ADVANCED

MARKETING
RESEARCH
JAGROOK DAWRA
 What does a manager do?
 What is the most important skill he should have?
 Why MR?
 Why not to take decisions in absence of information?
 Shooting from the Hip – Lee Iacocca
WHAT IS MR?

 Information
 Information processing
 Tools and techniques
 SWOT
 Observation
 One-on-one interviews
 Complex statistical tools and techniques – Regression, Factor,
Cluster, etc.
FROM MARKETING RESEARCH TO MARKETING ANALYTICS…

MR
 Focus is on analysis of survey-based carefully
collected data.
 Data size typically in a few hundreds.
 Focus on prediction from consumer’s psychological
aspects – perceptions/ attitudes.
 Data collection efforts are directed towards
obtaining a representative sample
MA
 Focus is on analysis of existing data obtained from
past behavior.
 Data size enormous.
 Focus is on prediction from past occurrences.
 Data collection is automatic and often mechanical.

4
 Evolution of an organization and its information needs.
 As an organization grows, its information needs also grow.
 Is there a real need for marketing research?
 Research takes time and costs money.
 Value of information vs. Cost of information?
DEFINITION – AMA

 Marketing research: the function that links the consumer, customer, and
public to the marketer through information – information used to identify
and define marketing opportunities and problems; generate, refine, and
evaluate marketing actions; monitor marketing performance; and improve
the understanding of marketing as a process.
Business problem/
Business objective Research objective Research method Interpretation Recommendations
symptom

PROCESS
 Research Method Interpretation Recommendation

• Establish Research • Analyze Data • Prepare Final

Design Report
• Identify Information • Present Final
types and sources Report
DETAILED • Determine
Methods of
PROCESS Accessing Data
• Design Data
collection
Instruments
• Determine Sample
Plan and Size
• Collect Data
THE MOST IMPORTANT STEP?
THE MOST IMPORTANT STEP

Alice: Oh, no, no. I was just wondering if you could help me find
my way.
Cheshire Cat: Well that depends on where you want to get to.
Alice: Oh, it really doesn't matter, as long as...
Cheshire Cat:Then it really doesn't matter which way you go.
 Alice in Wonderland (Lewis Carroll)
 Research objectives, when achieved, provide the information necessary to solve the problem
identified.
 Research objectives state what the researchers must do.
IT ALL STARTS WITH DETERMINING OBJECTIVES

“The formulation of a problem is often more

essential than its solution, which may be
merely a matter of mathematical or
experimental skill”
 Albert Einstein.
QUALITATIVE VS. QUANTITATIVE TOOLS

 Measuring the world

 Exploratory vs. confirmatory research.
RESEARCH METHOD

 Types of Research
 Exploratory Research: collecting information in an unstructured and informal manner.
 Descriptive Research refers to a set of methods and procedures describing marketing
variables
 Causal Research (experiments): allows isolation of causes and effects.
SOURCES OF DATA

 Primary Data: information collected specifically for the problem at hand

 Secondary Data: information already collected
RESEARCH INSTRUMENT

 Survey instrument
 Asking the right question in the right way
DESIGN DATA
COLLECTION QUESTIONNAIRE

 Software programs are available to assist marketing researchers in preparing

questionnaires.
SAMPLE PLAN AND SIZE

 Sample plan refers to the process used to select units from the population to be included
in the sample. Gives you representativeness!
 Sample size refers to determining how many elements of the population should be
included in the sample. Gives you accuracy!
DATA COLLECTION

 Data collection is very important because, regardless of the data analysis methods used,
data analysis cannot fix bad data.
 Data collection errors may be attributed to field workers or respondents.
 Researchers must know the sources of these errors and the controls to minimize them.
DATA ANALYSIS

 Data analysis involves entering data into computer files, inspecting data for errors, and
running tabulations and various statistical tests.
 SPSS
 Excel
 Any other software

 It is important to master the method and not the software!

