Chapter 3

Page 49
 CHAPTER III : RESEARCH METHODOLOGY

A research methodology is a systematic path searching for information about a particular

problem. It tells about the implementation of the scientific methods and tools for analysis of
the data. Research is broadly classified into two types - fundamental research and applied
research. Fundamental research is asking a question about “Why an act has happened?” and
solves a general problem. It formulates a theory to explain a problem. Applied research is
asking a question about “How an act has happened?”and solves practical problems. It is
dedicated finding an immediate solution to a problem. A research methodology is a study of
the definition of a problem, formulating a hypothesis, collecting the data, analyzing the data,
interpreting the data, findings the conclusion, and preparation of the report. The major
features of the research are systematic, logical, replicable, and logical. It helps in decision
making for formulating regulations, policies, and strategies in an organization or government
bodies. Research may be qualitative or quantitative, sometimes research may be of both
types – qualitative and quantitative.

3.1. Research Problem

The research problem is a crucial part of research and requires to be processed in a stepwise
manner. It starts with the statement of the problem, recognizing the nature of the problem,
collecting literature, and finding the conclusion. This chapter discusses the research
methodology used for collecting and analyzing the data of “A study of Online Marketing
and its impact on Consumer Decision Making”. It comprises a target study area, selection
of sample respondents, sampling procedures, data collection, and interpretation.

The objectives of the study are as described below.

i. To segment the stages of the consumer decision making process in online marketing.
ii. To examine the influence of demographics on consumer decision making in online
marketing.
iii. To understand the impact on consumer decision making in online marketing.
3.2. Selection of Study Area

The study area has been taken in two districts of Madhya Pradesh and Chhattisgarh. The two
districts in each state have been purposefully selected for the study. Jabalpur and Bhopal have
been taken in Madhya Pradesh while Bilaspur and Raipur have been chosen in Chhattisgarh.
The reason for the selection of these districts is its demographics structure and proper

Page 50
availability of the internet for using online marketing. The study has been carried out in both
rural and urban areas of these districts. Table 3.1 are the blocks of Jabalpur and Bhopal in
Madhya Pradesh and Raipur and Bilaspur in Chhattisgarh.

Table 3.1 Blocks in Selected Districts of M.P. and C.G.

Name of the Block
Bhopal Jabalpur Bilaspur Raipur
Berasia Jabalpur Belha Abhanpur
Hujur Kundam Kota Arang
Kolar Majholi Takhatpur Dharsiwa
Panagar Masturi Tilda
Patan
Shahpura
Sihora

3.2 Identification of Consumers

The selection of the consumers has been done on the purchasing of consumer durable through
online marketing in the past year. The major products taken in durable are Television,
Washing Machine, Refrigerator, and Juicer Mixer Grinder. The selection of respondents has
been routed to the consumers of these durable items. The consumers have been identified in
all these four districts in urban and rural areas for the collection of the data. Equal
opportunity for all respondents has been ensured irrespective of gender, social category, and
religion.

3.3 Research Design

The research design states the method for conducting a particular study. It is a collection of
various stages of the research in sequential order to consider the problem. The collection of
the data and its analysis with suitable methods helps in designing an appropriate research
design. Research design can be classified as follows:

1. Exploratory Research Design - When the problem is not clearly defined for
investigation, this research is called an exploratory research design. It does not give
a conclusive result. The result can be qualitative or quantitative. The primary

Page 51
 research methods used in exploratory research design are surveys, interviews, focus
groups, observations. The secondary research methods used in exploratory research
are an online collection of information, literature research, and case study research.
Exploratory research is not structured research, but it is flexible and scattered.
Exploratory research starts with the identification of the problem, hypothesis
development, and further analysis of the data is done to develop a result.
2. Descriptive Research Design - When the research is describing a phenomenon of
characteristics of the population, this research is called a descriptive research design.
Some major characteristics of this research design are quantitative research,
uncontrolled variables, cross-sectional studies. It discusses demographic elements to
collect quantifiable information for statistical analysis of the population.
3. Experimental Research Design - When the research is done among two variables
and a scientific approach is used to establish a cause and effect behaviour, that
research is called an experimental research design. Time plays an important role in
determining a relationship in cause and effect experimental design. The three major
types of experimental research design are pre-experimental, true experimental, and
Quasi-experimental research design.

