Engineering Management

Research Methods for Engineering Projects

Brunel University London

Understanding Customers

Without this there is little chance that a product or service will succeed in the market. When
laptops were first developed the major computer manufacturers viewed it as unlikely that
many consumers would purchase them – they believed that customers would be deterred by
the higher prices resulting from the increased engineering costs of miniaturisation. The
majority of computing firms did not manufacture laptops in large numbers based on this
belief.

However, Dell Computers operate a “direct business model” which means that customers
contact Dell directly (by phone or online etc..) to place an order when purchasing a computer.
This gave Dell as a company an advantage –their regular contact with customers led them to
realise that customers really valued the space saving and convenience of owning a laptop.
This insight into their customers enabled them to manufacture large numbers of laptops and
meet the unexpected market demand when laptops were first introduced to the market.

1
Research Approaches

You may be wondering how do researchers go about understanding whether or not customers
are likely to be interested in a new product or process or technology innovation?

Engineering firms and manufacturers put a great deal of effort into understanding customers
– to assist this we often ‘segment’ customers which means that we break the market into
customer groups with similar needs and characteristics. For example, the purchasing decision
for a home computer will be quite different for families versus a business person.

There are two main research approaches which we now introduce.

Deductive approach: This approach entails a research topic where a wealth of literature is
available, and theory can be formulated and hypothesis made to test the theory.

Inductive approach: involves collecting data first and developing theory from data analysed,
this is most appropriate on a new topic where there is little literature.

Although both approaches can be used on a research project, the basic differences of the two
approaches are summarised here:

Deductive approach Inductive approach

• Scientific principles • Gaining understanding of meanings
that people attach to events
• Moving from theory to data
• In-depth knowledge of topic
• Collection of quantitative data
• Collection of qualitative data
• Highly structured approach
• More flexible structure to change
research emphasis

These two research approaches may be viewed as complementary as shown by the ‘Research
Wheel’.

2
The Research Wheel

Research types can be classified into either exploratory; descriptive or explanatory studies,
although a research project can employ more than one of these. For an innovative product a
researcher might use an exploratory technique to of find what is happening, seek new insight
or find out the nature of a problem. This ‘inductive’ research can then be followed by
‘deductive’ research as shown by the ‘Research Wheel’ where research often starts with
observations (shown at the bottom) and may lead to a further research phase where
hypothesis are developed (shown at the top) and then tested in the field:

Conceptual
Framework
(theory, literature)
Data
Proposition
analysis Research
questions/hypothesis
Empirical
Inductive Deductive
observation
Data
collection
Source: Rudestam and Newton, 2001, p5

Research into product or process innovation typically involves Field Research where the
Researcher observes people’s activities or reactions in the appropriate context. The data
collected then needs to be analysed and patterns looked for. It can be helpful to use a software
package to analyse the data from the Field Research. Various software packages exist which
can be aid with this task including SPSS (originally, Statistical Package for the Social
Sciences) now owned by IBM which is often used by post graduate students and researchers
with Quantitative data (numerical data such as answers to survey questions where participants
give numeric answers).

Decisions on which research approach to use are made as part of the research design. The
research design according to “Research methods for business students” by Saunders, M.,
Lewis, P. and Thornhill, A. (2009) from Prentice Hall is:
“ a plan of how a researcher intend to go about answering his/her research
question(s), evolving on how data will be collected, constraints, like access to data,
location, time and money, also why research is carried out in a particular
organization, department and particular set of people in the department or
organization”.

3
Inductive research starts with Field Research (which may be Qualitative in nature) as follows:

Inductive Procedure
for Research Source: Hart, 1998, p192

Researcher gathers information and data

Questions are asked

about the phenomenon

The data is classified and

placed into categories

Patterns are looked for in the data and

potential theories are proposed

Theories are tested and developed and patterns

compared with other patterns and theories

The module materials (videos and accompanying materials) shall introduce some of the data
collection and analysis techniques used for research studies involving qualitative data.

In contrast, Deductive research involves establishing a ‘hypothesis’ which is a ‘tentative

explanation for an event’ and the causal factors based on existing knowledge in the literature:

Hypothesis: Establishing a
relation between variables

EXAMPLE
Variable Variable

Variable Event

Variable

4
The basic steps for this type of research are as follows:

Deductive procedure
for research
Source: Hart, 1998, p192

The researcher tests a theory

Hypothesis or research questions

are derived from the theory

Concepts and variables are

operationalized

An instrument is used to measure the

variables in the theory

Verification of the hypothesis

This type of research often involves testing Hypothesis (or Research Questions) on a large
scale (with a large number of consumers). Surveys are often used for this as they are
relatively easy to distribute – particularly online surveys many of which are available for free.
A researcher can use surveys in the form of questionnaires (after a literature review) to
collect the quantitative data.