RESEARCH AND ETHICS
 A research company decides to leave a message on prospective respondents’ answering machines
telling them that if they call back in the next 24 hours, they will receive a valuable prize if they take
part in a survey.

 Ethical as long as true

 Upon completion of an interview, the respondent is asked to provide the names and telephone
numbers of others he or she thinks should take part in the survey.

 Ethical - snowball sampling, referral sampling

 A door-to-door salesman finds that by telling people that he is conducting a survey, they are more
likely to listen to his sales pitch.

 Unethical – sugging
 Selling under the guise of research

 What is frugging?
 Fund-raising under the guise of research
 The cover letter of a mail questionnaire says that it will "only take a few minutes to fill out." But
pretests have shown that at least fifteen minutes are needed to fill it out.

 Unethical as “few” is vague

 Telephone interviewers are instructed to assure the respondent of confidentiality only if the
respondent asks about it.

 Ethical as long as confidentiality is true

 A client insists on inspecting the completed questionnaires to assess their validity, but the
researcher suspects that the client is really interested in finding out what specific respondents said
about the client.

 Unethical if the survey is confidential or anonymous.

 A client insists on inspecting the completed questionnaires to assess their validity, but the
researcher suspects that the client is really interested in finding out what specific respondents said
about the client.

 Unethical if the survey is confidential or anonymous.

 Respondents get a gift by Lottery.

 Considered unethical by some – Gambling.

 Institutional Review Board
 Reviews/ Scrutinizes research to ensure ethical practices and protect respondent rights.
 Code of Ethics
 CASRO: www.casro.org
 MRA: www.mra-net.org
 ESOMAR: www.esomar.org
 PMRS: www.pmrs-aprm.com
PANELS

 Marketing research companies are making greater use of panels.

 Recruiting respondents who agree to participate in future studies.
WHAT DOES THE COURSE ENTAIL?
 Module 1: Fundamentals - Problem Formulation, Research Design, Data collection
methods, sampling, validity and reliability, qualitative vs. quantitative research.
 Module 2: Qualitative research techniques – Indepth interviews, Focus group discussions
 Module 3: Research instrument – Validity and Reliability, Questionnaire design.
 Module 4: Scale construction and usage, EFA, CFA, SEM
 Module 5: Behavioral research – Design of experiments, Basic analysis – difference of
means, ANOVA, Regression
 Module 6: Segmentation – Cluster analysis
 Module 7: Positioning – Multidimensional scaling
 Module 8: Product and Pricing research – Conjoint analysis, Test marketing
 Module 9: Modelling choice – LPM, Logit, Probit
 Module 10: Business and marketing strategy – PIMS
 Module 11: Brand Valuation – Models of Brand equity
 Module 12: Writing a professional research report
EVALUATION

 Mid-term, end term and surprise quizzes

 Three mini projects
 Experimental design
 Segmenting and Positioning study
 Preference study
 Small sample sizes (25 for each, from classmates/ friends) allowed.
 Purpose of the project is to learn from it – experiential learning.
TO DO

 Read the case “The Coop” for the next class on Thursday/ Wednesday.
 Open up a free account on qualtrics.com and familiarize yourselves with its features.
 CR to form groups and mail the list to me by Wednesday.
 CR to form a group id for students of MR, coordinate with the PGP/ IT office and
communicate the same to all of us.
EXERCISE 1 [IN-CLASS]

 Please write the names of five people in the class you would like to form a group with.
EXERCISE 2 [OPTIONAL]

 Writing a research proposal

 A sample shall be mailed to you.
 Assume the role of a consultant.
 Think of a business problem and study objective(s).
 Create a proposal
THE COOP:
MARKET
RESEARCH
REAL ESTATE MARKET OF GOA
2013
VM SALGAOCAR GROUP

 Ban on Mining in Goa

 Venture into Real Estate
 Start with a residential project at Talegaon in Panjim
 Offer premium flats with a lot of indoor and outdoor recreational facilities.
 How would you formulate a business/ research problem for this project?
BUSINESS PROBLEM

 To study was to understand the Goa Real Estate market

 The findings of the study aimed at providing a basic and preliminary understanding of
 Environment
 Customers
 Competitors
 Enablers
 Study 1: Real Estate environment.
 Study 2: Consumer choice and preference structure.
 Study 3: Customer profiling and segmentation.
 Study 4: Competitor analysis.
 Study 5: Study of enablers.
 Objective: Study macroeconomic factors, past trends both local and global, risk
factors affecting the real estate market.
 Key deliverables: Detailed literature review of real estate market in India, macro-
level factors impacting the prices.