Here, the study is related to a descriptive research design that describes a problem and finds
a solution for it. The nature of the study is specific and purposive. The descriptive research
design can tell about where, when, what, and how types of the questions of the study while it
can answer ‘why’ about the problem. The descriptive research design helps to find about
opinions and views of the consumers, and the development of a questionnaire to gather the
data to establish a relationship among the variables.

The descriptive research design is of three types:

1. Observational Method - This method is used for quantitative and qualitative

observations. A qualitative observation method is based on monitoring activities
or characteristics. The characteristics of the respondents are natural and realistic,
the observer can act as an observer, an observer as a participant, or completely
participant. A quantitative observation method has the primary focus on the
collection of numeric data and figures. The result is drawn by statistical and
numerical methods.

Page 52
 2. Case Study Method - This method is an in-depth study and widens the further
scope of the study by developing a hypothesis. It can’t reveal the cause-effect
relationship because it can not produce appropriate predictions. The output might
be biased at the researcher’s end.
3. Survey Method - This method comprises of a questionnaire or polls. It is just like
to collect feedback from the respondents by filling the questionnaire. It can be a
blend of open-ended and close-ended questions. It can be done by pen and paper
or online filling of the questionnaire. In most of the cases, the questionnaire based
research is in a part of descriptive research when the sample size is enough to
have a better statistical analysis.

This study is based on the survey method of descriptive research design. A structured
questionnaire is designed to study the objectives of the research and significant and relevant
questions are included to get the desired collection of data (Saunders et. al. 2009, pp. 360-
363). The reliability and validity of the questionnaire are checked through academic and
professional experts. The questions should be written in a clear, concise, and comprehensive
way so that they can have a single meaning. Open and closed-ended questions can be taken
into consideration for the study. The open-ended question gives a wide space for the research
but it is very difficult to code the data for statistical analysis. Closed-ended questions can be
easily analyzed through statistical tools but it limits the no. of responses in a question.
However, it is a current trend to frame the Likert type scale question for imparting research. It
can be easily analyzed statistically and can produce a sound output of the research.
Qualitative forms of questions can also be taken as required in the research.

3.3.1 Sampling Design

Sampling Design is a part of statistics that can tell about the characteristics of the population.
Since the population can’t be considered for the research as it would large. The main aim is to
draw a sample from the population which can accurately represent the population. However,
the constraints of financial support and time limit the efficient sampling effort.

Population – The study is based on online marketing specified in the consumer durable
category in Madhya Pradesh and Chhattisgarh. Every person has an equal chance of being a
respondent for filling the questionnaire. Thus, the population comprises of males and females
with different socioeconomic and cultural background. Their level of education and
profession would be different from each other. Thus, the population seems to be infinite. The

Page 53
whole population is not considered for the study, however, the researcher has selected a part
of the items representing the universe (Zikmund et al., 2009).
a. Sampling Unit – Within the infinite size of the population, the researcher has chosen
only those respondents who have purchased a single item of consumer durable
(Television, Washing Machine, Refrigerator, and Juicer Mixer Grinder) through online
marketing in past one year. The researcher has also chosen those respondents whose
family income is below three lacs per annum, and they know about how to use online
marketing. The respondents have been taken from both urban and rural areas in Madhya
Pradesh and Chhattisgarh.
b. Sampling Method – The researcher has adopted a simple random sampling method to
collect the data. The researcher has personally visited the study area for filling the
questionnaire by the prospective respondents. The areas are nearby railway stations, bus
stops, IT parks, Malls and shopping complexes, entertainment parks, educational
institutes, and colleges, and some other relevant locations to randomly select the sample.
Equal opportunity has been given among males and females of both rural and urban
areas of the concerned districts in Madhya Pradesh and Chhattisgarh.
c. Sample Size – The sample size is determined as ten times the number of items in the
questionnaire for using principle component analysis. Since the researcher has taken 64
items in the questionnaire for the analysis of the data, the required sample size would be
640. However, the total sample size in the study is 648 comprising 162 samples from
each district (Jabalpur, Bhopal, Raipur, and Bilaspur). A total of 900 questionnaires have
been distributed to collect the data, but the filled questionnaire is returned by only 670
respondents. Out of 670, 12 questionnaires are rejected due to incomplete information.
Responses of 10 questionnaires are removed as a part of outliers; checked through box
plot. The final shortlisted questionnaire for the study is left with 648 in counting.
According to (Babbie, 2015), a 60% response rate is assumed as a good and when the
response rate is greater than 70%, it is considered as very good. In this study, the
response rate is 72.00%. The data collection is at a very good stage.
3.4 Reliability and Validity of the Measures

The reliability is a measurement that specifies the amount of error-free instrument to ensure
consistency across time and the related measurement in the instrument (Sekaran et. al., 2010).
The reliability of a scale can be measured by Cronbach’s Alpha (Nunally, 1978). The value
of Cronbach’s Alpha should be greater than 0.7 for a reliable scale. The validity of an

Page 54
instrument is an indicator of relevancy which checks about the subjective statement and its
actual meaning which is intending to measure in the developed instrument.

