Research Design

UNIT 3 RESEARCH DESIGN

Objectives

After studying this unit you should be able to:

• Explain the concept and importance of research design in the context of
marketing research problems
• Discuss the causality factor in the context of research design
• Describe the various kinds of research designs and their applications for
different research situations.
Structure
3.1 Introduction
3.2 Research Design - Meaning and Importance
3.3 Causality: The Basis of Classification of Various Types of Research Designs
3.4 Exploratory Research Design
3.5 Descriptive Research Design
3.6 Factors Influencing Experimental Validity
3.7 Quasi-Experimental Designs 3.8 Experimental Designs
3.9 Summary
3.10 Self-Assessment Questions
3.11 Further Readings

3.1 INTRODUCTION
In the last two units you studied about the nature and importance of marketing
research. In this unit you will learn about the first phase in planning the research
project, which is formulating a research design. Based on causality, research designs
have been divided into four categories of exploratory, descriptive, quasi-experimental
and experimental designs. While exploratory and descriptive studies provide
information on events and attributes from which inferences can be drawn, they can
only offer tenuous conclusions. In order to draw conclusions it is important to
formulate quasi-experimental or experimental designs. This units describes different
types of designs within the four categories mentioned above as well as the marketing
situations where these designs could be most appropriately applied.

3.2 RESEARCH DESIGN-MEANING AND IMPORTANCE

The research design is a comprehensive master plan of the research study to be
undertaken, giving a general statement of the methods to be used. The function of a
research design is to ensure that requisite data in accordance with the problem at
hand is collected accurately and economically. Simply stated, it is the framework, a
blueprint for the research study which guides the collection and analysis of data. The
research design, depending upon the needs of the researcher may be a very detailed
statement or only furnish the minimum information required for planning the
research project. To be effective, a research design should furnish at least the
following details.

a) A statement of objectives of the study or the research output.

b) A statement of the data inputs required on the basis of which the research
problem is to be solved.

c) The methods of analysis which shall be used to treat and analyse the data inputs. 23
Marketing Research : Let us try to understand the elements through an example:
Concepts and Design
A newly opened supermarket sells a broad line of merchandise ranging from
provision to household appliances and kitchenware. The general manager (sales)
believes that the total profits of the supermarket can be enhanced by getting people to
buy in larger quantities which could be achieved by offering attractive cash discount
on bigger purchases. As the other executives are doubtful about this. The hypothesis
can be tested by carrying out a marketing research exercise.
1) The objective is to calculate the margin earned on sales when this discount is
offered and compare it with the margin when discount is not being offered.
2) The data to be collected over a period of time
a) Sales in rupees to a selected sample of customers during the period when
the discount is offered.
b) Sales in rupees to the same customer when the discount is not being
offered.
c) The average order size in the two periods.
d) The average margins earned during the two periods.
e) The cost of promotional inputs regarding the discount.
3) The analysis of the data will be
a) Sales in rupees in period I minus those in period II.
b) Subtract cost of incentives.
c) Also subtract cost of promotional inputs.
The importance of research design lies in the fact that it makes a statement of what is
to be done in order to achieve the research objectives and how it is to be done. It is an
expression of what is expected of the research exercise in terms of results and the
analytical input needed to convert data into research findings.
The research design furnishes a clear idea as to the activities that would need to be
undertaken in order to achieve the research objective. It is therefore, helpful if the
research design after being finalised is put in writing to enable the researcher to have
a frame of reference and prevent the study from deviating.
At the stage of analysis and interpretation also, the research design helps in providing
direction to the computation and interpretation process to arrive at solution and
recommendations. This is however not to suggest that a design is a rigid straitjacket
to which the study must always conform. The research design only represents an
expectation of likely results but as the study proceeds, many unexpected results may
come forth which may necessitate framing of new hypothesis or at least modifying
some. The research design is only a guiding and not a limiting framework for
research study.

3.3 CAUSALITY: THE BASIS OF CLASSIFICATION OF

VARIOUS TYPES OF RESEARCH DESIGNS
There are four types of research designs which are broadly classified as:
1) Exploratory Designs
2) DescriptiveDesigns
3) Quasi-Experimental Designs
4) Experimental Designs
These designs will be discussed in detail very shortly. However, one point may be
noted that the basis of classification of various types of research designs is causality,
which is the subject matter of discussion of this section.

Let us start our discussion with an example. Suppose the sales manager of a company
manufacturing table fans carried out a training programme for its salesmen operating in
24 a state. Three months after the training programme, it was observed that there was an
increase in sales in that state by 40 per cent. The sales manager claimed that the Research Design
training programme was effective and therefore, the salesmen operating in other
states should also undergo the same training programme. We can see that the sales
manager is inferring that the training programme has resulted in a sales increase of 40
per cent. That is to say that training programme is the cause of increased sales.

