns. The ones
that exist generally assume known return probabilities or consider specific
cases where the most-informed method leads to the best forecast. However,
in environments where data is scarce or unreliable, or in environments that
are volatile, information may be misleading. This motivated us to assess the
impact that misinformation has on inventory management.
In order to bring the first insights on this matter, we focused on a static
environment where consistently the forecasted return rate or the return dis-
tribution does not correspond to the true behavior. We considered a single
product, single echelon, periodic review inventory system. We followed the

approach of Kelle and Silver (1989). Accordingly, we employ a base-stock
policy and four different methods are used to forecast the net demand dur-
ing the lead time. The four methods use different levels of information, with
respect to returns:
• Method A - Average behavior. This method estimates returns during the

lead time through the overall return probability. Demand and returns
during the lead time are simply netted. No historical information is
used with respect to demands and returns, so in a static environment
the resulting base stock level is constant in time.
• Method B - Return distribution. This method makes use of the return

distribution (return probabilities, per period) to determine the number
and moment of the returns during the order lead time.
• Method C - Return distribution & return information per period. This

method updates the number of items that have returned, which are
observed, and takes this into account to compute future returns.
• Method D - Return distribution & tracked individual returns. To employ

this method, one has to be able to trace back the returns, i.e. to know
from which order period each specific return is coming from.
We identified situations in which the most informed method does not nec-
essarily lead to the best performance. This is the case when one consistently
overestimates or underestimates the return rate per period.
Both for misinformation on the return rate and on the return distribution,
the differences between the methods become smaller with decrease of the
return probability and an increase of the lead time.
The most robust method given misestimation of the return rate is Method
B. Method B systematically outperforms Methods C and D if the return rate
is misestimated by merely 10%,or more. The cost differences are particularly
high if return rates are overestimated. With respect to misestimating the
conditional time to return distribution, Method C and again Method B are
much more robust than Method D.
The findings strongly suggested that Method B has a sufficient level of
sophistication both in the case of perfect estimation and imperfect estimation.
In the latter case Method B is far more robust than the most informed method,
Method D.
Chapter 8: To assess which future do academicians expect for re-

verse logistics as a research field
The field of Reverse Logistics has been expanding fast and it remains a field
in formation. To reflect on the future of reverse logistics as an academic
field, we carried out a Delphi Study with an international panel of academics
working on reverse logistics issues. The study consisted of two rounds of
anonymous questionnaires with an intermediary feedback phase. In short,
every member received information on the answers to the first questionnaire,
which they could rank and rate. The overall analysis allows us to evaluate
reverse logistics issues with respect to research and teaching.
We observed that the view of the panel on past research contributions
reflected the way the field has been growing up. The panel pointed out current
limitations of reverse logistics research. Some of those are inherited from
reverse logistics being a young field of research. In addition, the panel exposed
that more research rigor, more realism and more relevance need to be brought
to the field; otherwise it will remain very restricted. By the opportunities
pointed out, the field seems to be stretching itself and becoming rather linked
with other fields of research like marketing. The panel showed confidence in
the plausibility of strengthening the field by e.g. identifying critical factors
for reverse logistics and putting together a coherent theory.
With respect to teaching reverse logistics, the main challenges are, accord-
ing to the panel, finding interesting cases as well as good materials. Accord-
ingly, the panel suggests that top priority is given to instruments like textbooks
with exercises and case material. Time is a pressure factor as reverse logistics
issues are many times taught as an appendix of another course. The panel
believes that material demonstrating strategic impact on firms would relieve
some of the pressure of identifying clear lessons without going into much de-
tails. Besides this, we also noticed that there was some incoherency between
the successful reverse logistics cases and the examples that teachers discuss
in class. Thus, to facilitate the accessibility to successful cases,” we suggested
the organization of educators days/sessions in workshops or conferences.
In addition we compared the views of the European and North American
participants are. The North American research community is starting to ac-
cept the importance of the field, giving appreciable importance to recognition
of the field, description of practice, and modelling as past contributions. The
Europeans, who have been for long investigating reverse logistics issues, have
waited longer for the theoretical development of the field giving more weight
to it as a current limitation. As the top research opportunity, both North
American and European agree that it is to link reverse logistics with other
fields. The favored discipline to establish a link for the NA researchers is mar-
keting, followed by the extended product concept, reflecting how important
values like market, and service are in North America. Europeans give also
substantial weight to future modelling as an opportunity, and within it they
often mentioned IT. This indicates that Europeans believe to be prepared to
support reverse logistics with advanced tools. Such confidence is certainly
a result of their long history of contacts with practice. On this, and in the
past, North American research on reverse logistics has been very much theo-
retically driven. However, North American researchers are willing to deviate
from this traditional research paradigm, as they reasonably rate field research
as a research opportunity. Finally, the European panel was more condescend-
ing with respect to the plausibility of top research breakthroughs. Again, the
long experience of the European research community with reverse logistics is
likely to contribute to this.
We assessed which future academicians expect for reverse logistics as a
research field. In essence the panel claimed for more rigor, more realism and
more relevance of the field at the same time that links should be made with
other disciplines. These are undoubtedly ambitious expert aspirations. We
also stressed how we can work to make this happen. First of all, the fact that
there is willingness and belief is very positive. This because the panel had the
representation of the three types of researchers: the senior, the more junior
and the accidental. We advanced that as a matter of fact only a coordinated
exploitation of the three research types can bring through the desired future
for reverse logistics research. Putting it plainly, the senior researcher will
bring realism, the junior rigor, the accidental researcher will facilitate the
links with other disciplines: all together produces relevance.
9.2 Raising questions and further research
This thesis was carried out throughout three phases: 1) design, 2) data col-
lection and analysis, and 3) reflection. During the design phase a careful
plan was put together with input both from experts on reverse logistics and
from the literature of the field. The next phase, data collection and analysis
was organized in several projects, tuned with the pre-set objectives. In each
of these projects, there was collaboration with one or more researchers from
the European Working Group on Reverse Logistics, RevLog . This provided
triangulation with respect to the role of the researcher. Triangulation is in
essence a plural approach to research, aiming at diminishing research biases.
This thesis benefited not only from investigator triangulation, but also from
9.2. Raising questions and further research 283
data triangulation (different sources of data as both professional and academic

renowned journals) and combination of different methods (e.g. analytical and
simulation). During the reflection phase, we followed the standard informal
procedure of the Erasmus University Rotterdam: each chapter in turn was
discussed with the promoters and revised accordingly. Special attention was
paid to coherence, significance, and lacunas, improving the internal, the con-
struct, and the internal validity of this thesis.
However, a scientific research project and its findings is also meant to raise
questions, otherwise it would not meaningfully contribute to the advance of
scientific knowledge. Every chapter of this thesis ends with research limita-
tions and/or research opportunities. Next, we propose overall directions for
future research on the field of reverse logistics, based on two main questions
that arise after a reflection on this thesis.
How can scientific reverse logistics research support decision-

making in practice?
In our view, the course to tackle this is twofold: 1) on the one hand,
to learn more about reverse logistics practices, 2) on the other hand, to
adapt/develop methods to aid specific critical decisions. Obviously, the first
would be an input to the second. To learn more about practice, one can
employ qualitative research methodologies, such as comparative case study
research. We suggest the use of theoretical sampling, i.e. having in mind the
foreseeable degree of understanding that the case study in question brings to
the field. Regarding decision-making methods, a great help can come from
quantitative models. Yet, as the Delphi study showed, more realism is needed
(see Chapter 8). There is a need to break with the traditional assumptions,
for instance in the literature on inventory modelling with returns. A necessary
direction is to develop models suitable for the non-stationary (real) situations,
as it happens in the presence of seasonal products (see Chapter 4).
How can scientific reverse logistics research contribute to estab-

lish a reverse logistics theory?
It seems inevitable that reverse logistics will grow as a research field and
that its scope will widen. In Chapter 1 we described how the definition of
reverse logistics changed over time. We are confident that a reverse logistics
theory can be established but to do that both inner and outer approaches
are needed. By an inner-approach we mean a focus on the essence of reverse
logistics, by paying special attention to consistency and meaningfulness with
respect to reverse logistics management. The framework for Reverse Logistics
presented in Chapter 2 can be the departure platform to put forward propo-

sitions, which upon testing can lead to the identification of critical factors for
reverse logistics and therefore to a theory. By an outer-approach we imply the
following: to relate the field with other fields of research, like the Delphi study
suggested (see Chapter 8). Both similarities and contrasting points should be
brought forward, so reverse logistics can be a field with its own individuality.
We would like to recall that this research project benefited from co-
operation with many European researchers in this topic (especially within
the RevLog group). This in itself is an added-value, and we hope that other
researchers can benefit from similar working groups in the future. On the
other hand, this gives the research a remarkable European flavor, which can
be seen by some sort of limitation. A way to reduce geographic bias is to
establish bridges with researchers in other continents. Between Europe and
North America, there is already a semi-formal and structured collaboration
on the field, as the recent book of Guide and van Wassenhove (2003) attests.
Besides that, in the conferences on the field, contributions from colleagues
from other continents are not unusual. Reverse logistics is a worldwide phe-
nomenon (see Chapter 8) and we can be optimistic with respect to a wider
collaboration especially embracing areas that systematically lag behind. For
instance the European Association of OR Societies (EURO, 2003) is helping
African researchers to revive scientific research in Africa. A living example is
the set up of AORN, 2003, the African Operational Research Network.
As final words, we come back to the initial question of this thesis:
Managing reverse logistics

or
It goes without saying that reverse logistics is also logistics, so a priori

one can also transfer lessons from traditional logistics management to the
management of reverse logistics. On the other hand, traditional logistics
management mainly thinks “forward.” Logistics managers are ultra-focused
on dispatching the goods, on moving them forward to the client. They are
not to blame, after all, this has been the formula to keep businesses going,
and many companies have reward schemes based on it. But this is not going
to be enough in the future, where companies are eager to add value to their
products, by providing more service and by becoming “corporate citizens.”
9.2. Raising questions and further research 285
As reverse logistics stretches out worldwide, in all the layers of the supply
chain, some actors in the chain are forced to take products back, while others
will do so pro-actively (attracted by the value in used products). One way or
the other, reverse logistics is a key competence in modern supply chains.
Thus, in managing reverse logistics, one also has to be able of reversing
traditional logistics management:
Logistics cannot go forward without thinking reverse!