STUDY 1 OBJECTIVE
 Objectives:
 Measure consumer preference, price-feature trade-offs and willingness to pay.

 Key deliverables: Utility curves, simulation results, benefit segments

STUDY 2 OBJECTIVES
 Objectives:
 This study explored the size and type of customer segments.
 The study shall also profile the customers in these segments
based on various demographic factors, media habits,
psychographics, etc.
 Key deliverables:
 Consumer segments, size of segments, profiles of segments,
create top target groups (who they are, what they’re like, how to
reach them)

STUDY 3 OBJECTIVES
 Objectives: Competitor analysis. Who are
the major competitors in Goa? How are they
positioned in consumer minds? What are the
possibilities of differentiating a new offering?
 Key deliverables: competitor analysis,
positioning strategies.

STUDY 4 OBJECTIVES
STUDY 5 OBJECTIVES

 Study of enablers in the real estate business in Goa. Different government/ legal bodies, Banks, etc., act as enablers
in the real estate business
 Key deliverables: Detail activities of enablers, Interdependence of enablers on each other and their influence on
real estate business, Preference structure of property dealers.
 Study was done and completed by students of MR course of PGP 12-14 batch, guided by two professors.
FUNDAMENTAL RESEARCH PARADIGMS
 Problems

Converting
this P to K Instruments
The world

Knowledge
 CONVERTING P TO K

Approach 1: Interpretivism Approach 2: Positivism

Define Problem Define Problem

Make Predictive statements

from past knowledge
Collect data
Collect data

Interpret Analyze data to verify if the

predictive statements are true

Convert P to K Convert P to K
 JAANE BHI DO
YAARON…

Approach 1: Interpretivism Approach 2: Positivism

Who murdered Who murdered D’Mello?
D’Mello?
Corporator Taneja, involved in
substandard bridge construction
Interview all who committed the murder
knew D’Mello
Interview all

Interpret
Analyze data to verify if the
predictive statement is true

Convert P to K Convert P to K
WHAT IS A BETTER PROMOTION STRATEGY?
EDLP
6

5.5

Price in $
4.5

 MRP: $5.50
4

3.5

 EDLP: Everyday Low 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Price HiLo
Time

6
 HiLo: High-Low 5.5

Price in $ 4.5

3.5

3
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Trip
 Research question: What effect does HiLo promotion strategy have on trust on the store?
APPROACH 1: INTERPRETIVISM

 Conduct an experiment:
 Subject one group of people to EDLP strategy over a period of time.
 Subject another group of people to HiLo strategy
 Collect data on how much they trust each store.
 Conclude
APPROACH 2: POSITIVISM

 Studies show that people hate uncertainty.

 Walmart that follows EDLP is more successful than Macey’s that follows HiLo.
 HiLo involves more uncertainty than EDLP
 We therefore postulate/ conjecture/ hypothesize:
 H0: People will trust a store that follows EDLP strategy more than one that follows HiLo
 Conduct an experiment:
 Subject one group of people to EDLP strategy over a period of time.
 Subject another group of people to HiLo strategy
 Collect data on how much they trust each store.
 Accept or reject H0.
 BASIC PARADIGMS IN RESEARCH

Positivism Interpretivism
 Governed through explicitly stated theories and  Focuses on understanding what is happening in a
hypotheses. particular context.
 Researchers try to be emotionally neutral and make  Let the data ‘lead the way’
a clear distinction between reason and feeling,
science and personal experience.
 Positivists maintain a clear distinction between facts
and value judgments.
QUALITATIVE RESEARCH
FORMS OF DATA

 Written replies to open ended questions.  Magazine articles

 Photos/ documents  Case studies written by 3rd parties

 Print ads/ story boards  Online comments
 Director’s reports  On Consumer sites
 Annual reports  Social networking sites
 Company minutes of meetings  Etc.
 Etc.  Field notes
 Recorded interviews (Voice/ Videos)  FGD records, etc.