The sales manager is referring to sales training as a causal variable and the sales
increase as the effect variable. We may now ask a question whether this statement is
valid? Has the sales training really resulted' in increase in sales? The possible answer
is that we cannot say with certainty that the sales training has caused the increase in
sales. There could have been other factors which might have been responsible for the
increased sales. Increase in sale might have been caused by decrease in the price of
the table fans, a strike at a competitor's factory, increased product penetration in the
distribution channel, weather conditions, etc. Therefore, it is very essential that the
sales manager should know that conditions under which proper causal statements can
be made. To be able to make any causal statements, one should ensure whether the
following three conditions are met:
i) We must have very strong evidence to say that there exists a strong association
between an action (causal variable) and the ultimate outcome (effect variable).
ii) The other condition for the causal relationship is that the action (causal factor)
must precede the observed outcome (effect variable).
iii) We must have strong evidence to say that there were no other possible factors
(causal factors) which could have resulted in the observed outcome.
The first condition is that of concomitant variation. Concomitant variation is the
extent to which a cause, X, and effect Y occur or vary together. In our example, the
sale training programme and increase in sales would need to occur together. To be
able for us to conclude causality, the condition of concomitant variation between
variables in question must exist. However, it may be worth mentioning that a strong
association between two variables does not constitute a proof of a causal relationship,
It is quite likely that the strong association between two variables may be as a result
of random variation or both the variables may be influenced by an extraneous
variable.
The second condition for causal relationship is the requirement that the causal factor
(action) must occur either prior or simultaneously with effect factor(outcome).
However, the fact that the action precedes outcome does not establish causal
relationship. It might be a mere coincidence that sales training took place prior to
increase in sales of the table fans. There is also a possibility that sales training and
increase in sales of table fans are strongly associated. This however, does not prove
the existence of causal relationship.
The third condition for inferring causation is the absence of other possible causal
factors. This means that all other possible factor influencing the outcome (in our case
increase in sales of table fan) are either absent or are kept constant. It is only then we
could say logically that the sales training has resulted in increase of sales of fable
fans. In reality, it is impossible to find the absence of other factors or to hold some
factors constant. For example, we know that the sales of table fans is influenced by
weather conditions. Is it possible to keep weather conditions constant? Or can we be
sure that the competitor would not change the price? The obvious answer is "No".
In a case where the outcome is completely determined by only one causal factor, we
can say that causal factor is the deterministic cause of the outcome. That is the `causal
factor' in this case is both necessary and sufficient condition for the occurrence of
outcome. However, in a situation where the outcome is influenced by a host of causal
factors, any of the causal factor is the probablistic cause of the outcome. That is to
say it is a necessary but not a sufficient condition for the occurrence of outcome.
There are three possible ways to control the influence of extraneous variables. Firstly,
we may physically control the extraneous variable. For example, a company trying to
study the impact of two different packagings of a product on sales may control the 25
Marketing Research : extraneous variable like price by keeping it constant for both packaging containing
Concepts and Design the same amount and quality of good. The second way to control the effect of
extraneous variables if the physical control is not possible is to randomize the
assignment of treatments to test units. The third way to control the extraneous
variables is through the use of experimental designs, the discussion of which -would
follow in the subsequent sections. If the control of extraneous variable on the
dependent variable is not possible by any one of the method, we say that experiment
is confounded and such an extraneous variable is called a confounding variable.

Activity 1

Suppose the manufacturer of a particular brand of a desert cooler decreased the price
of his sets by 5%. It was observed that there was an increase in sales during the
succeeding four months as compared to what the company had prior to price
reduction. Has the price reduction increased the sales? Justify your answer.

.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.......................................................................................................................................

3.4 EXPLORATORY RESEARCH DESIGN

Ideally all marketing research projects must start with an exploratory research as this
helps in providing a sharper focus of the situation and a clearer definition of the
problem at hand. The exploratory research design, as the name suggests, involves
getting a feel of the situation and emphasises a discovery of ideas and possible
insights that may help in identifying areas of further rigorous study. For example a
food product manufacturer, wanting in introduce a breakfast cereal may be in
knowing the desirable attributes of such a product before really defining the product
concept. The main objective of the exploratory research is to fine tune the broad
problem into specific problem statement and generate possible hypotheses. It
therefore, gives useful direction for farther research. The exploratory studies are
mainly used for:

1) Providing information to enable a more precise problem definition or hypothesis

formulation.

2) Establishing research priorities.

3) Giving the researcher a feel of the problem situation and familiarising him with
the problem.

4) Collecting information about possible problems in carrying out research, using

specific collection tools and specific techniques for analysis.

Since exploratory studies are not conclusive studies, the design of the study is highly
flexible and informal. However, rarely ever does formal design exist in case. of
exploratory studies. Structured and/or standardised questionnaires are replaced by
judgement and intuitive inference drawing on the basis of collected data.
Convenience sampling rather than probability sampling characterises exploratory
designs. The generally used methods in exploratory research are:

a) Survey of existing literature

b) Survey of experienced individuals

c) Analysis of selected case situations.

26
Survey of existing literature Research Design

Published literature presents a very economical source of study for the purpose of
hypothesis generation and problem definition. A large variety of published and
unpublished data is available through books and journals, newspapers and
periodicals, government publications and individual research projects as well as data
collected by trade associations. A lot of data is also generated internally in the
company and some of it could be relevant to given problem situations. You will read
more about types of secondary data and their use in marketing research in the unit on
Data Collection. While survey of existing literature may not provide solutions to the
research problem, it can certainly provide direction to the research process.