Bibliography
[1] ACRR (2003). ACRR, The Association of Cities and Regions for Recy-
cling. http://www.acrr.org.
[2] Advanced Recovery (2003). http://www.advancedrecovery.com.
[3] Akkermans, H., Bogerd, P., Yucesan, E., and van Wassenhove,
L. N. (2003). The impact of erp on supply chain management: ex-
ploratory findings from a european delphi study. European Journal of
Operational Research, 146:284–301.
[4] Amazon.com (2003). www.amazon.com.
[5] Anderson, S., Browne, M., and Allen, J. (1999). Logistics impli-
cations of the uk packaging waste regulations. International Journal of
Logistics: Research and Applications, 2(2):129–145.
[6] Andriesse, F. G. (1999). Successful implementation of reverse logistics

at Philip Morris. In Van Goor, A. R., Flapper, S. D. P., and Clement,
C., editors, Handboek Reverse Logistics. Kluwer, B.V., Deventer, The
Netherlands. (In Dutch).
[7] Anupindi, R. and Bassok, Y. (1999). Supply contracts with quantity

commitments and stochastic demand. In Tayur, S., Ganeshan, R., and
Magazine, M., editors, Reverse Logistics. Kluwer Academic Publishers,
The Netherlands. pp. 197-232.
[8] AORN (2003). African operational research network (aorn).

http://www.euro-online.org/africanOR/.
[9] Arge Altauto (2003). http://www.arge-altauto.de/.
[10] Austrian Federal Chamber of Labour (1998a). Employment

in the environmental sector. Report, Austrian Federal Chamber of
Labour, Austria.
287
288 Bibliography
[11] Austrian Federal Chamber of Labour (1998b). Working condi-

tions in the waste management sector. Report, Austrian Federal Cham-
ber of Labour, Austria.
[12] Auto Recycling Nederland (2003). http://www.autorecycling.nl/.
[13] Bain, L. J. and Engelhardt, M. (1987). Introduction to Probability

and Statistics. Duxbury Press, Boston.
[14] Bakkers, F. T. P. G. and van Amstel Jr., J. J. A. P. (2000).

Ortes Lecluyse return- and recycling system for PVC slats. In Van Goor,
A., Flapper, S., and Clement, C., editors, Handboek Reverse Logistics.
Kluwer, B.V., Deventer, The Netherlands. (In Dutch).
[15] Barros, A. I., Dekker, R., and Scholten, V. (1998). A two-

level network for recycling sand: A case study,. European Journal of
Operational Research, 110(2):199–215.
[16] Bartel, T. (1995). Recycling program for toner cartridges and optical
photoconductors. In Proceedings IEEE Symposium on Electronics and
the Environment, pages 225–228, Orlando, Florida.
[17] Bartels, J. J. C. (1998). Collection and processing of batteries in The

Netherlands - the Stibat execution scheme. In Van Goor, A., Flapper,
S., and Clement, C., editors, Handboek Reverse Logistics. Kluwer B.V.,
Deventer, The Netherlands. (In Dutch).
[18] Bartholdi, J. J., Bunimovich, L. A., and Eisenstein, D. D.

(1999). Dynamics of two- and three-worker ‘bucket brigade’ production
lines. Operations Research, 47(3):488–491.
[19] Bartholdi, J. J. and Eisenstein, D. D. (1996). A production lione

that balances itself. Operations Research, 44(1):21–34.
[20] Bartholdi, J. J. and Gue, K. R. (2000). Reducing labor costs in

an ltl crossdockiing terminal. Operations Research, 48(6):823–832.
[21] Bayindir, Z. P., Erkip, N., and Güllü, R. (2003). A model to eval-
uate inventory costs in a remanufacturing environment. International
Journal of Production Economics, 81–82:597–607.
[22] Bentley, D., Rothstein, S. R., and Witt, R. (1986). Apply-

ing MRP II to remanufacturing: a case study. In APICS Conference
Proceedings, pages 18–21.
Bibliography 289
[23] Beukering, P. J. H. (2001). Recycling, international trade and the

environment: an empirical analysis. Kluwer Academic Publishers, Dor-
drecht, the Netherlands. Vrije Universiteit Amsterdam, PhD thesis.
[24] Beullens, P. (2001). Location, process selection, and vehicle routing
models for Reverse Logistics. PhD thesis, Catholic University of Leuven,
Leuven, Belgium. PhD Thesis.
[25] Beullens, P., van Oudheusden, D., and Cattrysse, D. (1998).
Bi-destination waste collection: impact of vehicle typre and operations
on transportation costs. working paper 98-23, University of Leuven,
Leuven, Belgium.
[26] Beullens, P., van Wassenhove, L. N., and van Oudheusden,
D. (2004). Collection and vehicle routing issues in reverse logistics. In
Dekker, R., Inderfurth, K., van Wassenhove, L., and Fleischmann, M.,
editors, Reverse Logistics. Reverse Logistics. Quantitative Models for
Closed-Loop Supply Chains, chapter 5. Springer-Verlag.
[27] Bloemhof-Ruwaard, J., Krikke, H., and van Wassenhove,
L. N. (2004). Or models for eco-eco closed-loop supply chains. In
Dekker, R., Inderfurth, K., van Wassenhove, L., and Fleischmann, M.,
editors, Reverse Logistics. Quantitative Models for Closed-Loop Supply
Chains, chapter 15. Springer-Verlag, Berlin, Germany.
[28] Boks and Tempelman (1997). Delphi forecasting of recycling tech-
nology to support design for end-of-life of consumer electronic products.
In Riitahuhta, A., editor, Proceedings of the 11th International Confer-
ence on Engineering Design, Tampere, Finland. ICED ’97.
[29] Box, G. E. P., Jenkins, G. M., and Reinsel, G. C. (1994). Time
Series Analysis - forecasting and control. Prentice-Hall Inc., New Jersey,
3rd edition.
[30] Bozer, Y. A., Meller, R. D., and Erlebacher, S. J. (1994).
An improvement-type layout algorithm for single and multiple floor
facilities. Management Science, 40(7):918–932.
[31] Bozer, Y. A. and White, J. (1984). Travel time models for auto-
mated storage/retrieval systems. IIE Transactions, 16(4):329–338.
[32] Bozer, Y. A. and White, J. (1990). Design and performance models
for end-of-aisle order picking systems. Management Science, 36(7):852–
866.
290 Bibliography
[33] Brodin, M. H. (2002). Logistics systems for recycling - on the influ-

ence of products, structures, relationships and power. PhD thesis, In-
stitute of Technology, Linköping Universitet, Linköping, Sweden. PhD
Thesis.
[34] Brundland, G. (1998). European union and the environment. Re-

port, European Union. Luxemburg.
[35] Buchanan, D. J. and Abad, P. L. (1998). Optimal policy for a pe-

riodic review returnable inventory system. IIE Transactions, 30:1049–
1055.
[36] Caldwell, B. (1999). Reverse logistics. InformationWeek, 729:48–56.
[37] Camm, F. (2001). Environmental management in proactive commercial

firms.
[38] Canon (2003). http://www.canon.com.
[39] Carrol, A. B. (1979). a three-dimensional conceptual model of cor-

porate performance. Academy of Management Review, 4(4):497–505.
[40] Carter, C. R. and Ellram, L. M. (1998). Reverse logistics: A

review of the literature and framework for future investigation. Inter-
national Journal of Business Logistics, 19(1):85–102.
[41] Cattani, K. and Souza, G. (2003). Inventory rationing and ship-

ment flexibility alternatives for direct market firms. Production and
Operations Management. (forthcoming).
[42] CERN (1997). Stores management & logistics lims.

http://asdbwww.cern.ch/lims.htm.
[43] CERN (1999). Cern stores. CERN Bulletin 47/99.
[44] CERN (2000). European organisation for nuclear research.

http://cern.web.cern.ch/CERN/.
[45] Chain Store Age (2002). Returns are products too.
[46] Chang, N.-B. and Wei, Y. L. (2000). Siting recycling drop-off sta-
tions in urban area by genetic algorithm-based fuzzy multi-objective
nonlinear integer programming modeling. Fuzzy Sets and Systems,
114:133–149.
Bibliography 291
[47] Cho, D. I. and Parlar, M. (1991). A survey of maintenance mod-

els for multi-unit systems. European Journal of Operational Research,
51:1–23.
[48] CLM (2003). The council of logistics management.

http://www.clm.org/.
[49] Coffman, E. G. J., Courcoubetis, C., Garey, M. R., Johnson,

D. S., Shor, P. W., Weber, R. R., and Yannakakis, M. (2000).
Bin packing with discrete item sizes, part 1: Perfect packing theorems
and the average case behavior of optimal packings. SIAM Journal on
Discrete Mathematics, 13(3):384–402.
[50] Corbett, C. J. and Tang, C. S. (1999). Designing supply contracts:

Contract type and information asymmetry. In Tayur, S., Ganeshan,
R., and Magazine, M., editors, Quantitative models for supply chain
management, pages 269–298. Kluwer Academic Publishers.
[51] Dalkey, N. and Helmer, O. (1963). An experimental application

of the delphi method to the use of experts. Management Science,
September:458–1467.
[52] Davis, S., Gerstner, E., and Hagerty, M. (1995). Money back
guarantees in retailing: matching products to consumer tastes. Journal
of Retailing, 71(1):7–22.
[53] Davis, S., Hagerty, M., and Gerstner, E. (1998). Return policies
and the optimal level of “hassle”. Journal of Economics and Business,
50:445–460.
[54] De Bodt, M. and van Wassenhove L.N. (1990). Cost increase due
to demand uncertainty in mrp lot sizing. Decision Sciences, 14(3):345–
353.
[55] De Brito, M. P. (2003). Promising areas for future research on re-

verse: an exploratory study. In 8th International Symposium of Logis-
tics, Sevilla, Spain. Logistics and Networked Organisations.
[56] De Brito, M. P. and de Koster, M. B. M. (2004). Product

and material returns: handling and warehousing issues. In Dekker,
R., Inderfurth, K., van Wassenhove, L., and Fleischmann, M., editors,
Reverse Logistics. Quantitative Models for Closed-Loop Supply Chains,
chapter 6. Springer-Verlag.
292 Bibliography
[57] De Brito, M. P. and Dekker, R. (2003). Modelling product returns

in inventory control - exploring the validity of general assumptions.
International Journal of Production Economics, 81–82:225–241.
[58] De Brito, M. P. and Dekker, R. (2004). A framework for reverse

logistics. In Dekker, R., Inderfurth, K., van Wassenhove, L., and Fleis-
chmann, M., editors, Reverse Logistics. Quantitative Models for Closed-
Loop Supply Chains, chapter 1. Springer-Verlag, Berlin, Germany.
[59] De Brito, M. P., Dekker, R., and Flapper, S. D. P. (2003).

Reverse logistics: a review of case studies. Report Series Research in
Management ERS-2003-012-LIS, Erasmus University Rotterdam, the
Netherlands.
[60] De Brito, M. P. and van der Laan, E. A. (2003). Inventory

management with returns: the impact of misinformation. Erasmus re-
search institute series, Erasmus University Rotterdam, Rotterdam, the
Netherlands.
[61] De Brito, M. P. and van Wassenhove, L. N. (2003). The future

of reverse logistics: a delphi study. working paper, Erasmus University
Rotterdam, the Netherlands.
[62] De Kool, R. (2002). Erp and reverse logistics. Master’s thesis, Rot-
terdam School of Management, Erasmus University Rotterdam, the
Netherlands. unpublished Master’s Thesis (in Dutch).
[63] De Koster, M. B. M. (1994). Performance estimation of pick-to-

belt orderpicking systems. European Journal of Operational Research,
72:558–573.
[64] De Koster, M. B. M. (1996). Ontwerp van een eenvoudig magazijn.