3.5 Hypothesis of the Study

Following are the hypothesis and sub-hypothesis of the study.
Hypothesis 1

H0 = There is no significant difference in education in online web characteristics and their a)

Psychological b) personal c) social d) political e) cultural impact on consumer decision
making.
H1= There is a significant difference in education in online web characteristics and their a)
Psychological b) personal c) social d) political e) cultural impact on consumer decision
making.
Hypothesis 2:
H0 = There is no significant effect of gender in online web characteristics and their a)
Psychological b) personal c) social d) political e) cultural impact on consumer decision
making.
H1 = There is a significant effect of gender in online web characteristics and their a)
Psychological b) personal c) social d) political e) cultural impact on consumer decision
making.
Hypothesis 3:
H0 = There is no significant effect of age in online web characteristics and their a)
Psychological b) personal c) social d) political e) cultural impact on consumer decision
making.
H1 = There is a significant effect of age in online web characteristics and their a)
Psychological b) personal c) social d) political e) cultural impact on consumer decision
making.

3.6 Pilot Study

A pilot study is conducted on a small size scale to specify the research design. It helps in
recognizing the hurdles which are faced in data collection and improving the designed
measurement scale for data analysis. It increases the experience of the researcher to signify
the sample size. A pre-test has been conducted among 50 respondents to test the reliability
and validity of the instrument measures of the research. Content and Construct validity have

Page 55
been checked in consultation with experts and academicians to identify the actual
measurement.

3.7 Collection of the Data

Primary and secondary data have been used in this study, however, there is a dominance of
primary data collected through a structured questionnaire. The data has been collected
through the questionnaire filled independently by the respondents. A very rigorous strategy
has been adopted to collect data to minimize errors.

a. Collection of Primary Data – The collection of primary data has been done through a
structured questionnaire filling by the respondents. To frame an effective questionnaire,
the researcher has taken care of the sequence of the questions to maintain the quality of
the questionnaire. Moreover, the researcher has drafted the questionnaire after an
interaction with expert views and academicians to produce a meaningful and relevant
instrument for the study. The drafted questionnaire is first tested with a pilot study to
know about the understanding of the respondents regarding the questionnaire,
willingness and personal views in sharing such asked information, time is taken by them
in filling a completed questionnaire. The big universe and large sample size have been
handled with full care in collecting data.
b. Collection of Secondary Data – The secondary data is required to structure the
conceptual framework of the research. The collection of secondary data has been done
through several reports, journals, books, documents of govt. and non – government
agencies. The Census 2011 has been used for population data in the study. The
secondary data is the backbone of research. It helps in comparing the analysis with our
collected data and the government publications and reports.

The demographic profile of the respondents has been shown in Table 3.2.

Table 3.2 Demographic Profile of the Sample

Jabalpur Bhopal Bilaspur Raipur
Gender
Male 88 87 91 90
Female 74 75 71 72
Education
Literate 6 7 5 6

Page 56
 10th 9 7 7 6
12th 26 25 26 28
UG 69 69 71 68
PG & Above 52 54 53 54
Age
15-25 Years 31 27 33 30
25-35 Years 54 47 36 54
35-45 Years 35 51 51 47
Above 45 42 37 42 31
Income
Below 3 lacs 15 8 9 11
3-6 lacs 48 53 51 52
6-9 lacs 62 64 55 68
Above 9 lacs 37 37 47 31
Family Types
Nuclear 64 65 67 63
Joint 98 97 95 99
Source: Primary
3.8 Clustering Technique

The clustering technique is a tool of classifying a set of objects into a group of similar objects,
whose characteristics are unknown but resembles a set of specified variables. It is generally
used for classifying large samples. The class or group of similar objects is termed as a cluster.
A cluster is called as good when it is efficient and effective. A cluster can be efficient when it
uses the least number of clusters or class in counting while a cluster can be effective when it
is statistically and commercially feasible. The cluster analysis can be performed on both
continuous and categorical dataset. But it is not necessary to take both continuous and
categorical dataset.