Survey of experienced individuals

Clue to solving many a marketing problems can be had by talking to individuals who
have expertise and ides about the research subject. These individuals could be top
executives, sales managers, sales men and channel members who handle the product
or related products and consumers or potential consumers. The information collection
exercise does not involve a scientifically designed survey, it is merely an attempt
together all possible information about the subject of research from people who have
specific knowledge about it. The success of this type of experience survey depends
upon the freedom of response given to the respondent as well as upon the expertise
and communication ability of the people questioned. However, at this stage, since the
researcher also has very limited experience with the research problem, he may not be
able to elicit very valuable responsers from the individulse.

Analysis of selected cases

This method involves the selection of a few extreme examples reflecting the problem
situation and a thorough analysis of the same, In certain cases this sort of study may
help in identifying the possible relationships that exist between the variables in a
given marketing problem situation. The emphasis in this type of study is upon
understanding the research subject as a whole. For example, a company is interested
in finding out the reasons for the wide variation in sales productivity of its salesmen,
as an exploratory study it could thoroughly analyse the case of some of its best and
some of its worst salesmen. This exercise may help in identifying possible
relationships between demographic and /or personality variables which may affect
variation in sales productivity. The relationships, their extent and direction can then
be measured using conclusive research designs.

Activity 2

Study three marketing, research projects conducted in your organisation or any other
organisation that you aft familiar with:

1. What sort of preliminary research preceded these projects:

A) .........................................................................................................................

B) .........................................................................................................................

C) .........................................................................................................................

2. What was the type of exploratory design used for this preliminary research.?

.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
......................................................................................................................................
27
Marketing Research :
Concepts and Design 3.5 DESCRIPTIVE RESEARCH DESIGN
Primarily in use for preliminary studies, this type of designs are employed to
facilitate description and inference building about population parameters and the
relationship among two or more variables. Description or inference could be'
quantitative or qualitative in nature. Descriptive designs only describe the
phenomenon under study attempting to establish a relationship between factors. The
data collected may relate to the demographically or the behavioural variables of
respondents under study or some situational variables. For example, descriptive
research design could be suited to measure the various attributes of successful sales
people, or evaluate a training programme or a retailing situation.
The design could be used to study how customers behaved when a new sales
promotion programme was introduced. It does not, however, determine the extent of
association between the different variables i.e. the income and age of people as
associated with response to the sales promotion. Descriptive design can however, be
used to draw inferences about the possible relationships between variables.
As descriptive designs are aimed only at providing accurate descriptions of variables
relevant to the problem under consideration, they are generally used for preliminary
and explorative studies. Very often however, the decision makers choose to accept
descriptive data which would permit inference drawing about causality between
variables. They may not want to or may not be able to afford experimental studies in
terms of time involved and as such descriptive design may at times be used for
conclusive studies also.
Descriptive designs are a very frequently used, perhaps the most commonly used
category of research design. In short descriptive research can be used for the
following purposes.
a) To describe the characteristics of certain groups of interest to the marketer e.g.
users of the product, potential users, non users, possible receivers of promotional
communication by the company and so on.
b) To estimate the proportion of people in a given population who behave in a
certain way for example the proportion of consumers who are prone to deals.
c) To make specific predictions for specified future periods.
d) To develop inferences whether certain variables are associated, for example
income and shopping place preference.
In comparison to an exploratory research study, the descriptive study is more formal
and less flexible. As the descriptive design in directed at collecting qualitative and
quantitative data to enable, description of variables, it involves formulation of more
specific hypothesis and testing them through statistical inference of discipline designs
may include case research designs, longitudinal and cross sectional designs as well as
focus group studies. Of these we shall discuss the panel, cross-sectional and focus
group designs.

Panel Research Design

Also known as longitudnal research, the panel design involves the continual or
periodic information collection from a fixed panel or sample of respondents. The
elements of this panels may be retail outlets, deal9rs, consumers or just individuals.
A panel is expected to be constant over time and adjustments are made to provide for
dropouts and representativeness. A continual measurement of variables relating to the
elements of the panel provides a very good idea of their response patterns and may
provide clues for future behaviour .The longitudal analysis used involves repeated
measurements of the same variables to facilitate a variety of inferences to be drawn
about the behaviour of the elements of the panel. Recently the omnibus panel has also
been devised where though a fixed sample of elements is maintained but the
information collected from the sample members may vary over time or vary between
individuals. For example, at one time information on attitudes with respect to a
28 proposed packaging change may be
sought while at another time the information on attitudes with respect to a proposed Research Design
packaging change may be sought while at another time the information may pertain
to respondent reaction to an advertising copy.

Though relatively new in India, panel studies are being increasingly used by research
agencies to enable a continual tapping of consumer attitudes, developing buyer
profiles, brand behaviour in the market and predicting changes in market position of
different brands.

Cross Sectional Design

The cross sectional study is aimed at taking a one time stock of the situation or the
phenomenon in which the decision maker is interested. Cross sectional designs give
the picture of the situation at a given point of time. For example, if the marketer is
interested in knowing the market position of the company vis-a-vis its competitors or
the number of people who recall the advertisement of the company out of the total
population of viewers, a cross sectional design could be employed to gather
information. Since the data gathered in this type of design consists of responses from
a sample which contains a large number of sources, it provides a "Cross section" of
the situation. In quite a few marketing situations cross sectional data furnish good
inputs. Consider the following examples:

• An instant food manufacturing company has provided display racks for its
retailers so that its product can be prominently displayed. Before embarking
upon a second round of distribution of these racks, it wants to find out how
many retailers are actually using the racks and the type of retail establishments
that are using them.

