OPTIMIZATION METHODS AND BULLWHIP EFFECT

MARTINA KUNCOV
UNIVERSITY OF ECONOMICS, PRAGUE, CZECH REPUBLIK

1. Introduction
1

In this article I would like to make a connection between methods of the operational research
and bullwhip effect, which is usually mentioned when talking about supply chain management. Supply
chain management (SCM) is nowadays seen as one of the most important areas that should be
developed. Supply chain itself can be characterized as a net that consists of suppliers, manufacturing
centres, warehouses, distribution centres, and retail outlets, as well as raw materials, work-in-process
inventory, and finished products that flow between the facilities. The main effort is dedicated to the
interconnection of all subjects in the supply chain and to the optimization of all the flows that exist
among these subjects. But because of the fact that firms are used to compete and not to cooperate, they
usually dont want to share any information about their products, inventory, costs or profit. So it is not
easy to manage the whole supply chain the way that is acceptable for everyone. Ive tried to describe
some examples connected with inventory optimisation and show, how the effort of
minimization of the enterprise inventory cost can cause the bullwhip effect.
2. Bullwhip Effect
Whenever we read any book or an article about the supply chain or the supply chain
management, we probably find there something about the bullwhip effect (also called demand
amplification effect or whiplash). Forester was the first man who described this effect but the
experts of Procter & Gamble gave it its name (after the way the amplitude of a whip increases
down its length) and also its publicity. Its well known, that during the monitoring of the
customers demand and retailers and distributors orders of the diaper product Pampers they
found out the amplification of order variability, though the demand was almost fixed.
Recently the interest in the exploration and measurement of the bullwhip effect has grown up
because of the fact that the amplification of orders influences the distributors and the
manufacturers costs, inventory, reliability and other important business processes. By reason
that nowadays the main effort is dedicated to the coordination and communication among
customers, suppliers, distributors and manufacturers (it means it is dedicated to the
development of the supply chain management), some specialists of this branch have started to
look for the reasons for the bullwhip effect, for the methods of its measurement and also for
144
the methods of its elimination.
2.1 Four Causes of the Bullwhip Effect
Demand forecast. Its not assumed that the retailer knows the exact form of the
customer demand process. Instead, he uses historical data and some forecasting
techniques to estimate the demand. The supplier doesnt know the retailers data and
so he supposes the retailers order to be the real demand. Due to this, the forecast
could be very different and thats why the orders can vary.
Order batching. In most cases (see below) the cost (or inventory) policy is the main
why for ordering in batches. But due to this, the next link of the supply chain has to
have higher inventory to avoid the depletion of inventory.
Price fluctuation. Customers are driven to buy in larger quantities by attractive offers
on quantity discounts or price discounts. If their behaviour is rational, they buy more
when the price is down or less when the price is up. However this strategy doesnt
reflect their true needs, and so this may give birth to the bullwhip effect.
Rationing and shortage gaming. This cause might be similar to the price fluctuation.
It occurs when demand exceeds supply or when the customers think it may happen.
Then they start to exaggerate their real needs to be sure that the existing demand will
be satisfied. The demand amplification effect will grow up even further if customers
are allowed to cancel their orders when their real demand is fulfilled.
Its possible to consider more reasons, for example the information distortion, lead
times or the maximization of the profit (or the minimization of the costs) that can influence
each of the causes mentioned above.
Every retailer usually knows a lot of information about customers, demand, prices,
discounts, inventory, etc. but it doesnt want to share this information with another company
even though it might be its supplier. Then the supplier and the manufacturer dont know much
about the real situation on the market, and therefore they can hardly follow the customers
needs. This effect is called the information distortion.
Lead times arent conventionally mentioned separately as the main cause of the
bullwhip effect, although this factor is included in the bullwhip effect formula. Its clear that
the longer the lead times are, the larger the demand and the safety stock must be (to avoid the
inventory shortage), and thats why the larger the demand amplification could be.