60
 “Every animate and inanimate object on earth will soon be generating data, including our homes, our car, and yes,
even our bodies.”
 Anthony D. Williams, Author – ‘Wickinomics’

 An average person today processes more data in a single day than a person in 1500s did in an entire lifetime.
 Much of this data is qualitative, though the analysis done on it may sometimes be quantitative.

61
PROCESS OF CONDUCTING QUALITATIVE RESEARCH

 Discussion guide/ Interview Guide

 FGG/ Interview
 Sample size: How many?
 Sampling method:
 Discussion/ Interview
 Transcribe data
 Analysis
 ISSUES

 Sampling and sample size

 Triangulation
 Analysis
 Reporting results
 Analysis
Type of Purpose Example
Sampling
Maximum Document diverse variation and common patterns – Actively looking for parting-off
variation involves looking for outliers to see if patterns still hold. ceremonies of private vs. public
Search actively for confirming, disconfirming and typical cases. companies.
Homogeneous Focusses, reduces, simplifies finding patterns Focus on a homogeneous set of
companies.
Critical case Permits logical generalization and maximum application of
information to other cases.
Theory based Finding examples of a theoretical construct and thereby Need for special treatment –
elaborate and examine it. “How does a store/ retailer make
you feel special?
Confirming and Elaborating initial analysis, seeking exceptions, looking for Actively looking for the
disconfirming variation. exceptions in a study. 64

cases
SAMPLING METHOD
TRIANGULATION

 Multiple sources of data

 Multiple methods – e.g. observation and interviews
 Multiple analysts
EXERCISE – IN CLASS

 Create a discussion guide for the following business problem:

 Retail stores have their own brands called private labels or store brands. How do consumers perceive these brands? Are
private label brands perceived different from national brands?
 How do they perceive its quality? Manufacturers of Jeans (Jeans brands) define quality in terms of threads per inch or
weight (in ounces). Do consumers also perceive quality of jeans as such?
 Rules:
 Create the guide in groups.
 Please take no more than 15 minutes to create this guide. Feel free to refer secondary sources if necessary.
 Submit the discussion guide by email to the CR in pdf format only. Name of the file must be Ex_01_<group no>.pdf
FGD – 20 MINUTES
ANALYSIS OF QUALITATIVE DATA

68
COMMON SOFTWARE THAT HELP
IN QUALITATIVE DATA ANALYSIS

 NviVo
 Atlas Ti
 RQDA – Freeware uses R

69
USING NVIVO

70
USING NVIVO

 Helps organize qualitative data.

 Also helps do some basic data analysis.
 Basically gives a ‘structure’ to qualitative data and makes the analysis more scientific.

71
 72

• Review • Key ideas

• Merge • Reflect
• Refine • Link

Coding Memoing

Visuals Queries

• Models • Text search

• Matrices • Word frequency
• Charts • Matrix
• Coding comparison

QUALITATIVE ANALYSIS INVOLVES

WHAT KIND OF DATA DOES NVIVO HANDLE?

 Written text – can be read directly.

 Other formats like Photos, Audios, Videos, social media, etc. can be analyzed if appropriately coded.
 Data can be easily exported/ imported to and from Excel, Word, pdf and other file formats.

73
WHERE DOES THE DATA COME FROM?

 Interviews
 FGDs
 Secondary sources – company files, memos, minutes of meetings, etc.
 Social media like Facebook, Twitter, Linkedin, web pages,YouTube, Picasa, etc.
 Field notes
 Etc.