In Duijker, J. P., Koster, M. B. M. D., and van Amstel, M. J. P., editors,
Praktijkboek Magazijnen/Distributiecentra, chapter 3.14, pages 3.14–01
–3.14–25. Kluwer, Deventer, the Netherlands.
[65] De Koster, M. B. M., de Brito, M. P., and van Vendel, M. A.

(2002). How to organise return handling: an exploratory study with
nine retailer warehouses. International Journal of Retail and Distribu-
tion Management, 30:407–421.
[66] De Koster, M. B. M., Flapper, S. D. P., Krikke, H., and Ver-

meulen, W. S. (2003). Networks for the collection and processing of
Bibliography 293
end-of-life goods: a case example of implemmenting producer respon-

sibility for large white goods. In Flapper, S. D. P., van Nunen, J. A.
E. E., and van Wassenhove, L. N., editors, Managing closed loop supply
chains. Springer-Verlag, Heilderberg, Germany.
[67] De Koster, M. B. M., Flapper, S. D. P., Krikke, H. R., and
Vermeulen, W. S. (1999a). Reverse logistics in the wholesale white-
good sector. In Van Goor, A. R., Flapper, S. D. P., and Clement,
C., editors, Handboek Reverse Logistics. Kluwer, B.V., Deventer, the
Netherlands. (in Dutch).
[68] De Koster, M. B. M. and Neuteboom, J. (2001). The logistics of
supermarket chains. Elsevier, Doetinchem, the Netherlands.
[69] De Koster, M. B. M., van der Poort, E. S., and Wolters, M.
(1999b). Efficient order batching methods in warehouses. International
Journal of Production Research, 37(7):1479–1504.
[70] Debo, L. S. and van Wassenhove, L. N. (2004). Coordination in
closed-loop supply chains. In Dekker, R., Inderfurth, K., van Wassen-
hove, L., and Fleischmann, M., editors, Reverse Logistics. Quantitative
Models for Closed-Loop Supply Chains, chapter 12. Springer-Verlag.
[71] Dekker, R., de Koster, M. B. M., van Kalleveen, H., and
Roodbergen, K. J. (2002). Quick response practices at the ware-
house of ankor. working paper, Erasmus University Rotterdam, the
Netherlands.
[72] Dekker, R., Inderfurth, K., van Wassenhove, L., and Fleis-
chmann, M. (2004). Reverse Logistics. Quantitative Models for Closed-
Loop Supply Chains. Springer-Verlag, Berlin, Germany. forthcoming.
[73] Dekker, R. and van der Laan, E. (1999). Gestion des stocks pour
une fabrication et une refabrication simultanes: synthese de resultats
recents. Logistique & Management, 7(2). (in French).
[74] Dekker, R. and van der Laan, E. (2003). Inventory control in
reverse logistics. In Guide Jr., D. and van Wassenhove, L. N., editors,
Closed-Loop Supply Chains. Carnegie Mellon University Press, Pitts-
burgh, Pennsylvania, U.S.A.
[75] Delbecq, A. L., van de Ven, A. H., and Gustafson, D. H. (1975).
Group Techniques for Program Planning, a guide to nominal group and
Delphi processes. Scott, Foresman and Company,, USA.
294 Bibliography
[76] DelCastillo, E. and Cochran, J. (1996). ptimal short horizon

distribution operations in reusable container systems. Journal of the
Operational Research Society, 47(1):48–60.
[77] Desproges, J. and David, G. (1998). Le contenu en emplois des fil-

ires de collecte et de traitements des dchets mnagers. Report, ADEME.
in French.
[78] Dethloff, J. (2001). Vehicle routing and reverse logistics: the vehicle
routing problem and simultaneos delivery and pick-up. OR Spektrum,
23:79–96.
[79] Diaz, A. and Fu, M. C. (1997). Models for multi-echelon repairable

item inventory systems with limited repair capacity. European Journal
of Operational Research, 97(3):480–492.
[80] Dijkhuizen, H. P. (1997). Reverse Logistics at IBM. In Van Goor,

A., Flapper, S., and Clement, C., editors, Reverse Logistics, chapter
E1310. Kluwer, B.V., The Netherlands. (In Dutch).
[81] Dillman, D. A. (1978). Mail and telephone surverys, the total design
method. John Wiley & Sons, Inc., USA.
[82] Donker, S. and van der Ploeg, T. (2001). Multi-echelon inventory

control of repairable spare parts at Lucent Technologies Nederland. In
Van Goor, A., Flapper, S., and Clement, C., editors, Handboek Reverse
Logistics. Kluwer, B.V., Deventer, The Netherlands. (In Dutch).
[83] Doppelt, B. and Nelson, H. (2001). Extended producer respons-

ability and product take-back: Applications for the pacific northwest.
Report, The Center for Watershed and Community Health, Mark O.
Hatfield School of Government, Portland State University, Portland,
Oregon, USA.
[84] Drejer, A., Blackmon, K., and Voss, C. (2002). Worlds apart? - a
look at the operations management are in the US, UK and Scandinavia.
Scandinavia Journal of Management, 16:45–66.
[85] Driesch, H. M., Flapper, S. D. P., and van Oyen, J. E. (1998).

Logistic control of engine re-use at Daimler-Benz MTR. In Van Goor,
A., Flapper, S., and Clement, C., editors, Handboek Reverse Logistics.
Kluwer, B.V., Deventer, The Netherlands. (In Dutch).
Bibliography 295
[86] Duhaime, R., Riopel, D., and Langevin, A. (2000). Value analysis
and optimization of reusable containers at Canada Post. Interfaces,
31(3):3–15.
[87] Duijker, J. P., de Koster, M. B. M., and van Amstel, M. P., editors
(1994–2000). Praktijkboek Magazijnen/Distributiecentra. Ten Hagen
Stam, The Hage, the Netherlands. in Dutch.
[88] ELA (1999). Insight to impact, results or the fourth quinquennial Euro-
pean logistics study. European Logistics Association (ELA). Brussels.
[89] Emmons, H. and Gilbert, S. (1998). Note. the role of return poli-
cies pricing and inventory decisions for catalogue goods. Management
Science, 44(2):276–283.
[90] EUREKA (2003). A network for market oriented R&D,.

www.eureka.be/.
[91] EURO (2003). http://www.euro-online.org/.
[92] EUROPA (2003). Europa, the European Union online.

http://www.europa.eu.int/.
[93] European Environmental Agency (2002). Environmental signals

2002 - benchmarking the millennium. Environmental Assessment Re-
port 9, European Environmental Agency (EEA), Copenhagen, Den-
mark.
[94] Faria de Almeida, A. and Robertson, A. (1995). Domestic batter-

ies and the environment: A life-cycle approach to consumer electronic
products. In Proceedings of the International Conference Clean Elec-
tronics Products & Technology, pages 162–167, Edinburgh, Scotland.
[95] Farrow, P. H. and Jonhson, R. R. (2000). Entrepreneurship, inno-

vation and sustainability strategies at walden paddlers, inc. Interfaces,
30(3):215–225.
[96] Fernandéz, I. (2003). The concept of Reverse Logistics. A review

of litertaure. In Juga, J., editor, Proceedings of 15th Annual Confer-
ence for Nordic Researchers in Logistics, NOFOMA 2003, “Stiving for
Leading Edge Logistics”, pages 464–478, Oulu, Finland. NOFOMA.
[97] Fernie, J. and Hart, C. (2001). Uk packaging waste legislation.

implications for food retailers. British Food Journal, 103(3):187–197.
296 Bibliography
[98] Ferrer, G. and Ketzenberg, M. (2003). Value of yield information

for remanufacturing complex products. Working paper, Kenan-Fleger
Business School.
[99] Flapper, S., van Nunen, J., and van Wassenhove, L. (2003).
Managing closed loop supply chains. Springer-Verlag, Heilderberg, Ger-
many. (forthcoming).
[100] Fleischman, E. A. and Mumford, M. D. (1991). Evaluating clas-

sifications of job behaviour: a construct validation of the ability require-
ment scales. Personnel Psychology, 44:523–575.
[101] Fleischmann, M. (2000). Quantitative Models for Reverse Logistics.

PhD thesis, Erasmus University Rotterdam, Rotterdam, the Nether-
lands. PhD Thesis.
[102] Fleischmann, M., Bloemhof-Ruwaard, J., Beullens, P., and

Dekker, R. (2004). Reverse logistics network design. In Dekker,
chapter 4. Springer-Verlag, Berlin, Germany.
[103] Fleischmann, M., Bloemhof-Ruwaard, J. M., Dekker, R.,

van der Laan, E. A., van Nunen, J. A. E. E., and van Wassen-
hove, L. N. (1997). Quantitative models for reverse logistics: A review.
European Journal of Operational Research, 103:1–17.
[104] Fleischmann, M., Krikke, H. R., Dekker, R., and Flapper,

S. D. P. (2000). A characterisation of reverse logistics networks for
product recovery. Omega, 28(6):653–666.
[105] Fleischmann, M., Kuik, R., and Dekker, R. (2002). Controlling

inventories with stochastic returns: a basic model. European Journal
of Operational Research, 138(1):63–75.
[106] Fontrodona, J. (2002). Pragmatism and Management inquiry.

Quorum Books, Westport, USA.
[107] Fuller, D. A. and Allen, J. (1997). A typology of reverse chan-

nel systems for post-consumer recyclables. In Polonsky, J. and Mintu-
Winsatt, A. T., editors, Environmental marketing: strategies, practice,
theory and research. Haworth Press, Binghamton, NY.
Bibliography 297
[108] Gademann, A. J. R. M., van den Bergand, J., and van der
Hoff, H. H. (2001). An order batching algorithm for wave picking, in
a parallel-aisle warehouse. IIE Transactions, 33(5):385–398.
[109] Gan, F. and Koehler, K. (1990). Goodness-of-fit tests based on

p-p probability plots. Technometrics, 32(3):289–303.
[110] Ganeshan, R., Jack, E., Magazine, M. J., and Stephens, P.