1
This article was supported by Grant No. 402/01/0771 from the Grant Agency of the Czech Republic
145
The most often aim of every company is to maximize their profit or to minimize their
costs irrespective of their suppliers or customers. If they try to fulfil this stated goal, they have
to realize some of the actions mentioned above (discounts, order batching, etc.) and by this
means they contribute to the demand amplification effect.
2.2 Problems With Measuring of the Bullwhip Effect
According to other similar articles, the bullwhip effect is usually measured as the ratio
of variances Var(q*) / Var(Q), where Var(q*) represents variance of (optimal) order and
Var(Q) is variance of demand faced by the enterprise. Although this formula is used very
often, Ive found out that it is dependent on the type of data aggregation.
I have obtained the real data about the intensity of the sales and orders of one type of
summer tyres
2
from the tyre repair and service shop. Chart 1 show the real demand, order and
inventory per month.
0
10
20
30
40
50
60
70
January February March April May June July August September October November December
demand order inventory

Chart 1 Monthly demand, order and inventory

Demand is between 0 and 47 units (tyres) per month, but the firm orders every month
(as you can see later, sometimes even every week or twice a week). The variance of the sales
is 241 and the variance of the orders is 210, so there is no demand amplification effect
because the quotient of these variances is 0.87.
As I know, the customers dont buy the tyres once a month but conventionally they
want to buy it every week or every day and so does the retailer. When you look at the Chart
2, you can see that the inventory level is much higher than the demand level (they can afford
it because of the fact that the inventory holding costs are very small). Although the orders are
close to the demand, the bullwhip effect is suddenly 1.34.


2
Brilantes 160/70/R13 OR60 usually used for the Czech cars Skoda Fabia and Skoda Felicia
146
-10
0
10
20
30
40
50
60
70
w
e
e
k 2 4 6 8
1
0
1
2
1
4
1
6
1
8
2
0
2
2
2
4
2
6
2
8
3
0
3
2
3
4
3
6
3
8
4
0
4
2
4
4
4
6
4
8
5
0
5
2
demand order inventory

Chart 2 Demand, order and inventory per week

Finally Ive investigated the daily demand and order (and then also the same but
without weekends because the shop is closed in those days). When Ive taken into account all
work days of the year 2001, the bullwhip effect has been 3.27. The question is whether the
days with zero demand and zero order should be involved or not. If they arent involved, the
bullwhip effect is surprisingly 5.84. As you can see in Chart 3, the bullwhip effect also
depends on the period of data aggregation.
Another possibility, how to count this effect, is the ratio of coefficients of variation of
order and demand, where coefficient of variation is equal to standard deviation divided by
average demand (order). This formula is much worse than the previous one, especially when
the standard deviations of demand and order are nearly equal, because the ratio of average
orders (demands) doesnt tell us nearly anything about the oscillations.
Because of these difficulties with measuring, Ive tried to count the mean costs of
every member of the supply chain (in the next example) and define the bullwhip effect (with
regard to costs) as a percentage growth of costs.
Bullwhip Effect
0
1
2
3
4
5
6
7
monthly weekly daily daily without weekends daily without zeroes

Chart 3 The bullwhip effect when using various periods for data aggregation

147
3. Stochastic Models of Inventory and the Bullwhip Effect
It is supposed that demand is a random variable (usually a discrete random variable)
that can be approximated with a normal distribution with known mean and standard deviation.
If we want to calculate the optimal order quantity (to minimize the inventory costs), we can
use EOQ formula as follows (I assume the inventory shortage doesnt occur, so this model
doesnt reflect the inventory shortage costs):
EOQ formula:
1
2 Q *
c
c 2
q

=
where: q* = the optimal order quantity (EOQ) for minimization of the inventory costs

Q
= the mean demand in units per one period
c
1
= the inventory carrying costs (per one unit per one period)
c
2
= the costs per order
Since we dont know the actual demand, it is necessary to build up a safety stock (w)
that should compensate the variations of demand. The safety stock level depends on the
desired percents product availability ( = the probability that the inventory shortage wont
occur). This desired service level is a function of the normal loss curve, which provides the
area in a right tail of a normal distribution. If the factor that corresponds with is z and we
know the lead times (which are constant), then we can calculate the safety stock level and also
the reorder point (the inventory level showing the necessity to order) as:

L * * z w
Q
= w * L s
Q
+ =
where: w = safety stock in units, z = the z factor that corresponds with service level
L = lead time, s = reorder point in units,
Q
= the mean demand

Q
= the standard deviation of demand
Now lets say that the customers demand oscillates randomly between 20 and 30 units
per month (mean is 25 units and standard deviation is 1,67). The fixed costs are 1,5 crowns
per month and variable costs are 120 crowns (for all members of the supply chain), the lead
times are 2 weeks (0.43478 per month) and the service level is 99%. Then the optimal
characteristics for the retailer are in the Table 1. In this case the bullwhip effect is 316
because the oscillations of demand are very small compared to the oscillations of the orders.
Since the demand is random, the supplier cant find out when the retailer really orders
(which week or day). Because the retailers order is now its demand, which the supplier faces,
148
and it is not normally distributed, lets suppose that the supplier has 3 possible strategies (A,
B, C). All characteristics are in Table 1.