• An electronics goods company which provides high quality after sales service to
its customers and wants to evaluate the customer response to it.

None of the examples given above seek to establish causal relationships. Cross
sectional studies may however uncover relationships, which may be conclusively
tested by using the experimental designs.

The cross sectional studies may have an informal design rather than an explicitly
described design. When the latter is the case, the design only includes the nature and
the source of data, the analytical methods and the nature of the expected results.

Focus Groups

The focus group, an informal design, basically consists of bringing together a group
of people to have an extensive, free flowing discussion relating to their experiences
or opinion.

The usual procedure is to tape record the conversations and later analyse them. The
discussion in the focus group is led and guided by a moderator, who has a fairly
standardised way of establishing a relaxed and congenial atmosphere; initiating the
discussion and keeping it on the desired focus. The purpose of the focus group design
is exploratory, seeking to get respondent opinions of experiences on a particular
subject matter. The objectives of the focus group study are clearly specified, but there
is no structure of the discussion pattern. In fact, a distinct effort is made to keep the
discussion free flowing. The success of the focus group design depends upon a well
trained and disciplined moderator who can keep the discussion on to the desired
course and elicit responses from all the participants of the group. Consider the
following example.

An established toilet soap manufacturer wants to change the advertising theme for his
prestigious brand of soap, so as to strengthen its image of a luxury soap using a focus
group of young women college students, housewives and some executives. A free
wheeling discussion about expectations from toiletries in general and bathsoap in
particular was generated and recorded to give valuable ideas and insights for the 29
creative strategy to be used for the advertising theme.
Marketing Research : Focus groups serve as excellent tools for preliminary research as they reveal
Concepts and Design problems and opportunities which emanate from the free and frank discussions by the
consumers about their problems, fancies, expectations, satisfactions and
dissatisfactions. They are easy to organise and interesting to analyse, and become
very good sources of ideas for campaigns, product improvement or repositioning,
desirability of deals, dealer selection and so on. The danger associated with focus
group lies in the possibility that once analysed, the focus group data may be used for
conclusive purposes, to draw conclusions or how people behave. The size of the
focus group being very small, it is dangerous to draw conclusions about consumers'
behaviour only on the basis of the recorded and duly analysed discussions.

3.6 FACTORS INFLUENCING EXPERIMENTS

VALIDITY
Before taking up the discussion on Quasi-Experimental Designs and Experimental
Designs, it would be worthwhile to understand the factors that influence experiment's
validity. There are two important concepts of validity that are related to
experimentation namely internal validity and external validity.

Internal Validity: It means that no other plausible cause of the observed results
should exist except those tested. This is the basic minimum that should be present in
an experiment before any conclusion about treatments can be made. The presence of
other extraneous factors or the absence of internal validity, makes the experiment
confounded.

External validity: It relates to the question of "generalization" of the results obtained

from the experiments. We are interested in knowing whether it is possible to project
the conclusions of the experiment to what population, geographical areas etc.

Although, we would want an experiment to satisfy both internal and external validty,
it is not uncommon to find experiment meeting all the internal validity requirements
and yet be invalid externally. This could possibly happen because of the following
reasons:

i) Environment at the time of test may be different from the environment of the
real world Where the decisions are taken.

ii) Results obtained in a six week test may not hold good in an application of, say,
twelve months.

iii) The population used for experimentation of the test may not be similar to the
population where the results of experiment could be applied.

iv) Treatments in the test may be different from the one used in the real world.