3.8.1 Categorization of Cluster Analysis

There are several methods of cluster analysis. In general, the following are broadly
categorized methods of cluster analysis.
A. Hierarchical Method
B. Partitioning Method

Page 57
 C. Model based Method
A. Hierarchical Method

This method designs the group of data into a cluster tree is called a dendrogram. The
hierarchical cluster is non-overlapping. The parent cluster signifies only common similarity.
However, if the object is specialized for a particular variable, it is assigned to the descent
cluster. The cluster analysis measures the degree of similarity between objects across
variables in the dataset. It consists of two types of measures – similarity and dissimilarity
coefficients.

 Correlation Coefficients –The correlation coefficient is a measure of similarity

between two variables. It can be used among two respondents rather than two
variables to draws the pattern of responses of both respondents. The correlation
coefficient remains unaffected by dispersion differences across variables but there is
no information about the elevation of scores. The correlation coefficient doesn’t tell
about the distance between two respondent’s profiles. It is denoted by ‘r’.

 Euclidean Distance –Euclidean distance is the alternative measure of the geometric

distance between two objects. It has been drawn that as small would be the
Euclidean distance, there would be more similarity among the cases. But, the
Euclidean distance can be influenced by variables in large dataset. There is a need
for standardizing scores before analysis.

 Average Linkages–In this stage of hierarchical cluster method, the two most similar
case is chosen to form the nucleus of the cluster. The average similarity within the
cluster is calculated to compare with the similarity of each remaining case of the
cluster. That case is added to the cluster which has the highest similarity for the
average similarity.

 Ward’s Method – The ward’s method is a complex form of linkage method. The
main motto of the ward’s method is to join cases into clusters to minimize the
variance of the cluster. Each case starts as its cluster which is later merged to reduce
the variability within the cluster. In other words, it can be said that the average
similarity of the cluster is measured at every stage. The difference between average
similarity and each case within a cluster is calculated and then squared, which is
called as a deviation. The sum of squared deviations gives a measure of error within
a cluster which can produce the least increase in the error.

Page 58
The hierarchical clustering method can be further classified into the following methods.

a. Agglomerative Hierarchical Clustering – This method comprises of bottom-up

strategy by putting each object in its cluster. The cluster is then merged to form a large
cluster. The large cluster is the topmost level in the hierarchy of the cluster. This process
continues until all clusters are grouped in a single entity or the termination situation
arrives in the entity (Jain & Dubes, 1988).
b. Divisive Hierarchical Clustering – It is opposite to agglomerative hierarchical
clustering which comprises of top-down strategy by putting all objects in a single cluster.
Further, it divides its cluster into small segments to form its cluster or termination
situation arrives. It terminates after achieving the desired no. of clusters or the distance
between two clusters is above the defined threshold distance. The most common method
used for divisive hierarchical clustering is as follows (Zhang et al., 1996; Guha et al.,
2000).
i. Density based Divisive Hierarchical Clustering – It is based on the density of
the clusters in the data space. This method helps to grow a cluster until it
exceeds a threshold density of the neighbourhood. The neighbourhood cluster
has to contain at least a minimum number of data points under a certain radius
of the cluster. It filters outliers and gives an arbitrary shape. (Ester et al., 1996,
Ankerst et. al., 1999, Berkhin, 2002)
ii. Grid based Divisive Hierarchical Clustering – This method divides data space
into a finite number of cells to form a grid structure (Sheikholeslami et al., 2000;
Agrawal et al., 1998). Its computational complexity is independent of data space,
it is dependent on the number of cells in each dimension of quantized data space.
B. Partioning Method

When a database of n objects are given to form k clusters, the Partioning method performs
the object into k partitions (k ≤ n). The formed partition is termed as a cluster. The cluster is
formed to optimize an objective criterion, called a similarity function. It results in similar
objects within a cluster and objects within other cluster is entirely different from it. The
commonly used partitioning are
a. K-means cluster method – Mean value of objects is represented in each cluster.
Cluster similarity is measured in terms of the mean value of the objects It has high
intra-cluster similarity but it has low inter-cluster (Macqueen, 1967, Dubes and Jain,
1979).