Table 1 Main characteristics of retailer and supplier
Retailer Supplier A Supplier B Supplier C

Q
(units) 25 21 63 31,5

Q
(units) 1,67 30 0,67 10,5
q* (units) 63,24 57,96 100,6 71
t* (months) 2,53 2,76 1,6 (3-5) 2,25
w (units) 3 47,59 1,06 16,66
s (units) 14,63 57,56 32 32
S (units) 79 116 133 103

N (crowns/month)
99,37 158,33 152,19 131,48

As you can see in all examples shown above, when any subject of the supply chain
uses the EOQ method for order quantity calculation (to minimize its inventory costs), it
always orders in batches after some period and so it should enhance the bullwhip effect. But
when Ive tried to simulate this small supply chain (customer-retailer-supplier) and than count
the effect, Ive found out that the point of view (or the measuring method) is the critical
factor. When you look into the table above, you can see that the lower the safety stocks are,
the higher the optimal order must be, and vice versa. But the most used formula for counting
the demand amplification effect doesnt care about the safety stocks although they are very
important when you are interested in costs and inventory levels. So the bullwhip effect
counted as the ratio of variances differs from the ratio of mean costs (see Table 2).

Table 2 Comparison of the classical BE and the growth of mean costs
Retailer vers.
Supplier A
Retailer vers.
Supplier B
Retailer vers.
Supplier C
Ratio of variances (BE) 0,959 2,168 1,07
Ratio of mean costs 1,59 1,53 1,32

In the first case there is no bullwhip effect, because the supplier orders nearly with the
same frequency and nearly the same amount as the retailer does, but it doesnt reflect that the
supplier must hold high level of safety stock which increase its inventory holding and
149
carrying costs. So it is possible to say that the bullwhip effect is hidden in inventory level or
in safety stocks.
4. Conclusion
Supply chain management (SCM) is nowadays seen as one of the most important areas
that should be developed because the present trend aims at cooperation and managing the
whole chain. The bullwhip effect is one of the main problems inside any supply chain because
it has bad impact on those enterprises, which are not closed to the end customer. Ive tried to
show how it may come into existence, for example when using the EOQ policy (to minimize
the inventory holding and carrying costs), which leads to ordering in batches, and how it is
possible to measure it. Ive found out that the usual method for its measuring shows only the
growth of amplification of orders (if occurs), but it doesnt tell us anything about the impact
on the inventory level or inventory costs. If each member of the supply chain wants to
minimize its inventory costs, the bullwhip effect always appears. As Ive described above,
sometimes it doesnt appear as the demand amplification effect, but it is hidden in the growth
of the inventory level and the inventory costs. Unfortunately, it is nearly impossible to avoid
it, because any company must count with uncertainty.
References
1. Bowersox D.J., Closs D.J.: Logistical Management The Integrated Supply Chain
Process. Singapore, McGraw-Hill 1996
2. Chen F., Drezner Z., Ryan J.K., Simchi-Levi D.: Quantifying the Bullwhip Effect in a
Simple Supply Chain: The Impact of Forecasting, Lead Times, and Information.
Management Science 2000, http://proquest.umi.com/
3. Fiala P., Pelzbauerov V.: Analysis of the Bullwhip Effect in Supply Chains. Proceedings
of the 18
th
International Conference on Mathematical Methods in Economics, The
University of Economics, Prague 2000
4. Fransoo J.C., Wouters M.J.F.: Measuring the Bullwhip Effect in the Supply Chain. Supply
Chain Management, Bradford 2002, http://proquest.umi.com/
5. Hausman W.H.: Information Distortion and Collaboration: The Bullwhip Effect,
http://www.supplychain.com/
6. Hausman W.H.: Introduction to Supply Chain Management,
http://www.supplychainonline.com/
7. Lee H.L., Padmanabhan V., Whang S.: The Bullwhip Effect in Supply Chains. Pensylvania
State University 2001, Sloan Management Review 38 (3), p.93-102
150
151
8. Simchi-Levi D., Kaminsky P., Simchi-Levi E.: Designing and Managing the Supply
Chain. USA, McGraw-Hill 2000
9. Taylor D.H.: Measurement and Analysis of Demand Amplification Across the Supply
Chain. International Journal of Logistics Management, Ponte Vedra Beach 1999,
http://proquest.umi.com/


Ing.Martina Kuncov
Department of Econometrics, University of Economics, Winston Churchill sq. 3, 130 67
Prague 3, Czech republic, tel. +420 224 095 449, kuncovam@vse.cz