It is desirable to understand the sources of errors that are responsible for reducing
experiment's internal and external validity. This will help the researcher guide the
30 choice of design to use.
An experiment's validity is influenced by a host of extraneous factors. These are Research Design
explained below:
1) History: History deals with the occurrence of specific nonrecurring events that
are external to the experiment but occur at the some time as the experiment.
Suppose we measure the sales of a territory before and after salesmen were
trained The difference in the sales may indicate the treatment effect (sales
training). However, an increase in sales can not be attributed to sales training
alone . An improvement in business condition before and sales training might
have resulted in the increase in sales. Therefore, longer the time period involved
between observations, greater is the chance that history will confound an
experiment of this type and hence will significantly affect the results.
2) Maturation: Maturation is similar to history except this it is concerned with the
changes that occur with the passage of time in the people involved in the design.
In our sales training examples, an increase in sales may be due to the fact the
salesmen have become more experienced with the passage of time and hence
they know their customers better. It may be noted that it is not only people that
change over time, but also stores, geographic regions, and organisations etc. The
duration of time period has a direct impact on results of maturation effect.
3) Testing: In the process of an experiment, respondents may know that their
behvaviour is being observed or the results are being measured. This would
sensitize and bias respondents. In our example of training salespersons, if the
salespersons know that they are being trained to know how effective the training
is, they might respond differently than what they would have done otherwise.
For example, if sales in a territory are recorded before training of the
salespersons and another observation on sales is taken after the training and if
the respondents are aware of it, they are likely to become sensitized and behave
differently.
As another example, consider a case where the respondents have filled a pre-
treatment questionnaire. The respondents are likely to respond differently if they
are asked to fill the same questionnaire after they are exposed to the treatment.
This is because they have now become `experts' with that questionnaire.
4) Instrumentation: This refers to the effect caused by changes in measuring
instrument or process during the experiment.
If difference in total rupees sales volume per territory on a "before" and "after"
training is used as the yardsticks to know the effectiveness of sales training, a
price change during this time interval could make a substantial difference. This
is one example of changing instrument. The experience of the investigators, a
change in the mood of the investigators, a change in the investigators are some
other examples which may affect the measurment and hence the interpretation of
the results.
5) Selection Bias: The selection bias occurs when test units are selected in such a
way as not to be representative of the population. It refers to assigning of test
units to treatment groups in such a way that the groups differ before being
subjected to treatments. If groups are selected on the judgement of the
researcher,. there is always a possibility of selection bias. The selection
procedure should be random as this will result in a measureable random
variation. However, if a non-random procedure is 'adopted, the results will be
affected by a non-measurable error.
6) Test Unit Mortality: It is quite possible that some of test units might drop out
from the experiment while it is in progress. Referring to the sales training
example, some of the sales persons, who are undergoing training might quit
before the training is over. There is no way of knowing whether it is those who
were not improving left the training. Also, it is not possible to determine
whether those who left the training would produce the same results as those who
stayed till the completion of the training programme. In short, the interpretation
of the results will become difficult task.
31
Marketing Research : Activity 4
Concepts and Design
Explain with the help of suitable examples the factors that influence experiment's
validity.

.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................

3.7 QUASI-EXPERIMENTAL DESIGNS

In order to facilitate the discussion on Quasi-experimental Designs and Experimental
Designs (to be discussed in 3.8), the following notational system will be used.
i) X represents the exposure of a test group to an experimental treatment whose
effect is to be observed or measured.
ii) refers to the measurement or observation of the dependent variable on the test
units. Suppose more than one observation or measurement is taken, the
subscripts O1, O2, O3 etc., will be used to indicate temporal order.
iii) R symbolizes that individuals have been selected at random to separate
treatment groups or groups have been allocated at random to separate
treatments.
iv) Movement from left to right indicates a time sequence of events.
v) All notations in one row indicate that subjects belong to that specific treatment
group.
vi) Vertical arrangements of the notations mean that the events occur
simultaneously.
To make the national scheme clear , let us consider the following example:
Example 1
The symbols 0, x 02 represents that the measurement were taken on a group both
prior to (01) and after (02) of receiving the treatment (X).
Example 2
The symbols
R O1 x.O2
R x O3
Indicate that two groups of individuals were assigned at random to two different
treatment groups at the same time. The groups received the same treatment (X). One
group received both a pretest (O1) and post test measurement (O2). The second
received the post test measurement (O3) at the same time as O2.
Now having understood the notation scheme, we begin our discussion on Quasi
experimental design.
Quasi – experimental Designs: In these designs the researcher has control over data
collection procedures but lacks control over the scheduling of the treatments and also
lacks the ability to randomize test units ‘exposure to treatments. There are various
designs which fall under the category of quasi-experimental designs. Some of these
will be discussed here. These design have inherent weakness as their internal validity is
questionable. They lack the control attributes of the truly experimental designs .The
32
following quasi-experimental designs will be discussed in this section. Research Design

1) After-Only without Control Group

2) Before-After without Control Group
3) The Static-Group Comparison
4) Logitudinal Designs (Time Series Designs)
5) Multiple Time-Series Designs.
1) After-Only without Control Group: This design is also called a “One-shot case
study.” The design may be presented symbolically as follows:

XO

In this design, test units are not selected at random. The test units are either self-
selected or arbitrarily selected by the experimenter. A single group of test units
are first exposed to treatment (X), and then measurements are taken on the
dependent variable afterwards (O).

For example, a sales manger may request the members of his sales force to
volunteer participation in the sales training programme and their sales
performance is measure after training programme is completed.

It is not possible to draw meaningful conclusions from such a design. This is

because we do not have any observation (measurement) on the sales
performance of the sales persons prior to their undergoing training.

The level of O is the result of various uncontrolled factors. The effect of history,
maturation, selection and mortality are substantial and non measurable making
the design internally invalid.

2) Before-After without Control Group: This design is also termed a “One group
pretest-post test design.” This design is superior to the design” After-Only
without Control Group.” It provides a measurement of what existed prior to the
test group being subjected to the treatment. This design is presented
symbolically as follows.