Page 59
 b. K-medoids method – A medoid is defined as the point whose dissimilarities with
other points of the cluster are minimum. Each cluster is represented by one of the
objects located near the center of the cluster. It is a fast method and can converge in a
fixed number of steps (Fox et al.,1991).
c. K-modes method – K-modes method is an extension of the k-means method. This is
used to handle categorical data which defines the cluster based on matching categories
between the data points (Huang, 1998).
C. Model Based Method

The model based method optimizes the data with the help of a mathematical model. It follows
two approaches:
a. Statistical Approach – This method is based on Bayesian classification theory to
measure the relative reliability of the sample. It needs to characterize the reliability of the
source of the data.
b. Neural Network Approach – This method is used to study the complex and rigorous
sample. It is distribution free and is independent of knowledge of the statistical
distribution of the objects in the dataset. It requires the weight of each object to analyze
its influence through iteration techniques.

3.9 Statistical Tools

Part A of the questionnaire comprises demographic questions namely gender, education, age,
occupation, family income of the respondents. The descriptive statistics are applied to
determine the share of the respondents in the sample.

Part B of the questionnaire comprises questions related to stages of the consumer decision
making process. It will help in determining the behaviour of the respondents in their purchase
decision making through online marketing.

Part C of the questionnaire comprises questions related to the impact of consumer decision
making in purchasing consumer durable through online marketing. This will help in
determining the variables which can produce an impact on consumer decision making in
online marketing.

Inferential statistics have been used in part B and part C to analyze the data and test the
hypothesis through factor loading, correlation, and regression analysis. The statistical
package for social sciences 22.0 has been used in the study for statistical assessment.

Page 60
Incorrect and missing values have been carefully checked before the start of the analysis.
Moreover, normality has been also checked to identify and remove outliers for the good
result of the analysis.

3.10 Measurement Scale

The measurement scale has been developed to classify the data for statistical analysis. There
are four types of measurement scales - Nominal Scale, Ordinal Scale, Interval Scale, and
Ratio Scale. The nominal scale classifies the numbers of the data as a ‘tags’. It is assumed
that the non-numeric variables in a nominal scale is not a value and doesn’t define the
characteristics of the variable. The ordinal scale shows the relative order without resembling
any degree of the variation between them. The relative ranking of the ordinal scale signifies
the magnitude of the variable. The interval scale is a quantitative measurement scale that
quantifies the difference between the values of the variable. Ratio scale is also a quantitative
form of measurement scale to compare differences or intervals. It embarks an origin or zero
points characteristics. Ratio scale can be used for simple mathematical calculations like
addition, subtraction, multiplication, division. It has also featured unit conversion, mean,
median, and mode calculation.

The measurement scale for consumer decision making has been developed based on some
works of literature. Table 3.3 shows the measurement scale for the consumer decision making
process and to measure characteristics of online marketing and its impact on consumer
decision making.

Table 3.3 Measurement Scale of CDMP

Needs Recognition (Chiu et al., 2014)

Information Search (Satish and Bhardwaj, 2010)
Evaluation of Alternatives (Hsin Chang & Wen Chen, 2008)
Purchase Decision (Gefen, 2000) (Pavlou, 2003)
Post Purchase Decision (Van der Heijden, 2003)
Social Influence (Venkatesh et al., 2012)

3.11 Scope of the Study

Online marketing is growing at a fast pace in India. Several companies have started their
business through online marketing. It has been expected that the consumer durable industry

Page 61
will grow at a pace due to rural consumption, reduction in replacement cycles, availability of
several brands and varieties of consumer durable. However, heavy items for household
categories are still facing a challenge in the growth of their market through online marketing.
Online marketing will help in recognizing the impact of consumer purchase behaviour of
durable. The period of the study has been covered from July 2018 to July 2019. The data has
been collected from the respondents by visiting the concerned districts of the study area, and
a structured questionnaire is used as a research instrument to obtain the data for the research.

3.12 Ethical Perspectives

The data collection is a rigorous work to visit the study area, meet with the respondents, and
give an equal opportunity in selecting the sample. This process requires ethics in research so
that there is no biased case for getting a fruitful result. The data collection process also
requires privacy and confidentiality of the responses which are given to the prospective
respondent. Before the start of filling responses by the respondents, the objective and purpose
of the research are to be explained in short so that they can have faith in the researcher and
also ensure them to use their data for academic purposes only. The respondents are briefed
about the time to contribute to filling the responses. They are independent of filling the
responses as per their wish. No forceful behaviour is to be done to get the questionnaire
completely answered. The respondent's name, address, work, mobile no. email Ids are not
used in the analysis of the data.

Page 62