(1999). A taxonomic review of supply chain management research. In
Tayur, S., Ganesham, R., and Magazine, M. J., editors, Quantitative
Models for Supply Chain Management. Kluwer Academic Publishers,
Massachussets, USA.
[111] Gardner, E. (1990). Evaluating forecast performance in an inventory

control system. Management Science, 36:490–499.
[112] Garner, A. and Keoleian, G. A. (1995). Industrial Ecology: an

introduction. National Pollution Prevention Center, Michigan, USA.
[113] Gentry, C. R. (1999). Reducing the cost of returns. Chain Store

Age, 75(10):124–126.
[114] Global Investment Recovery (2003).

http://girpreciousmetals.com.
[115] Goetschalckx, M. (1992). An interactive layouot heuristic based

on hexagonal adjacency graphs. European Journal of Operational Re-
search, 63:304–321.
[116] Goetz, W. G. and Egbelu, P. J. (1990). Guide path design and

location of load pick-up/drop-off points for an automated guided vehicle
system. International Journal of Production research, 28(5).
[117] Goggin, K. and Browne, J. (1978). Reverse distribution channels

for recycling. California Management Review, 20(3):72–81.
[118] Goggin, K. and Browne, J. (2000). Towards a taxonomy of

resource recovery from end-of-life products. Computers in Industry,
42:177–191.
[119] Goh, T. N. and Varaprasad, N. (1986). A statistical methodology

for the analysis of the life-cycle of reusable containers. IEE Transac-
tions, 18:42–47.
298 Bibliography
[120] Gooley, T. (1998). Reverse logistics: five steps to success. Logistics

Management and Distribution Report, pages 49–54.
[121] Guba, E. G. and Lincoln, Y. S. (1994). Competing paradigms

in qualitative research. In Denzin, N. K. and Lincoln, Y. S., editors,
Handbook of qualitative research. SAGE, Thousand Oaks, CA, USA.
[122] Guide Jr., V. D. R. (2000b). Technical note: Production planning

and control for remanufacturing: industry practice and research needs.
Journal of Operations Management, 18:467–483.
[123] Guide Jr., V. D. R., Jayamaran, V., Srivastava, R., and Ben-
ton, W. C. (2000). Supply chain management for recovarable manu-
facturing systems. Interfaces, 30(3):125–142.
[124] Guide Jr., V. D. R. and Spencer, M. (1997). Rough-cut capacity

planning for remanufacturing firms. Production Planning & Control,
8(3):237–244.
[125] Guide Jr., V. D. R. and Srivastava, R. (1997a). Buffering from

material recovery uncertainty in a recoverable manufacturing environ-
ment. Journal of the Operational Research Society, 8(48):519–529.
[126] Guide Jr., V. D. R. and Srivastava, R. (1997b). Repairable inven-

tory theory: models and applications. European Journal of Operational
Research, 102:1–20.
[127] Guide Jr., V. D. R. and Srivastava, R. (1998). Inventory buffers

in recoverable manufacturing. Journal of Operations Management,
106:551–568.
[128] Guide Jr., V. D. R., Srivastava, R., and Spencer, M. S. (1997).

An evaluation of capacity planning techniques in a remanufacturing
environment. International Journal of Production Research, 35:67–82.
[129] Guide Jr., V. D. R. and van Wassenhove, L. N. (2000). Closed-

loop supply chains. working paper 2000/75/TM/CIMSO 15, INSEAD,
Fontainebleau Cedex, France.
[130] Guide Jr., V. D. R. and van Wassenhove, L. N. (2001). Man-

aging product returns for remanufacturing. Production and Operations
Management, 10:142–155.
Bibliography 299
[131] Guide Jr., V. D. R. and van Wassenhove, L. N., editors (2003). Busi-
ness Perspectives in Closed-Loop Supply Chains. Carnegie Mellon Uni-
versity Press. Pittsburgh, Pennsylvania, USA.
[132] Guiltinan, J. and Nwokoye, N. (1974). Reverse channels for re-
cycling: an analysis for alternatives and public policy implications. In
Curhan, R. G., editor, New marketing for social and economic progress,
Combined Proceedings. American Marketing Association.
[133] Gungor, A. and Gupta, S. M. (1999). Issues in environmentally
conscious manufacturing and product recovery: a survey. Computers &
Industrial Engineering, 36:811–853.
[134] Gupta, T. and Chakraborty, S. (1984). Looping in a multi-
stage production system. International Journal of Production Research,
22:299–311.
[135] Hackman, S. T. and Platzman, L. K. (1990). Near-optimal so-
lution of generalized resource allocation problems with large capacities.
Operations Research, 38(5):902–910.
[136] Hausman W. H., L. B. Schwarz, S. C. G. (1976). Optimal storage
assignment in automatic warehousing systems. Management Science,
22(6):629–638.
[137] Healy, M. and Perry, C. (2000). Comprehensive criteria to
judge validity and reliability of qualitative research within the real-
ism paradigm. Qualitative Market Research: An International Journal,
3(3):118–126.
[138] Hendrickson, C., Matthews, H. S., Cagan, J., and
McMichael, F. (2003). Green engineering. In Guide Jr., V. D. R.
and van Wassenhove, L. N., editors, Business Perspectives in Closed-
Loop Supply Chains. Carnegie Mellon Press, U.S.A.
[139] Heskett, J. L. (1963). Cube-per-order index - a key to warehouse
stock location. Transportation and Distribution Management, 3:27–31.
[140] Hess, J., Chu, W., and Gerstner, E. (1996). Controlling product
returns in direct marketing. Transportation and Distribution Manage-
ment, 7(4):307–317.
[141] Heyman, D. P. (1977). Optimal disposal policies for a single-item
inventory system with returns. Naval Research Logistics Quarterly,
24:385–405.
300 Bibliography
[142] IGD.com (2002). Information, research and education, fact sheets.

http://www.igd.com/.
[143] Indefurth, K. and van der Laan, E. A. (2001). Leadtime effects

and policy improvement for stochastic inventory control with remanu-
facturing. International Journal of Production Economics, 71(1-3):381–
390.
[144] Inderfurth, K. (1997). Simple optimal replenishment and disposal

policies for a product recovery with leadtimes. OR Spektrum, 19:111–
122.
[145] Inderfurth, K., Flapper, S. D. P., Lambert, A. J. D., Pappis,

C., and Voutsinas), T. G. (2004). Production planning for product
recovery management. In Dekker, R., Inderfurth, K., van Wassenhove,
L., and Fleischmann, M., editors, Reverse Logistics. Quantitative Mod-
els for Closed-Loop Supply Chains, chapter 10. Springer-Verlag, Berlin,
Germany. forthcoming.
[146] Inderfurth, K. and Jensen, T. (1999). Analysis of mrp policies

with recovery options? In et al., U. L.-W., editor, Modelling and De-
cisions in Economics, pages 189–228. Physica-Verlag, Heidelberg-New
York.
[147] Jahre, M. (1995). Logistics Systems for recycling - efficient collection

of household waste. PhD thesis, Chalmers University of technology,
Götenberg, Sweden.
[148] Jordan, K., Gonser, J., Radermaker, F., and Jorgensen,

R. (2001). European commission dgxi.e.3, european packaging waste
mangement systems. Report, ARGUS.
[149] Kanji, G. K. (1999). 100 Statistical tests. SAGE Publications.
[150] Kelle, P. and Silver, E. (1989). Forecasting the returns of reusable

containers. Journal of Operations Management, 8(1):17–35.
[151] Kenis, D. (1995). Improving group decisions. Designing and testing

techniques for group decisions support systems, applying Delphi tech-
niques. PhD thesis, University of Utrecht, Utrecht, the Netherlands.
[152] Kiesmüller, G. P. (2003). The use of information in a one depot

N retailer system with returns. International Journal of Production
Economics. forthcoming.
Bibliography 301
[153] Kiesmüller, G. P. and van der Laan, E. (2001). An inventory

model with dependent product demands and returns. International
Journal of Production Economics, 72(1):73–87.
[154] Klausner, M. and Hendrickson, C. (2000). Reverse-logistics
strategy for product take-back. Interfaces, 30(3):156–165.
[155] Klausner M., W. M. G. and Hendrickson, C. (1998). Reuse
of electric motors in con-sumer products, design and analysis of an
electronic data log. Journal of Ind. Ecology, 2(2):89–102.
[156] Kleijn, M. and Dekker, R. (1999). An overview of inventory sys-
tems with several demand classes. In Grazia Speranza, M. and Stahly,
P., editors, New Trends in Distribution Logistics, volume 480 of Lec-
ture Notes in Economics and Mathematical Systems. Springer Verlag,
Berlin.
[157] Kleineidam, U., Lambert, A. J. D., Blansjaar, J., Kok, J.,
and Heijningen, R. (2000). Optimising product recycling chains
by control theory. International Journal of Production Economics,
66(2):185–195.
[158] Kobeissi, A. (2001). Contribution a la mise en ouvre du systeme
d’evaluation des produits techniques en fin de vie. PhD thesis, Institut
National Polytechnique de Grenoble, Grnoble, France. PhD Thesis (in
French).
[159] Koehorst, H., de Vries., H., and Wubben, E. (1999). Standard-
isation of crates: lessons learned from the versfust (freshcrate) project.
Supply Chain Management, 49(2):95–101.
[160] Kokkinaki, A., Zuidwijk, R., van Nunen, J., and Dekker, R.
(2004). Information and communication technology enabling reverse
logistics. In Dekker, R., Inderfurth, K., van Wassenhove, L., and Fleis-
chmann, M., editors, Reverse Logistics. Quantitative Models for Closed-
Loop Supply Chains, chapter 16. Springer-Verlag.
[161] Kopicky, R. J., Berg, M. J., Legg, L., Dasappa, V., and Mag-
gioni, C. (1993). Reuse and recycling: reverse logistics opportunities.
Council of Logistics Management, Oak Brook, IL.
[162] Krikke, H. (1998). Recovery strategies and reverse logistic network
design. PhD thesis, Institute for Business Engineering and Technology
Application (BETA), University of Twente, Twente, the Netherlands.
302 Bibliography
[163] Krikke, H., le Blanc, I., and van de Velde, S. (2003). Creating
value from returns the impact of product life cycle management on cir-
cular supply chains - and reverse. Report, Center of Applied Research,
Tilburg University, Tilburg, the Netherlands.
[164] Krikke, H., van Harten, A., and Schuur, P. (1999a). Business
case oce: Reverse logistic network re-design for copiers. OR Spektrum,
34(3):381–409.
[165] Krikke, H., van Harten, A., and Schuur, P. (1999b). Business
case roteb: recovery strategies for monitors. Computers & Industrial
Engineering, 34(4):739–757.
[166] Kroon, L. and Vrijens, G. (1995). Returnable containers: An ex-

ample of reverse logistics. International Journal of Physical Distribution
& Logistics Management, 25(2):56–68.
[167] Krumwiede, D. W. and Sheu, C. (2002). A model for reverse

logistics entry by third-party providers. Omega, 30(5):325–333.
[168] Lambert, D. M. and Stock, J. R. (1981). Strategical Physical

Distribution Mangement. Homewood,, IL.
[169] Landers, T. L., Cole, M. H., Walker, B., and Kirk, R. W.

(2000). The virtual warehousing concept. Transportation Research
Part-E- Logistics and Transportation Review, 36(2):115–125.
[170] Landrieu, A. (2001). Logistique inverse et collecte des produits tech-

niques en fin de vie. Tourneées de véhicles avec contraintes. PhD thesis,
Institut National Polytechnique de Grenoble, Grenoble, France. PhD
Thesis (in French).
[171] Lariviere, M. A. (1999). Supply chain contracting and coordination

with stochastic demand. In Tayur, S., Ganeshan, R., and Magazine, M.,
editors, Quantitative models for supply chain management, pages 233–
268. Kluwer Academic Publishers.
[172] Laxe, F. G. (1999). Protection de l’environnement comme source

d’emplois. Report, Council of Europe.
[173] Lee, H., Padmanabhan, V., and Whang, S. (1997). Information

distortion in a supply chain: the bullwhip effect. Management Science,
43(4):546–558.
Bibliography 303
[174] Linstone, H. A. and Turoff, M. (1975). The Delphi method.