O1 x O2

Referring to the sales training example as discussed in "After-only without

Control Group", a pretest measurement of sales performance has been added to
the design discussed earlier. Now the question is whether the difference is sales
performance as measured by O2 -- O1 can be attributed, to the sales training
programme? One could list out a number of extraneous factors which could be
responsible for this difference (O2 - O1) thereby making this design useless and
disqualified for internal validity. Some of these factors are listed below:
i) The test units had volunteered for the sales training programme (selection
bias)
ii) Economic environment might have improved during that period (history)
iii) The salespersons might have gained more experience during the training
period which would have resulted in the improved sales performance
(maturation).
iv) The pretest measurement might have affected performance(testing)
v) Prices might have changed during that period (instrumentation)
vi) Some test units might have left during the period of training (moratality)
3) The Static-Group Comparison Design: In this design we use two groups of
subjects. Group 1 is exposed to the treatment whereas Group 2 has not been 33
Marketing Research : exposed to it. Group 2 is called a control group and is used as the baseline for
Concepts and Design comparison since it has not been exposed to the treatment. In marketing, the
control group treatment is called the current level of marketing activity. The
design may be diagrammed symbolically as follow:

Group 1 (Experimental Group): X1 O1

Group 2 (Control Group): X2 O2

Here X2 represents baseline marketing programme (existing programme) which

is to be compared with the new programme (treatment) X1 . For example X1
could be new sales training programme which is to be compared with the
existing programme X2 (may be called control group treatment).

The experiment result O1-O2 could be attributed to the selection procedure

adopted for two groups. The test units were not randomly selected. So the result
may be attributed to the pretest difference caused by selection procedures. Test
unit mortality could also be another probable cause. These two reasons put
together could make this design invalid.

4) Longitudinal Designs (Time-Series Design): This is an extension of one-group

pretest-post test design. Here we take periodic measurements for some test units.
The treatment is introduced and the measurement on the test units are again
taken at various points of time.

For example, a company may be interested in studying the market share of one
of its products. It takes measurement on the market share at periodical intervals.
Then a new advertising compaign is introduced and periodic measurements on
the market shares are taken again.

In order to explain this design, let us consider the following figure and explain
the results of this type of design.

At the outset, this design may appear to be similar to one group pretest-post test
design. However, there is a difference between the two designs. In the pretest-
post test design we would have obtained only two measurements namely 04 and
05 Here in this design we have in fact taken many pretest-post test observations
so that there is more control over extraneous variables. In the pretest-post test
design, where we use only two observations 04 and 05, the analysis of various
situations is as follows:

a) The results for situations A, B and C are identical namely that the market
share has increased.
34
 b) For situations D, E and F, the conclusion is that the new advertising Research Design
compaign (X) has no impact on market share as it has remained constant.
The conclusions arrived at from the various situations of the Longitudinal
designs are as follows:
i) In situation A, the compaign has had a short-run positive effect, after which
the market share sustained.
ii) In situation B, it is observed that the rise is market share was temporary.
The new compaign had a short-run positive effect. We note that the market
share reverted to the level existing before the application of the treatment.
iii) In situation C, we observe that the changes that occur after the application
of the treatment are a continuation of a trend (they also existed prior to the
application of the treatment). Therefore, it is quite likely that the new
advertising compaign had no effect on the market share.
iv) In situation D, the conclusion would be same as obtained for situation C.
v) In situation E, it appears that the new advertising campaign seems to have
stopped the decline in market share.
vi) In situation F, the treatment (the new advertising campaign) had a delayed
effect and therefore it required a longer period to appear.
Because of multiple measurements in this design, the effect of maturation,
testing and instrumentation can be ruled out. Further if the test units are selected
randomly and measures are taken to prevent panel members (test units) dropping
out, one could atleast reduce selection and test unit mortality factors. However,
the main weakness of this design is that here it is not possible to control history.
However, the researcher could maintain a record of various events like unusual
economic activity, economic changes etc. and if no changes are observed, one
could reasonably conclude that treatment has had an effect.
5) Multiple Time-Series Designs: This designs may be diagrammed as:
Experimental Group O1 O2 O3 O4 X O5 O6 O7 O8
Control Group O'1 O'2 O'3 O'4 O'5 O'6 O'7 O'8
Here the experimental group is the one which was selected for sales training (X)
and the observation O1,O2…………………. represents the sales volume of this
group. A comparable group or the sets of the salesmen could be treated as
control group. Their measurements on sales volume are symbolised as O'1 O'2
O'3 O'4…………………etc. After training, the measurements on the sales volumes of
both the groups are taken. The effect of the sales training may be obtained by
comparing the average sales volume for the two groups before and after
treatment. The main weakness of this design is the possibility of an interactive
effect in an experimental group.
Activity 5
Describe a few specific situations from the functional area of marketing where
you would recommend each of the following designs as being best suited.
Justify your answer.
a) After-Only Without Control Group
b) The Static-Group comparison
c) Longitudinal Designs.
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
....................................................................................................................................... 35
Marketing Research : Activity 6
Concepts and Design
Compare and contrast One Group Pretest-Post Test Design with Longitudinal Design.
Use a suitable example from marketing.

.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................