According to "Surveys in social research" by De Vaus, 2002 from Routledge, a questionnaire

is a technique of data collection in which persons are asked to respond to the same set of
questions in a predetermined order, it therefore includes structured interviews and telephone
questionnaires. Although questionnaires are usually not very good for exploratory or other
research that requires large number of open ended question, instead they work best with
standardized questions that you can be confident will be interpreted same way by all
respondents (Saunders et al, 2009, p.362).

Think Point

What are your views on ‘Smart Homes’ – are you comfortable with using your mobile
handset to access your home devices and make changes (e.g. adjusting heating level)?

5
Comment

As consumers we have differing attitudes towards adopting new technologies and products –
some of us are keen to try out new technologies, others of us are less willing!

People can be categorised by their willingness to adopt new innovations – this is known as
Adoption Categories as follows:

“Innovative Consumers”

People who are keen to try out the newest product (even though the teething problems
may not have yet been resolved) are known as “Innovative Consumers”,

“Early Adopters”

People who are keen to try new products but prefer that they have been “proven” in
the market are known as the “Early Adopters”.

“Early Majority” or “Late Majority”

The majority of the population are more cautious in their attitude towards adopting
new innovations and are known as the “Early Majority” or “Late Majority”.

“Laggards”

The people who are most reluctant and are very late to adopt new technology are
known as “Laggards” as shown by the following figure.

The horizontal timeline below shows the profile of consumers adopting an innovation:

Proportion
of Adopters

Time

6
 Source: Jeffrey Baumgartner (2009)
www.noweurope.com

As shown above the design and development of new products or processes involves various
activities. The starting point is shown at the top – as the “Innovation Challenge”.

We are suggesting that development of the firm’s products or processes should be made
according to the principles of “User-Centered Design” whereby the perspective of the
customer (or end user) is key to design. Towards this research into engineering management
processes and systems usually involves Field Research where the Researcher observes
people’s activities and technology usage in the appropriate context – such as at their typical
place of work. Alternatively, we may use “Prototypes” as a means of getting user feedback.

Field research involves identifying a suitable approach to holding data gathering sessions.
Data gathering can be approached in various ways - either remotely such as via surveys or
online forums - or in person such as interviews or focus groups (where the researcher acts as
a moderator to facilitate discussion).

Preparing questions in advance is usually advisable, however, when discussing

devices/systems and their usage with people one needs to be flexible as the direction of the
discussion may vary. As a researcher you need to be prepared to capture the comments made
and also to “steer” the discussion back to your research interests.

Aids can be used to make a session more “structured” – such as the use of “index cards”
which may prompt people in their thinking as they review the cards on which key items are
shown. Repertory Grid is another structured approach that is useful for small discussion
groups where items raised are recorded as “Constructs” for the “Events” being discussed.

7
Customer Value

During our studies we emphasise the importance of taking the perspective of the customer (or
end user) – taking their viewpoint in terms of their “Needs and Wants” and how a proposed
system or solution may address this. The Value Proposition Canvas is a good visual tool for
capturing customer perspectives and insights as shown:

PRODUCT / SERVICE CUSTOMER

BENEFITS EXPERIENCE

CONCERNS
WANTS

FEATURES

NEEDS

The right side of the canvas captures the customer’s view point and their “Needs and Wants”
as well as areas of “Concern” - or “Fears”. The left side of the canvas is where we identify
users’ anticipated “Experience” of the product/service along with the Features that it will
provide - as well as the Benefits for the end User. We distinguish between Features (the
product attributes) and Benefits as part of understanding what provides “Customer Value”.
Benefits refer to the value that having or using the product/service brings to a person – such
as the convenience of having a car (to assist with food shopping; commuting to work etc).

Using a flowchart or other visuals to identify proposed solutions and customer experiences
allows Developers to determine which design concepts provide a better User experience. A
flow chart can help us to plan and develop new processes where we can visualise the “flow”
from a customer’s perspective – this is known as a “Customer Journey Map”.