Techniques and applications. Addison-Wesley, London, U.K.
[175] Linton, J. D. and Jonhson, D. A. (2000). A decision support
system for planning remanufacturing at nortel networks. Interfaces,
30(6):17–31.
[176] Listes, O. and Dekker, R. (2001). Stochastic approaches for prod-
uct recovery network design: a case study. Econometric Institute Re-
port EI 2001-08, Erasmus University Rotterdam, the Netherlands.
[177] LOCOMOTIVE (1993). Logistics chain multidimen-
sional design toolbox with environmental assessment.
http://www.newcastle.research.ec.org/esp-syn/text/8615.html.
[178] Louwers, D., Kip, B. J., Peters, E., Souren, F., and Flapper,
S. D. P. (1999). A facility location allocation model for reusing carpet
materials. Computers & Industrial Engineering, 36:855–869.
[179] Luo, Z.-Q., Pong, J.-S., and Ralph, D. (1996). Mathematical Pro-
gramming with Equilibrium Constraints. Cambridge University Press,
Cambridge, UK.
[180] Lützenbauer, M. (1993). Mehrwegsysteme für transportverpackun-
gen (systems for returnable transport packaging). Deutscher Fachverlag
GmbH.
[181] Mahadevan, B., Pyke, D. F., and Fleischmann, M. (2002). Peri-
odic review, push inventory policies for remanufacturing. ERIM Report
Series ERS-2002-35-LIS, Faculty of Business Administration, Erasmus
University Rotterdam, The Netherlands.
[182] Martino, J. (1972). Technological forecasting for decision making.
Elsevier, New York, USA.
[183] Maslennikova, I. and Foley, D. (2000). Xerox’s approach to sus-
tainability. Interfaces, 30(3):226–233.
[184] Matson, J. O. and White, J. A. (1982). European journal of
operational research. Omega, 11:309–318.
[185] McGavis, D. (1994). The energy bucket and a not-so-drop-in-the-
bucket portion of waste stream, consumables. pages 267–272, San Fran-
cisco, California. IEEE International Symposium on Electronics & the
Environment.
304 Bibliography
[186] Meijer, H. W. (1998). Green logistics. In Van Goor, A. R., Flap-

per, S. D. P., and Clement, C., editors, Handboek of Reverse Logistics.
Kluwer B. V., Deventer, The Netherlands. (in Dutch).
[187] Melissen, F. W. and de Ron, A. (1999). Defining recovery prac-

tices - definitions and terminology. International Journal on Environ-
mentally Conscious Manufacturing and Design, 8(2):1–18.
[188] Merriam-Webster (2003). Merriam-webster online - the language

center. http://www.m-w.com/home.htm.
[189] Meyer, H. (1999). Many happy returns. The Journal of Business

Strategy, 20(4):27–31.
[190] MHIA (2002). Material handling industry of america. www.mhia.org.
[191] Minner, S. and Kleber, R. (2001). Optimal control of produc-

tion and remanufacturing in a simple recovery model with linear cost
functions. OR Spektrum, 23:3–24.
[192] Moffat, J. (1992). Resources for the repair of non-modular aero-

engines. Omega, 20(1):99–103.
[193] Moorthy, S. and (1995), K. S. (1995). Signaling quality with a

money-back guarantee:the role of transaction costs. Marketing Science,
14:442–446.
[194] Morphy, E. (2001). Newgistics gears up to deliver many happy re-

turns. CRMDaily.com. http://www.crmdaily.com.
[195] Mostard, J. and Teunter, R. (2002). The newsboy problem with

resalable returns. ERIM Report Series ERS-2002-89-LIS, Faculty of
Business Administration, Erasmus University Rotterdam, The Nether-
lands.
[196] Muckstadt, J. A. and Isaac, M. H. (1981). An analysis of single

item inventory systems with returns. Naval Research Logistics Quar-
terly, 28(1):237–254.
[197] Murphy, P. R. (1986). A preliminary study of transportation and

warehousing aspects of reverse distribution. Transportation Journal,
35(4):12–21.
Bibliography 305
[198] Murphy, P. R. and Poist, R. F. (2003). Green perspectives and

practices: a “comparative logistics” study. Supply Chain Management:
An International Journal, 8:122–131.
[199] Murphy, P. R. and Poist, R. P. (1989). Managing of logistics
retromovements: an empirical analysis of literature suggestions. Trans-
portation Research Forum, 29(1):177–84.
[200] Muther, R. (1973). Systematic layout planning. Cahners Books,
Boston, USA, 2nd edition edition.
[201] Nagel, C. and Meyer, P. (1999). Caught between ecology and
economy: end-of-life aspects of environmentally conscious manufactur-
ing. Computers & Industrial Engineering, 36(4):781–792.
[202] Office for Fair Trading (2003). http://www.oft.gov.uk/.
[203] Olsen, R. (2000). A checklist for e-commerce. Discount Store News,
39(4):15–16.
[204] O’Neill, S. and Chu, J. (2001). Online returns management strate-
gies. Executive strategy report, IBM Corporation.
[205] Pappis, C., Stavros, E., and Tsoulfas, G. (2004). Reverse logis-
tics and environmental management. In Dekker, R., Inderfurth, K., van
Wassenhove, L., and Fleischmann, M., editors, Reverse Logistics. Quan-
titative Models for Closed-Loop Supply Chains, chapter 14. Springer-
Verlag, Berlin, Germany.
[206] Piron, F. and Young, M. (2000). Retail borrowing: insights and
implications on returning used merchandise. International Journal of
Retail & Distribution Management, 28(1):27–36.
[207] Pohlen, T. L. and Farris, T. (1992). Reverse logistics in plastics
recycling. International Journal of Physical Distribution & Logistics
Management, 22(7):35–47.
[208] Rabinovich, E., Windle, R., Dresner, M., and Corsi, T.
(1999). Outsourcing of integrated logistics functions. International
Journal of Physical Distribution and Logistics Management, 29:353–
273.
[209] Realff, M. J., J. C. A. and Newton, D. (2000). Strategic design
of reverse production systems. Computers and Chemical Engineering,
24:991–996.
306 Bibliography
[210] ReCellular (2003). http://recellular.com.
[211] Reloop (1998-). Reloop, Reverse Logistics Chain Optimization.

http://www.kuleuven.ac.be/cib/projkt03.htm.
[212] RevLog (1998-). Revlog, the european working group on reverse

logistics. http://www.fbk.eur.nl/OZ/REVLOG/.
[213] Richter, K. (1996). The extended EOQ repair and waste model.
International Journal of Production Economics, 1–3:443–448.
[214] RLEC (2003). RLEC, The Reverse Logistics Executive Council.

http://www.rlec.org/index.html.
[215] Robison, S. J. (1992). The world of a class ”A” remanufacturing

company. In APICS Remanufacturing Seminar Proceedings, pages 1–3.
APICS.
[216] Rodrigue, J.-P., Slack, B., and Comtois, C. (2001). Green

logistics. In Brewer, A., Button, K., and Henshe, D., editors, The
Handbook of Logistics and Supply-Chain Management, Handbooks in
Transport N. 2. Pergamon/Elsevier, London, “UK”.
[217] Rogers, D. S. and Tibben-Lembke, R. S. (1999). Going Back-

wards: reverse logistics trends and practices. Reverse Logistics Execu-
tive Council, Pittsburgh, PA.
[218] Rogers, D. S. and Tibben-Lembke, R. S. (2001). An examination

of reverse logistics practices. Journal of Business Logistics, 22(2):129–
248.
[219] Roodbergen, K. (2001). Layout and routing methods for warehouses.

lands.
[220] Ross, S. (1996). Stochastic Processes. John Wiley & Sons, Inc., 2nd
edition.
[221] Rouwenhorst, B., Reuter, B., Stockrahm, V., van Houtum,

G. J., Mantel, R. J., and Zijm, W. H. M. (2000). Warehouse
design and control: framework and literature review. European Journal
of Operational Research, 122:515–533.
Bibliography 307
[222] Rudi, N., Pyke, D. F., and Sporsheim, P. O. (2000). Product

recovery at the norwegian national insurance administration. Interfaces,
30(3):166–179.
[223] Sanders, H., van Harten, A., and van der Heijden, M. C.
(2000). Logistics and reverse logistics at Wehkamp. In Van Goor, A.,
Flapper, S. D. P., and Clement, C., editors, Handboek of Reverse Logis-
tics. Kluwer B. V., Deventer, The Netherlands. (in Dutch).
[224] Schiffeleers, S. (2001). Reduction of distribution costs for emi

music in europe. Master’s thesis, Eindhoven University of Technology,
the Netherlands. unpublished MSc Thesis.
[225] Schinkel, H. (2000). Plaster remnants should not be thrown away!

In Van Goor, A., Flapper, S. D. P., and Clement, C., editors, Handboek
of Reverse Logistics. Kluwer B. V., Deventer, The Netherlands. (in
Dutch).
[226] Schmidt, R. A., Sturrock, F., Ward, P., and Lea-

Greenwood, G. (1999). Deshopping - the art of illicit consumption.
International Journal of Retail & Distribution Management, 27(8):290–
301.
[227] Schrady, D. A. (1967). A deterministic inventory model for re-

pairable items. Naval Research Logistics Quarterly, 14:391–398.
[228] Silver, E. A., Peterson, R., and Pyke, D. F. (1998). Inventory

management and production planning and scheduling. John Wiley &
Sons, New York, USA, 3rd edition edition.
[229] Simons, P. H. W. (1998). Reverse logistics at Trespa International

B.V. In Van Goor, A., Flapper, S. D. P., and Clement, C., editors,
Handboek of Reverse Logistics. Kluwer B. V., Deventer, The Nether-
lands. (in Dutch).
[230] Simpson, V. (1978). Optimum solution structure for a repairable

inventory problem. Operations Research, 26:270–281.
[231] Slackman, H. (1975). Delphi critique. Lexington,, Massachusetts,