3.8 EXPERIMENTAL DESIGN

True experimental design provide a stronger and more reliable basis for the existence
of casual relationship between variables. Here, the researcher is able to eliminate all
extraneous variables from the experimental and the control group through the use of a
random selection procedure. One of the advantages of using random selection
procedure is that we can use inferential statistical techniques for the analysis of
experimental results. One such technique is the analysis of variance. The following
experimental designs will be discussed in this section.
1) After-only with One Control Group
2) Before-After with One Control Group
3) The Solomon Four Group Design
4) Completely Randomised Design
5) Randondzaed Block Design
6) Latin Square Design
1) After-Only with One Control Group:-Here, two groups are experimental and
one control group are set up by randomisation procedure. The design requires
only one treatment and post test measurement of both of experimental and the
control group. The design may be written symbolically as follows:
Experimental Group R X O1
Control Group R O2
It may be noted that no pretest measurements are taken in any of the two groups.
The difference between the post test measurement of the two groups is taken as
a measure of treatment effect. Here 0, measurement is composed of treatment
effect and all the other extraneous variables and 0, is composed of extraneous
factors only. Therefore :
O1-O2 = (Treatment effect + Extraneous factors) - (Extraneous factors) =
Treatment effect.
Here the absence of pretest measurement provides control over the testing and
instrument effect. Also the interactive testing effect cannot occur. We, however,
assume in this design that the pretest measurement of the characteristics under
study are same in both the groups. It is also assumed that the test unit mortality
affects each group equally. These assumptions can be justified if we take large-
randomised samples. This design is very widely used in marketing.
2) Before-After with One Control Group: This design may be presented
symbolically as:
Experimental Groups R O1 X1 O2
Control Groups R O3 O4
This design is also called Pretest - Post test Design.
In this design two groups one `experiment and one `control' - are set up by
randomisation process. The extraneous variables operate equally on the experimental
36
 and control groups. Pretest measurement are taken in both the groups. However, Research Design
it is the experimental group which is exposed to treatment. Post test
measurement are taken for both the groups.
Since the extraneous factors operate equally on both the groups, therefore, the
only difference between the two groups is as a result of the treatment of the
experimental group.
Let A = Change in the measurements of the experimental group
O2 - O1
Let B = Change in the measurements of the control group
O4 - O3
Therefore, treatment effect - A-B = (O2 – O1)-(O4 -O3)
The main problem with this design is that it is not possible to remove sensitizing
testing effect from the measurement effect. The estimate of the treatment effect
given above also includes the sensitizing testing effect. The sensitizing test
effects occurs in the experimental group and not in the control group. The
individuals in the experiment group get affected because of the pretest
measurements and may become very cautious and behave differently after being
exposed to the treatment. Therefore the measure A = O2 – O1 will include a
sensitizing testing effect. Since the control group is not exposed to the treatment
and therefore this is missing in B = O4 - O3Therefore, this shows that the
treatment effects as given by A-B will include sensitizing testing effect and
therefore we cannot generalise the result of our treatment.
3) The Solomon Four Groups Design : This design is also called the four group
six study design. This test overcomes the limitations of pretest - post test control
group designs. This design is infact a combination of `Before After with one
control group' and `After only with one control Group' designs. The design is
symbolically presented as:

Experimental Groups : R O1 X O2

Control Groups : R O3 O4

Experimental Groups : R O6

Control Groups R O6

You may note that the experimental group 2 and control group 2 do not receive
any pretest measurements. The results of group 2 help us in measuring and
eliminating the sensitizing testing - effect since this type of sensitizing cannot
occur in measurement 05. The results of the difference of various pretest and
post test measurements give us the following effects.

37
Marketing Research :
Concepts and Design

Please note that in order to measure the change in experiment group 2 , we

required an estimate of pretest measurement on O5 , we assumed that groups are
equal before the experimental because of the random assignments of the subjects
to all four of our group I and control group 1 and control group 1 respectively .
Given the just mentioned assumption, the pretest measurement of experimental
group 1 and control group 1 could be taken as an estimate of pretest
measurement of experiment group 2. that is what we used in the equations (iii),
(iv),(v) and (vi) listed above).
This design is also referred to as "the ideal control experiment." As you have
seen that we are not only able to control all extraneous variables and sensitizing
testing effect, but also could get their estimates. However, this design is not very
commonly used in marketing because of increase in cost, time and effort
required to conduct the experiment.
The three experimental designs discussed above allow us to manipulate or
control an independent variable so as to measure its effect on the dependent
variable thereby enabling us to make proper casual statements. Below we will
discuss some designs where we can manipulate or control more than one level of
independent variable on the dependent variable..
4) Completely Randomised Design: The test is useful when the investigator is
interested in studying the effect of one independent variable on the dependent
variable. The independent variable should have a number of categories (nominal
scale). Each category of the independent variable is considered as a treatment.
Suppose we wish to sell a product and want to know what price should be
charged. Here the independent variable is price and the three different levels of
price namely low, medium and high are used as treatments. The test units
(shops) are randomly assigned to three different treatments (price levels).
Suppose our dependent variable is sales, we could always compare average sales
level corresponding to each of the treatment to determine which treatment (pi-
ice level) was the best.
The limitation of this design is that it does not take into account the effect of
various extraneous variables on the dependent variable. In the example which
we have considered, the possible extraneous factors would be competitor's price,
size of the shops, the price of the substitute of the product in question and so on.
It is assumed in this design that the extraneous factors have the same impact on
all the test units.
5) Randomized Block Design: The randomized block design is an extension of
completely randomized design. In the completely randomized design it was
assumed that all the extraneous factors have the same impact on all the test
units. This may not be true.
In the examples used in the completely randomized design, we had three price
38 levels tested in various shops. We did not take into consideration the fact that the
 differences in the result could be due to the differences in the size of the shops. Research Design
One could classify various shops on the basis of their size as:

Small size shops

Medium size shops

Large size shops

In the randomized block design it is possible to separate out the effect of one
extraneous factor from the results thereby providing a clearer picture of the
impact of treatment on sales.