A high technology firm and its designers and developers are better able to create a
“distinctive value proposition” by using tools such as “Customer Journey Maps” and a
“Value Proposition Canvas”.

With a target customer (group) in mind these tools provide a means of visualising products
and services from their perspective and to identify the system (or solution) features and
benefits that will provide value.

Use of these types of visual tools and collaborative teams are more likely to lead to successful
innovation as the focus is firmly on the customer (or end user) and developing products or
systems that satisfies users and provides customer value.

8
Market Segments

“Market segmentation” refers to dividing the potential market into similar (homogenous)
customer groups with similar needs and characteristics. For example, families may have
similar needs when it comes to buying cars or booking their holidays.

As engineering managers we may not be directly involved in the activities related to

understanding customers, however, we may well attend meetings where marketing specialists
(or sales account staff) discuss this and so it’s important that we have some appreciation of
the approaches they use.

We distinguish between firms selling to consumer markets (known as B2C) whose customers
are individuals or households versus firms selling to other businesses (known as B2B) whose
customers are organisations - such as companies that may themselves be selling to industrial
or consumer markets.

Examples of B2C segmentation bases:

Demographic: e.g. age, sex, occupation, income, family etc

Geographic: e.g. area type, neighbourhood type, region

Behavioural: e.g. user/non-user, frequency, loyalty

Psychographic: e.g. lifestyle, fashion orientation, benefits sought.

ACTIVITY

Consider a product that you purchased recently as a consumer –what might be the basis for
segmentation for this type of product?

9
Firms selling to other businesses whose customers are companies (rather than individuals),
may segment their market depending on the type of business, the industry sector and the size
of the company.

For a firm to have ongoing (repeat) business they need to not only produce products that
customers are interested in purchasing but also other perform in other aspects. We can use
surveys to ask “What do our customers value?”

We can then compare this to a firm’s current product and service offering and how well they
currently perform. We can examine how well the firm is performing in each of these areas
using a radar (or ‘spider’) map as shown:

P roduc t offering
8
6
Tec hnic al S upport 4 P ric e
2
C us tomer
0 E xpec tations
F irm's P erformanc e

R es pons ivenes s Quality

Delivery

Customer viewpoints are best assessed and presented for each key market segment for an
engineering or manufacturing firm. Therefore establishing the basis for segmentation is
required as early as possible in the research process.

Surveys can be a useful way for engineering and manufacturing firms to get comments and
feedback from customers.

Note that we need to consider the survey results in terms of:

• Reliability – are the results obtained reliable? If repeated will they be consistent?

• Validity – are the results valid? Did we interview people with knowledge?

• Representative – do the results from a limited data sample reflect the views of our entire
customer base for that product line? Is our data size/sample sufficient?

10
Normality

Surveys have the advantage in that they can provide a high level of responses which represent
a firm’s diverse customer base encompassing people from different geographic regions and
other demographics. Conducting quantitative research involves collecting and analysing field
research data that are numbers based. For example, we may run a survey where we ask
participants to indicate “How Strongly” they feel about an item or “How Important” it is to
them and ask them to Score a response from 1 (Low) to 7 (High) – known as a Likert Scale.

Once we have collected survey responses with quantitative data then we need to analyse it –
to look for patterns. However, many of these tests require that the data collected for each of
the variables should be ‘normally’ distributed such that they have a ‘bell-shaped curve’
around the sample mean as shown:

Frequency of
occurrence

The distance of a data point from the mean squared is known as its “deviation”. The square
root of the average value for the entire data sample is known as the Standard Deviation (SD)
and represents how widely spread the curve is. This is often denoted by ‘sigma’ (σ). Study
the chart – you should see almost all data is within +/- 3 σ of the centre point (the mean).

We can establish the likelihood (the probability) of a data point occurring by knowing its
distance from the mean expressed as a number of Standard Deviations - for which we use the
“z score”. We can use standard “look up” tables (such as ones available online) or the Excel
function (NORM.S.DIST) or other tools to find the probability once we have the z-score.

Where, z-score = point of interest (x) – sample mean (m)

sample Standard Deviation (SD)

11
Example:

We can compare data for two groups (for example, males and females) in terms of the sample
curve shapes - the mean values and Standard Deviations. Shown below are the results of a
sample of people who are regular gym users. The results are shown firstly for the ladies and
secondly for the guys:

LADIES USAGE
Mean of half an hour
Frequency

Time
Mean (m)

GUYS USAGE
Mean of 1 hour

Mean (m)

You should notice that the sample average (the Mean) for the ladies is lower at half an hour
whereas the sample average (the Mean) for the guys is higher at an hour. The other difference
is the shape of the curve – the ladies is more focused around the “peak” (the Mean) whereas
the guys is more “spread” out. How do you think this will be reflected in their Standard
Deviations?