USA.
[232] Smit, W. and Popma, W. (2002). Complacency, indecisiveness and

diy-centres threaten dutch garden centres’ flourishing market position.
European Retail Digest, European Regional Review, 34:36–40.
308 Bibliography
[233] Smith, N. C., Thomas, R. J., and Quelch, J. A. (2001). A strate-

gic approach to managing returns. Harvard Business Review, pages
102–112.
[234] Spengler, T., Puchert, H., Penkuhn, T., and Rentz, O.
(1997). Capacitated remanufacturing with service level constraints. Eu-
ropean Journal of Operational Research, 97(2):308/326.
[235] STIBAT (2003). STIBAT, the Dutch battery foundation.
http://www.stibat.nl/.
[236] Stock, J. R. (1992). Reverse Logistics. Council of Logistics Man-
agement, Oak Brook, IL.
[237] Stock, J. R. (1998). Development and implemmentation of Reverse
Logistics programs. Council of Logistics Management, Oak Brook, IL.
[238] Stock, J. R. (2001). The 7 deadly sins of reverse logistics. Material
Handling Management, 56(3):5–11.
[239] Stock, J. R. and Lambert, D. M. (2001). Strategic Logistics man-
agement. McGraw-Hill International edition, Singapore.
[240] Strauss, A. and Corbin, J. (2000). Grounded theory methodology.
In Denzin, N. K. and Lincoln, Y. S., editors, Handbook of qualitative
research. Sage, Thousand Oaks, CA, USA. 2nd edition.
[241] Stroeker, N. E. (1995). Second-hand markets for consumer
durables. PhD thesis, Erasmus University Rotterdam, Rotterdam, the
Netherlands. PhD Thesis, Tinbergen Institute research Series.
[242] Swinkels, J. and van Esch, C. F. A. (1998). Return packaging
and turnover rates. In Van Goor, A., Flapper, S. D. P., and Clement,
C., editors, Handboek of Reverse Logistics. Kluwer B. V., Deventer, The
[243] ’t Slot, W. M. E. and van Amstel, J. J. A. P. (1999). Expe-
riences from the national trial project apparetour. In Van Goor, A.,
Flapper, S. D. P., and Clement, C., editors, Handboek of Reverse Logis-
tics. Kluwer B. V., Deventer, The Netherlands. (in Dutch).
[244] Tashakkori, A. and C.Teddlie (1998). Mixed methodology. com-
bining qualitative and quantitative approaches. In Applied Social Re-
search Methods Series, volume 46. SAGE, Thousand Oaks, California,
USA.
Bibliography 309
[245] Teunter, R. (2001). Economic ordering quantities for remanufac-

turable item inventory systems. Naval Research Logistics, 48(6):484–
495.
[246] Teunter, R., Inderfurth, K., Minner, S., and Kleber, R.

(2000). Reverse logistics in a pharmaceutical company: a case. working
paper, Otto von Guericke-University, Magdeburg, Germany.
[247] Teunter, R. and Vlachos, D. (2002). On the necessity of a dis-

posal option for returned items that can be remanufactured. Interna-
tional Journal of Production Economics, 75(3):257–266.
[248] Thierry, M., Salomon, M., van Nunen, J. A. E. E., and van
Wassenhove, L. N. (1995). Strategic issues in product recovery man-
agement. California Management Review, 37(2):114–135.
[249] Thomas Jr., M. (1997). Emerging technology: production schedul-

ing matures. IIE Solutions, pages 25–29.
[250] Tibben-Lembke, R. S. and Rogers, D. S. (2002). Differences

between forward and reverse logistics in a retail environment. Supply
Chain Management: An International Journal, 7(5):271–282.
[251] Tichy, N., McGill, A. R., and Clair, L. S. (1997). Corporate

global citizenship. The new Lexington Press, San Francisco, USA.
[252] Tijms, H. C. (1994). Stochastic models, an algorithmic approach.

John Wiley & Sons.
[253] Tijms, H. C. (2003). A first course in stochastic models. Wiley, West

Sussex, England.
[254] Toktay, L. B. (2003). Forecasting product returns. In Guide Jr.,

V. D. R. and van Wassenhove, L. N., editors, Business Perspectives in
Full Cycle Supply Chains. Carnegie Mellon UniversityPress, Pittsburgh,
Pennsylvania, USA.
[255] Toktay, L. B., van der Laan, E. A., and de Brito, M. P.

(2004). Managing product returns: The role of forecasting. In Dekker,
chapter 3. Springer-Verlag, Berlin, Germany.
310 Bibliography
[256] Toktay, L. B., Wein, L. M., and Zenios, S. A. (2000). Inven-

tory management of remanufacturable products. Management Science,
46(11):1412–1426.
[257] Tompkins, J., White, J., Bozer, Y., Frazelle, E., Tanchoco,
J., and Trevino, J. (1996). Facilities planning. Wiley, New York.
[258] Trebilcock, B. (2002). Return to sender. Warehousing Manage-
ment, 9(4):24–27.
[259] Tsay, A. A., Nahmias, S., and Agrawal, N. (1999). Modeling
supply chain contracts: A review. In Tayur, S., Ganeshan, R., and
Magazine, M., editors, Quantitative models for supply chain manage-
ment, pages 299–336. Kluwer Academic Publishers.
[260] Ubbens, U. M. P. (2000). The closed loop of metal packaging. In
Van Goor, A., Flapper, S. D. P., and Clement, C., editors, Handboek
of Reverse Logistics. Kluwer B. V., Deventer, The Netherlands. (in
Dutch).
[261] UGSG (2001). Obsolete computers, “gold mine”or high-teck trash?
resource recovery recycling. USGS Fact Sheet FS-060-01, U.S. Depart-
ment of the Interior, Denver, U.S.A. U.S. Geological Survey.
[262] United Nations (2003). http://esa.un.org.
[263] Van Burik, A. M. C. (1998). Return flow problems of car wreck ma-
terials. In Van Goor, A., Flapper, S. D. P., and Clement, C., editors,
Handboek of Reverse Logistics. Kluwer B. V., Deventer, The Nether-
lands. (in Dutch).
[264] Van den Berg, J. P. (1999). A literature survey on planning an
control of warehousing systems. IIE Transactions, 31:751–762.
[265] Van den Berg, J. P. and Zijm, W. H. M. (1999). Models for ware-
house management: classification and examples. International Journal
of Production Economics, 59:519–528.
[266] Van der Laan, E. A. (1997). The effects of remanufacturing on
inventory control. PhD thesis, Erasmus University Rotterdam, Rotter-
dam, the Netherlands.
[267] Van der Laan, E. A., Kiesmüller, G., Kuik, R., Vlachos, D.,
and R.Dekker (2004). Stochastic inventory control for product re-
covery management. In Dekker, R., Inderfurth, K., van Wassenhove,
Bibliography 311
L., and Fleischmann, M., editors, Reverse Logistics. Quantitative Mod-

els for Closed-Loop Supply Chains, chapter 8. Springer-Verlag, Berlin,
Germany.
[268] Van der Laan, E. A., Salomon, M., Dekker, R., and van
Wassenhove, L. N. (1999). Inventory control in hybrid systems with
remanufacturing. Management Science, 45(5):733–747.
[269] Van der Meer, J. (2000). Operational control of internal transport.

lands.
[270] Van Donk, F. (1999). Environmental logistics in housebuilding and

public construction. In Van Goor, A., Flapper, S. D. P., and Clement,
[271] Van Goor, A., Flapper, S. D. P., and Clement, C., editors (1994-).
[272] Van Laarhoven, P., Berglund, M., and Peters, M. (2000).

Third-party logistics in europe - five years later. International Journal
of Physical Distribution and Logistics Management, 30(5):425–442.
[273] Van Nes, N. (2003). Replacement of Durables: Influencing product

lifetime through product design. PhD thesis, Erasmus University Rot-
terdam, Rotterdam, the Netherlands.
[274] Van Notten, R. A. P. (2000). Collection and processing for the re-
use of packaging glass. In Van Goor, A., Flapper, S. D. P., and Clement,
[275] Vis, I. F. A. (2002). Planning and control concepts for material han-
dling systems. PhD thesis, Erasmus University Rotterdam, Rotterdam,
the Netherlands.
[276] Vlachos, D. and Dekker, R. (2000). Return handling options

and order quantities for single period products. Report Econometric
Institute EI2000-29/A, Faculty of Economics, Erasmus University Rot-
terdam, The Netherlands.
312 Bibliography
[277] Vroom, J. A., van der Linden, A., and Kraal, A. (2001). Pack-
aging and reverse logistics at campina melkunie. In Van Goor, A., Flap-
per, S. D. P., and Clement, C., editors, Handboek of Reverse Logistics.
[278] VVAV (2003). Dutch waste processing association (vvav).
www.vvav.nl.
[279] Waimiti, H. (2001). Logistics in kenya. status, trends and develop-
ments. Master’s thesis, Erasmus University Rotterdam, Rotterdam, the
Netherlands. unpublished Master thesis.
[280] Walpole, E. E., Myers, R. H., and Myers, S. (1998). Probabil-
ity and statistics for engineers and scientists. Prentice Hall Inc., 6th
edition.
[281] Wäscher, G. (2002). Order picking: a survey of planning prob-
lems and methods. working paper 13/2002, Faculty of Economics and
Management, Otto von Guericke University Magdeburg, Magdeburg,
Germany.
[282] Wijshof, C. P. H. (1997). Reverse logistics, a case. In Van Goor,
A., Flapper, S. D. P., and Clement, C., editors, Handboek of Reverse
Logistics. Kluwer B. V., Deventer, The Netherlands. (in Dutch).
[283] Wood, S. (2001). Remote purchase environments: the influence of
return policy leniency on two-stage decision processes. Journal of Mar-
keting Research, 38(2):157–169.
[284] Yender, G. L. (1998). Battery recycling technology & collection
processes. In Proceedings IEEE International Symposium on Electronics
and the Environment, page 30 35, Oak Brook, IL, U.S.A.
[285] Yuan, X.-M. and Cheung, K. (1998). Modeling returns of mer-
chandise in an inventory system. OR Spektrum, 20(5):147–154.
[286] Zuidema, J. and de Koster, R. (2002a). Wehkamp - teaching note
-. Teaching case (exercise note), Rotterdam School of Management,
University Erasmus Rotterdam, Rotterdam, the Netherlands.
[287] Zuidema, J. and de Koster, R. (2002b). Wehkamp: rethinking re-
turn handling at DC Dedemsvaart. Teaching case, Rotterdam School of
Management, University Erasmus Rotterdam, Rotterdam, the Nether-
lands.
Samenvatting
In het verleden zorgden supply chains voor de logistieke fine-tuning van de
voorwaartse goederenstromen van grondstof tot de uiteindelijke klant. Van-
daag de dag echter vloeit een groeiende stroom producten terug in de keten.
Bedrijven moeten dus ook de retourlogistiek beheren. Toch denken managers
voornamelijk in ‘voorwaartse’ termen, waarbij de nadruk ligt op het leveren
van goederen aan de klant. De vraag is of bedrijven door kunnen gaan met
deze benadering of dat de focus meer moet worden gericht op retourlogistiek.
In andere woorden: Is het een zaak van ‘managing reverse logistics’ of van
‘reversing logistics management’ ?
Dit proefschrift draagt bij aan een beter begrip van retourlogistiek en wel
met betrekking tot de volgende hoofdlijnen:
1. retourlogistiek als (wetenschappelijk) onderzoeksgebied,
2. beslissingsondersteuning en praktijk van retourlogistiek.
De belangrijkste doelen van het verrichtte onderzoek zijn
• het structureren van retourlogistiek als onderzoeksgebied,
• een beter begrip kweken van de praktijk van retourlogistiek,
• het structureren en ondersteunen van beslissingen in retourlogistiek,
• speculeren over de toekomstige ontwikkeling van retourlogistiek als on-