In this experiment, we may for example, randomly assign nine small size shops
to the three price level in a way that there are three shops for each price level.
Similarly one could assign randomly nine medium size shops to the three price
levels and nine large size shops to the three price levels. Having done this, one
could use analysis of variance technique to separate out the effect of extraneous
variable (size of shops) from the total experimental error and also analyse the
effect of treatment (price level) on sales (dependent variable)

6) Latin Square Design: This design allows the researches to control and measure
the effect of two extraneous variables on the dependent variable. Here the two
extraneous variables need to be divided into as many categories as the number
of levels of the independent variable (treatment). The Table 1 below illustrates
the layout of Latin Square Design.

Table 1: Latin Square for Various Advertising Campaigns

Packaging
I I1 III IV
1 A B C D

Cities 2 B C D A

3 C D A B

4 D A B C

Suppose a marketing organisation wants to study the impact of four types of

advertising campaigns on the sales of a product. The sales could be affected by say
two extraneous variables like the type of city and the type of packaging. If Latin
Square Design is to be used the cities need be divided into four types namely 1, 2, 3,
4 and four different types of packaging say I, II, III and IV are required. In Table 1
we have identified four categories for the row variable; the type of city, and four
categories for the column variable, the type of packaging. The treatment variable the
types of advertising campaign is identified as A, B, C and D. We note that the
number of categories of extraneous variables is" the same as the number of
treatments (4).

The treatments (four advertising campaigns) are assigned randomly to the cells of the
table in such a way that each treatment appears once and only once in each row and
each column. One this arrangement is made, the analysis of variance technique could
be used to separate out and measure the effect of the two extraneous variables in
question and also analyse the effect of treatments (four types of advertising
campaign) on the dependent variable namely sales of the product.

This design is very complex to set up. Further, it can be very expensive to execute.

39
Marketing Research : Activity 7
Concepts and Design
How would you use 'Before-After' with one Control Group design to measure the
effectiveness of a new advertising on the sales of a particular brand of shampoo?
Also, mention the weakness of this design.

.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
........................................................................................................................................

Activity 8

List out a few marketing studies where “Latin Square Design” could be used.

.........................................................................................................................................
.........................................................................................................................................
.........................................................................................................................................
………………………………………………………………………………………….
.........................................................................................................................................
.........................................................................................................................................
........................................................................................................................................
…………………………………………………………………………………………
…………………………………………………………………………………………
…………………………………………………………………………………………

3.9 SUMMARY
This unit on research design gives an overiew of the research designs and their
application in marketing research. After discussing the meaning and benefits of a
research design, the discussion focusses on the four basic types of designs namely
exploratory, descriptive, quasi-experimental and experimental. Causality, the
relationship between data and conclusions, forms the basis of this classification. The
nature of causality and the need to measure it has also been discussed.

The major design categories and their applications have been described. Specific
subtypes within the major categories have also been discussed. Under the descriptive
design, panel design, cross sectional design and focus groups were discussed. The
quasi- i experimental designs discussed include after only without control group,
before-after without control group the static group comparison, longitudinal design
and multiple time series designs.

Under the experimental design categories the designs covered are after-only with one
control group; before-after with one control, Solomen four group design completely
randomized design, randomised block design and Latin square design. Questions of
internal and external validity of the experimental validity of experiment have also
been discussed .

3.10 SELF-ASSESSMENT QUESTIONS

1) Explain the solomon-four group design. How far 1) does this design succeed in i
controlling different extraneous variables? Illustrate your answer with the help
40 of a suitable example from marketing.
2) Explain the various extraneous variables which if not controlled in an Research Design
experiment may contaminate the effect of the independent variable.
3) What design would you use to
a) Test whether the coloured newspaper advertisement is more effective than
the existing black and white advertisement ?
b) Know which of the several promotional techniques is most efficient in
selling a particular product?
c) Determine whether additional shelf space allocated to a product would
increase the sales of that product in the supermarket?
d) Determine the type of customers who would be the first adopters of a newly
introduced brand?
Justify your answer.
4) What are the marketing research situations suitable for
a) Focus group study
b) Panel research design
c) Cross-sectional design
d) Quasi-experimental design
5) Distinguish between exploratory and descriptive research design.

3.11 FURTHER READINGS

Freund, John E. and Frank J. Williams, "Elementary Business Statistics - The
Modern kiproach", Prgntice Hall International Edition.
Kinnear, Thomas C. and James R. Tayler, "Marketing Research - An Applied
Approach", McGraw-Hill International Edition.
Luck, David J. and Ronald s. Rubin, "Marketing Research" Prentice-Hall of India
Pvt. Ltd.
Green, Paul E. and Donald S. Tull, "Research for Marketing Decisions" Prentice Hall
of India Pvt. Ltd.
Westfall, Boyd and Stasch, “Marketing Research - Text and Cases” Rechard D.
Irwin. Inc.

41