The curve for the guys is more widely spread which will lead to a higher Standard Deviation
value - showing that there is more variation in gym usage times compared to the ladies where
their gym usage times are much closer to the Mean (and so will have a lower Standard
Deviation value) .

We can also examine the two groups statistically using a ‘t-test’. For three or more groups we
may use a similar test – the “Analysis of Variance” (ANOVA) test which determines whether
there are any statistically significant differences between the means of three or more
independent (unrelated) groups.

For example, we may examine the survey results amongst different age groups to see whether
there are similarities or differences in the patterns for each age group (or segment).

12
ACTIVITY

Camera usage underwater (minutes):

Group 1 Group 2 Group 3

17.3 41.5 115.6
20.1 43.2 132.3
19.6 38.6 109.7
17.4 46.1 134.7
25.1 53.7 110.1

Examine the data provided in the table: How “diverse” are the groups?

Calculate the Arithmetic mean (m):

∑ data points
number of data points (n)

Find the deviation for each data point: (x-m) & square it
Standard Deviation: the square root of the average deviation for sample

To assist you the first group are calculated for you:

Measure Distance (d) Deviation

(x) using x-m using (x-m)^2
17.3 -2.600 6.76
20.1 0.200 0.04
19.6 -0.300 0.09
17.4 -2.500 6.25
25.1 5.200 27.04
∑ 99.50 40.18 TOTAL

Mean (m) = ∑ 99.50 / number of data points (n) of 5 = 19.900 minutes

Average deviations is Total deviations 40.18 / number of data points (n) of 5 = 8.036

Standard Deviation: the square root of the average for sample = √8.036 = 2.835 minutes

Hint: Mean of 44.620 mins and Standard Deviation (SD) or σ = 5.150 mins for Group 2
Mean of 120.480 mins and Standard Deviation (SD) or σ = 10.860 mins for Group 3

The “spread” of each data sample can be represented by the Standard Deviation (SD) which
is a measure of the distance of the data points from the mean. Given that for a data sample
with ‘Normality’ nearly all data is within +/- 3 σ of the centre point (the mean): the results
show that Groups 1 and 2 have camera usage times that are closely grouped around the
mean (within 3 σ ie 9 mins and 15 mins), whereas Group 3 have a more widely spread
sample with more variation in the camera usage times (within 3 σ ie 30 mins).

13
Regression Analysis

Surveys are a convenient way for conducting marketing research – especially for customer
feedback information using Likert Scales which are commonly used (eg score 1 to 7.) where
respondents indicate the importance of a factor (or their likely behaviour in future purchasing
decisions). After we have collected some marketing research data we examine it to see if
there is a pattern. We introduce a method commonly used by engineering managers for this.

Note we are not giving detailed advice here – check with a specialist before using a statistical
test as your data needs to meet the conditions for that specific test.

Regression analysis example - Sales Demand

3000
2900
2800
Units sold

2700
2600
2500
2400
2300
30 35 40 45 50 55 60
Time (Days)

Regression analysis may be used to examine relationships between factor(s) and the outcome.
Single Regression is for one factor where a “best fit” line is plotted as shown above. Note that
this – and many other statistical tests – are dependent on the Quantitative data sample having
a good level of ‘normality’ that approximates to a “bell shaped” curve.

Be aware that before we start running such tests we need to ‘clean up’ our data – remove any
‘outliers’. We also need to consider: Is the data reliable? Cronbach Alpha is commonly used
for this purpose where the results of such a test may be interpreted as follows:

Alpha > 0.7 is good (depending on sample size).

Further, before running a Regression analysis we may be advised to first use a correlation test
(such as the Pearson Coefficient, ‘r’) to examine the correlation (extent of a linear
relationship) between two data sets. The results of such a test may be interpreted as follows:

Results (-1 to +1): r > 0.5 (or < -0.5) is strong

Where a result <0 indicates a negative relationship

The module materials (videos and accompanying materials) shall introduce some of the data
analysis techniques used to examine quantitative survey results during our studies.

14