derzoeksgebied.
In het bijzonder werden verscheidene onderzoeksprojecten opgezet om speci-

fieke doelstellingen te bereiken. Elk hoofdstuk geeft een verslag van één van
deze doelstellingen en wel als volgt.
Hoofdstuk 2: Het structuren van het onderzoeksgebied en het

kweken van globaal begrip van retourlogistiek.
Dit hoofdstuk kweekt een globaal begrip van retourlogistiek door middel van
een structurering van het onderzoeksgebied. In feite construeren we een
raamwerk, dat wil zeggen een conceptionele structuur, voor retourlogistiek:
we identificeren de elementaire ingrediënten (dimensies) van retourlogistiek,
we structureren deze en beschrijven hun onderlinge relatie.
314 Bibliography
Voor het raamwerk gebruiken we drie bronnen van input: selectieve liter-
atuur die het onderzoeksgebied structureert, een bespreking van case stud-
ies (Hoofdstuk 3), en de kennis van retourlogistiek die is geaccumuleerd
gedurende het gehele onderzoekstraject.
De dimensies en typologieën van het raamwerk voor retourlogistiek geven
niet alleen context voor retourlogistieke situaties. Ook hun combinatie bepaalt
voor een groot gedeelte het soort issues dat een rol speelt bij het imple-
menteren, beheersen, en controleren van zulke situaties. Dit wordt geïllustreerd
aan de hand van een verzameling typische situaties in de retourlogistiek.
Hoofdstuk 3: Het aangeven van een diversiteit van retourlogistieke

activiteiten en het helpen structureren van het framework.
Dit hoofdstuk analyseert meer dan 60 casussen uit de literatuur die handelen
over retourlogistieke aspecten. De studie is georganiseerd met betrekking tot
de volgende onderwerpen: 1) Netwerk structuren, 2) Relaties, 3) Voorraad-
beheer, 4) Planning & Control, en 5) Informatie technologie.
De analyse heeft een aantal relaties en issues aan het licht gebracht. Zo
lijken end-of-life retouren bijvoorbeeld sterk gecorreleerd te zijn met recy-
cling. Als producten of modules zouden zijn ontworpen voor remanufacturing
echter, dan zou remanufacturing waarschijnlijk attractiever zijn, zelfs aan het
einde van de levenscyclus. De belangrijkste driver voor recycling is wetgev-
ing. In feite lijkt de huidige wetgeving niet in staat te zijn om andere vor-
men van hergebruik te stimuleren. Echter, als wetgeving zich meer pro-actief
zou richten op product-ontwerp vraagstukken in plaats van zich reactief te
richten op hergebruik, dan zouden andere vormen van hergebruik dan recy-
cling waarschijnlijk meer worden gestimuleerd.
Voor elk van de bovengenoemde beslisgebieden zijn die dimensies geïdentificeerd
die overwegend belangrijk zijn en zijn richtlijnen gegeven voor verder onder-
zoek.
Hoofdstuk 4: Het aangeven van de state-of-the-art van kwanti-

tatieve modellen voor de afhandeling van retouren en het identi-
ficeren van mogelijkheden voor toekomstig onderzoek.
Dit hoofdstuk
• bouwt een beslissingsraamwerk voor de afhandeling van retouren, en
• identificeert modellen voor beslissingsondersteuning.

Bibliography 315
De analyse laat zien dat tot nog toe expliciete modellen voor beslissing-
sondersteuning ten behoeve van de afhandeling van retouren in het magaz-
ijn zijn veronachtzaamd. Voor een aantal onderzoeksgebieden geven we aan
hoe voorwaartse modellen kunnen worden uitgebreid, of hoe nieuwe modellen
zouden kunnen worden opgezet, om retouren mee te nemen. Verscheidene
onderzoekshiaten worden geïdentificeerd en gevat in een onderzoeksagenda.
Hoofdstuk 5: Het identificeren van de belangrijkste factoren voor

het al of niet combineren van voorwaartse en retour stromen bij de
afhandeling van retouren.
Deze exploratieve studie helpt mee aan de ontwikkeling van theorie voor de
afhandeling van retouren en geeft aanbevelingen voor de praktijk als ook
wetenschappelijk onderzoek. Met betrekking tot de factoren voor het al of niet
combineren van voorwaartse en retour stromen introduceren we drie veronder-
stellingen. Deze veronderstellingen geven relaties tussen de volgende zaken:
1. het beleveren van winkels & de transport fase beslissing,
2. het retour-volume & de ontvangst in het magazijn,
3. de markt voor retouren & de opslag beslissing.
Hoofdstuk 6: Het onderzoeken van de validiteit van gangbare aan-

names in de literatuur van voorraadbeheer voor retouren.
In dit hoofdstuk wordt een raamwerk opgezet voor het statistisch toetsen van
gangbare aannames in de literatuur van voorraadbeheer voor retouren. Met
de ontwikkelde methodologie is het mogelijk om zelfs voor kleine steekproeven
de gewenste toetsen uit te voeren. We gebruiken data sets van producten
van drie verschillende Europese bedrijven. We vinden producten waarvoor de
gangbare aannames lijken te gelden, maar ook enkele waarvoor ze niet gelden.
Op basis van de analyse geven we theoretische inzichten en implicaties voor
de praktijk. Bovendien geven we aan welk onderzoek nog nodig is.
Hoofdstuk 7: Het evalueren van de waarde van (mis-)informatie

m.b.t voorraadbeheer.
Dit hoofdstuk onderzoekt de waarde van (mis)informatie met betrekking tot
voorraadbeheer. We beschouwen een voorraadmodel met één product en één
echelon, waarbij elke periode een bestelling kan worden geplaatst om de voor-
raadpositie aan te vullen tot een van tevoren bepaald niveau. Hierbij wordt
316 Bibliography
rekening gehouden met de retouren die zullen binnenkomen gedurende de

levertijd. Om deze te schatten worden vier methodes gebruikt, elk met een
bepaalde informatiebehoefte.
Met behulp van analytische resultaten en simulatie identificeren we situ-
aties waarin de meest geavanceerde methode niet noodzakelijkerwijs de beste
prestaties levert. Dit is bijvoorbeeld het geval als systematisch de retour
intentie wordt onderschat of overschat.
De analyse geeft vervolgens aanleiding tot een aantal implicaties voor man-
agement, in het bijzonder informatie management.
Hoofdstuk 8: Het evalueren van de toekomst van retourlogistiek als

onderzoeksgebied.
Dit hoofdstuk presenteert een Delphi studie, ondernomen door een panel van
academici die werken op het gebied van retourlogistiek, met het doel om te
reflecteren op de toekomst van het vakgebied. De analyse maakt het mo-
gelijk om aanbevelingen te doen over onderzoek en onderwijs binnen de re-
tourlogistiek. Het blijkt dat het type onderzoek afhankelijk is van het type
onderzoeker (junior, senior, accidental). Ook geven we aan wat mogelijke
consequenties zijn van de verschillen tussen Europees onderzoek en dat in de
Verenigde Staten.
Het spreekt voor zich dat retourlogistiek ook een vorm van logistiek is, zodat
a priori lessen uit de traditionele logistiek ook kunnen worden doorgetrokken
naar retourlogistiek. Aan de andere kant denken traditionele managers voor-
namelijk in termen van voorwaartse stromen. In de toekomst zal dit niet
voldoende blijken. Aangezien retourlogistiek zich wereldwijd uitspreidt in
alle lagen van de supply chain, worden sommige actoren in de keten gedwon-
gen om producten terug te nemen, terwijl anderen dit pro-actief doen De
laatsten worden ofwel aangetrokken tot de waarde die gebruikte producten
nog vertegenwoordigen, ofwel aangetrokken door hun institutionele verant-
woordelijkheid. Hoe dan ook, retourlogistiek is een cruciale competentie in
moderne supply chains.
Logistiek kan niet vooruit zonder ‘terug’ te gaan!

Curriculum Vitae
Marisa P. de Brito studied Statistics and Operations Research at the Uni-
versity of Lisbon in Portugal. On her last year, she visited the Econometric
Institute at the Erasmus University Rotterdam (EUR) where she wrote her
thesis on the topic of Inventory Control. On graduating in 1997, she en-
rolled for a 1-year post-graduation on Operations Research at the University
of Lisbon in Portugal. During the same academic year, she was a full-time As-
sistant Professor of the department of Quantitative Methods at the Institute
of Business Administration, ISCTE in Lisbon, Portugal. In October 1998 she
got engaged with an International Master on Urban Management at EUR,
having received the diploma one year later.
In October 1999 she became an assistant researcher at the Economet-
ric Institute (EUR), being a PhD candidate associated with the Erasmus
Research Institute of Management (ERIM). Her research was nested in the
European Working Group for Reverse Logistics (RevLog), financially sup-
ported by the European Commission. Within the RevLog network, she vis-
ited the Business School INSEAD of Fontainebleau in France for a 2-month
period, at which she worked with Prof. Luk van Wassenhove. Furthermore,
during her PhD, she followed international courses associated with the Eu-
ropean Doctoral School on Knowledge and Management (EUDOKMA), the
European Institute for Advanced Studies in Management (EIASM) and the
European Logistics Association (ELA). Her PhD was financially supported by
the Portuguese Foundation for the Development of Science and Technology,
“Fundação Para a Ciência e a Tecnologia,” as from January 2000.
Erasmus Research Institute of Management
ERIM Ph.D. Series Resesearch in Management
1. J. Robert van der Meer, Operational Control of Internal Transport

Promotor(es): Prof.dr. M.B.M. de Koster, Prof.dr.ir. R. Dekker
Defended: September 28, 2000, ISBN: 90-5892-004-6
2. Moritz Fleischmann, Quantitative Models for Reverse Logistics

Promotor(es): Prof.dr.ir. J.A.E.E. van Nunen, Prof.dr.ir. R. Dekker
Defended: October 5, 2000, Lecture Notes in Economics and Mathematical
Systems, Volume 501, 2001, Springer Verlag, Berlin, ISBN: 3540-417-117
3. Dolores Romero Morales, Optimization Problems in Supply Chain

Management
Promotor(es): Prof.dr.ir. J.A.E.E. van Nunen, dr. H.E. Romeijn
Defended: October 12, 2000, ISBN: 90-9014-078-6
4. Kees Jan Roodbergen, Layout and Routing Methods for Warehouses

Promotor(es): Prof.dr. M.B.M. de Koster, Prof.dr.ir. J.A.E.E. van Nunen
Defended: May 10, 2001, ISBN: 90-5892-005-4
5. Rachel Campbell, Rethinking Risk in International Financial Markets

Promotor(es): Prof.dr. C.G. Koedijk
6. Yongping Chen, Labour flexibility in China’s companies: an empirical

study
Promotor(es): Prof.dr. A. Buitendam, Prof.dr. B. Krug
7. Pursey P.M.A.R. Heugens, Strategic Issues Management: Implications

for Corporate Performance
Promotor(es): Prof.dr.ing. F.A.J. van den Bosch, Prof.dr. C.B.M. van Riel
8. Roland F. Spekl, Beyond Generics; A closer look at Hybrid and

Hierarchical Governance
Promotor(es): Prof.dr. M.A. van Hoepen RA
322 Bibliography
9. Pauline Puvanasvari Ratnasingam, Interorganizational Trust in Business

to Business E-Commerce
Promotor(es): Prof.dr. K. Kumar, Prof.dr. H.G. van Dissel
Defended: November 22, 2001, ISBN: 90-5892-017-8
10. Michael M. Mol, Outsourcing, Supplier-relations and

Internationalisation: Global Source Strategy as a Chinese puzzle
Promotor(es): Prof.dr. R.J.M. van Tulder
Defended: December 13, 2001, ISBN: 90-5892-014-3
11. Matthijs J.J. Wolters, The Business of Modularity and the Modularity
of Business
Promotor(es): Prof.mr.dr. P.H.M. Vervest, Prof.dr.ir. H.W.G.M. van Heck
Defended: February 8, 2002, ISBN: 90-5892-020-8
12. J. van Oosterhout, The Quest for Legitimacy; On Authority and

Responsibility in Governance
Promotor(es): Prof.dr. T. van Willigenburg, Prof.mr. H.R. van Gunsteren
13. Otto R. Koppius, Information Architecture and Electronic Market

Performance
Promotor(es): Prof.dr. P.H.M. Vervest, Prof.dr.ir. H.W.G.M. van Heck
14. Iris F.A. Vis, Planning and Control Concepts for Material Handling
Systems
Promotor(es): Prof.dr. M.B.M. de Koster, Prof.dr.ir. R. Dekker
15. Jos Bijman, Essays on Agricultural Co-operatives; Governance

Structure in Fruit and Vegetable Chains
Promotor(es): Prof.dr. G.W.J. Hendrikse
Defended: June 13, 2002, ISBN: 90-5892-024-0
16. Linda H. Teunter, Analysis of Sales Promotion Effects on Household

Purchase Behavior
Promotor(es): Prof.dr.ir. B. Wierenga, Prof.dr. T. Kloek
17. Joost Loef, Incongruity between Ads and Consumer Expectations of

Advertising
Promotor(es): Prof.dr. W.F. van Raaij, Prof.dr. G. Antonides
18. Andrea Ganzaroli, Creating Trust between Local and Global Systems
Promotor(es): Prof.dr. K. Kumar, Prof.dr. R.M. Lee
Bibliography 323
19. Paul C. van Fenema, Coordination and Control of Globally Distributed

Software Projects
Promotor(es): Prof.dr. K. Kumar
20. Dominique J.E. Delporte–Vermeiren, Improving the flexibility and

profitability of ICT-enabled business networks: an assessment
method and tool.
Promotor(es): Prof.mr.dr. P.H.M. Vervest, Prof.dr.ir. H.W.G.M. van Heck
21. Raymond van Wijk, Organizing Knowledge in Internal Networks. A

Multilevel Study
Promotor(es): Prof.dr.ing. F.A.J. van den Bosch
22. Leon W.P. Peeters, Cyclic Railway Timetable Optimization

Promotor(es): Prof.dr. L.G. Kroon, Prof.dr.ir. J.A.E.E. van Nunen
23. Cyriel de Jong, Dealing with Derivatives: Studies on the role,

informational content and pricing of financial derivatives
Promotor(es): Prof.dr. C.G. Koedijk
24. Ton de Waal, Processing of Erroneous and Unsafe Data

Promotor(es): Prof.dr.ir. R. Dekker
25. Reggy Hooghiemstra, The Construction of Reality

Promotor(es): Prof.dr. L.G. van der Tas RA, Prof.dr. A.Th.H. Pruyn
Defended: September 25, 2003, ISBN: 90-5892-047-X
26. Marjolijn Dijksterhuis, Organizational dynamics of cognition and action

in the changing Dutch and U.S. banking industries
Promotor(es): Prof.dr. F.A.J. van den Bosch, Prof.dr. H.W. Volberda
27. Dennis Fok, Advanced Econometric Marketing Models

Promotor(es): Prof.dr. P.H.B.F. Franses
28. Martin Wielemaker, Managing Initiatives. A Synthesis of the

Conditioning and Knowledge-Creating View
Promotor(es): Prof.dr. H.W. Volberda, Prof. dr. C.W.F. Baden-Fuller
Defended: November 16, 2003, ISBN: 90-5892-050-X
324 Bibliography
29. Douglas A.F. Berghe, Working Across Borders: Multinational

Enterprises and the Internationalization of Employment
Promotor(es): Prof.dr. R.J.M. van Tulder, Prof. dr. E.J.J. Schenk
30. Peter R. Miltenburg, Effects of modular sourcing on manufacturing

flexibility in the automotive industry. A study among German OEMs
Promotor(es): Prof.dr. J. Paauwe, Prof. dr. H.R. Commandeur
31. Gjalt Graag, Tractable Morality. Customer discourses of bankers,

veterinarians and charity workers
Promotor(es): Prof.dr. F. Leijnse, Prof. dr. T. van Willigenburg
32. Victor A. Gilsing, Exploration, Exploitation and Co-evolution in

Innovation Networks
Promotor(es): Prof.dr. B. Nooteboom, Prof.dr. J.P.M. Groenewegen
33. Bert Filer, Strategic Renewal of European Financial Incumbents

Promotor(es): Prof.dr. F. van den Bosch, Prof.dr. H. Volberda
34. Peter de Langen, The Performance of Seaport Clusters

Promotor(es): Prof.dr. B. Nooteboom, Prof.dr. H.W.H. Welters
Defended: January 20, 2004, ISBN: 90-5892-056-9
Managing Reverse Logistics or Reversing Logistics
Management?
Traditionally, supply chains fine-tune the forward logistics from raw
material to the end customer. Today an increasing flow of products is
going back in the chain, so companies have to manage reverse logistics
as well. This thesis contributes to a better understanding of reverse
logistics. The thesis brings insights on reverse logistics decision-making
and it lays down theoretical principles for reverse logistics as a research
field. In particular, it puts together a framework for reverse logistics
identifying the elementary dimensions, providing typologies, and
structuring their interrelations. With respect to aiding decision-making,
this thesis comprises return handling and inventory management.
Regarding the first, the focus is on identifying critical factors that
determine if forward and reverse flows should be combined, or if should
be dealt with separately. Regarding the second, the main research issue
is the value of information. One of the findings is that more informed
methods do not necessarily lead to the best performance in case of
misinformation. Furthermore, this thesis proposes a reflection on the
future development of the field. Through a Delphi study with an inter-
national panel of academics working on the area, recommendations are
made concerning both research and pedagogy.
This thesis also poses the following question: how should we deal with
reverse logistics? Is it a matter of simply managing reverse logistics or of
The message is: logistics cannot go forward without reverse thinking!
ERIM
The Erasmus Research Institute of Management (ERIM) is the Research
School (Onderzoekschool) in the field of management of the
Erasmus University Rotterdam. The founding participants of ERIM are
the Rotterdam School of Management and the Rotterdam School of
Economics. ERIM was founded in 1999 and is officially accredited by
the Royal Netherlands Academy of Arts and Sciences (KNAW). The
research undertaken by ERIM is focussed on the management of the
firm in its environment, its intra- and inter-firm relations, and its
business processes in their interdependent connections. The objective
of ERIM is to carry out first rate research in management, and to
offer an advanced graduate program in Research in Management.
Within ERIM, over two hundred senior researchers and Ph.D. candi-
dates are active in the different research programs. From a variety of
academic backgrounds and expertises, the ERIM community is united
in striving for excellence and working at the forefront of creating
new business knowledge.
www.erim.eur.nl ISBN 90-5892-058-5
View publication stats

Managing Reverse Logistics or Reversing Logistics

Uploaded by

Copyright:

Available Formats

Managing Reverse Logistics or Reversing Logistics

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Managing Reverse Logistics or Reversing Logistics

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Managing reverse logistics or reversing logistics management? /

Thesis · January 2004

Humanitarian Logistics View project

Entrepreneurship in Leisure, Tourism and Hospitality View project

The user has requested enhancement of the downloaded file.

Cover photo: ‘Reverse’

Beheersing van retourlogistiek

to obtain the degree of Doctor from

The public defense shall be held on

Other members: Prof.dr.ir. J.A.E.E. van Nunen

Finishing a PhD dissertation is certainly a moment to cheer. My PhD tra-

2 A Framework for Reverse Logistics 39

3 Reverse Logistics: a review of case studies 79

4 Return Handling: decision-making and quantitative support123

5 Return Handling: an exploratory study 147

6 Inventory Management: a critical analysis of assumptions 177

7 Inventory Management with product returns: the value of

7.6 Cost performance . . . . . . . . . . . . . . . . . . . . . . . . . . 219

8 Reverse Logistics research: what does the future bring? 245

9 Concluding remarks 271

Samenvatting (Summary in Dutch) 313

Curriculum Vitae 319

1.1 The body of the thesis . . . . . . . . . . . . . . . . . . . . . . . 30

2.1 Building the framework . . . . . . . . . . . . . . . . . . . . . . 42

3.1 IT support for Reverse Logistics in different phases of the life-

4.1 Return handling in the warehouse. . . . . . . . . . . . . . . . . 124

5.1 Reverse flows in the supply chain. . . . . . . . . . . . . . . . . . 148

5.4 Decision-scheme on general long- and medium-term warehous-

6.1 Statistical analysis scheme to test if a process is an HCPP. . . . 185

7.1 Relation between data and decision-making. . . . . . . . . . . . 207

7.9 Performance of Method A both under perfect correlation and

8.1 Representation of the sheet of paper given to the participants. . 250

1.1 Recycling quotas in Europe (data collected in 2003 with * being

3.1 Key words used for the search of case studies . . . . . . . . . . 81

4.1 A framework of decisions for return handling. . . . . . . . . . . 129

5.4 The return process at the nine retailers. . . . . . . . . . . . . . 160

6.1 The assumptions of the literature on inventory models with

7.1 Notation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210

8.1 Employed arguments during NGT. . . . . . . . . . . . . . . . . 250

“A long trip starts with the first step”

1.1 Reverse Logistics: definition and scope

“In the sweat of your face you shall eat bread

At the end of the nineties, Rogers and Tibben-Lembke (1999) portrayed

1.1.2 Delineation and scope

return logistics and retro logistics or reverse distribution, sometimes referring

1.1.3 The European dimension

EU Directive France Germany The Netherlands

1.2 Literature on Reverse Logistics

give an idea of opportunities for multidisciplinary reverse logistics studies, we

1.2.1 Books on reverse logistics

1.2.2 PhD theses on reverse logistics

1.2.3 Review articles on reverse logistics

and the ecological aspects in the supply chain.

1.2.4 Some literature on areas related to reverse logistics

Beukering (2001) investigates the economic and the environmental impor-

1.3 About the thesis

• Part I on the theoretical development of the field;