74
 Helps you appropriately code.
 Coded text can be easily organized into themes
 Emergent themes consistently appearing are often called propositions.
 Propositions are generalizable and have very few exceptions.
 Descriptive statistics – frequency of word occurrence, between and within group
analysis, cross tabulations, cluster analysis of word coding and similarity.
 Kappa coefficient [triangulation statistic between researchers]
 Does not establish cause and effect relationships.
 But data can be exported to Excel/ SPSS and analyzed.
WHAT KIND OF ANALYSIS DOES NVIVO 75

DO?
WHY IS CODING IMPORTANT?

 Qualitative analysis is often done to ‘explore’.

 Simple text description is often not adequate.

76
HERMENEUTICS

 Hermeneutics focusses on the meaning of the qualitative data.

 Video – credit card

77
HISTORICITY AND HERMENEUTICS

 Understanding of a topic/ quote/ concept in a particular historical context.

 For e.g. We note that ‘profits of a company have been going down’. One would be tempted to think that there is something
wrong with the company or its management.
 But if the context was put in place – for instance: The state of the economy was bad and that recession had set in, the
interpretation changes.
 Important to incorporate context at the time of taking field notes.

78
THE HERMENEUTIC CIRCLE

 It refers to the idea that one's understanding of the text as a whole is established by reference to the individual
parts and one's understanding of each individual part by reference to the whole.

79
  Interpretation of a text needs the researcher to have some prior
knowledge. For instance, she needs to know the language, its
vocabulary and social conventions.
 Social conventions change – SMS lingo: LOL, LMAO, TY, , , :-/ …
 The more poetic a person is, more cryptic is his/ her language and
interpretation is less mechanical.
 Mark Twain: “I don’t let my school interfere with my education”

80

HERMENEUTICS AND PREJUDICE

HOW MUCH SHOULD YOU TRUST WHAT IS WRITTEN ON FACEBOOK/
LINKEDIN? WRITTEN VS.VERBAL TEXT VS.VISUAL INTERPRETATION

 95% of interaction in the world is verbal.

 Written text (written by respondents) comes after due deliberation and a certain ‘caution’ exercised by the
respondent, while verbal data is not so much.
 Emotions associated with the statements can be captured only in verbal/ visual text and not otherwise.
 Verbal data cannot be accurately captured by existing technology.

81
 Attitude towards the brand
 Whose attitude – is there a
 Attitude towards the product
 Attitude towards the
performance
 Attitude towards the attributes
 Attitude towards the aftersales
service
 Attitude towards the ease of use
 Etc.
EXAMPLE – ATTITUDE
TOWARDS SONY HSC 300
difference between groups?
 Men vs. women
 Classification as per other
demographics.
 Amateur vs. professional
photographer.
 Usage – heavy vs. light user.
 Usage – purpose (wildlife/ friends
and family/ scenery/ etc.)
82
 Advertisement deconstruction – understanding the explicit and implicit
messages in an ad. A further analysis could be done to reconcile the
difference between the intended positioning strategy and the perceived
positioning of brands.
 Director/ chairman report in the annual report of a company – emphasis laid
down on CSR activities by companies and its relationship with the movement
of stock prices of that company.
 Analysis of Facebook/ twitter data on the pages of various companies to
bring out the attitude of consumers towards these companies. [objective may
need to be further focused]

83

POTENTIAL EXERCISES
POTENTIAL EXERCISES

 Brand failures: using the cases in the book on brand failures (and also perhaps the
same author’s book on brand successes) build a generalized theory around what
causes brands to succeed/ fail. See if your emerging theory conforms to the ones that
already exist in the theory.
 What type of comments/ content diffuse faster over social media?
 May need to control for size of the network.

84
BLUEMOUNTAIN RESORTS
WHAT DID WE LEARN FROM BLUE MOUNTAIN RESORT CASE?

 How important business issues are converted into research objectives.

 Sampling Plan, method
 Sample bias and response bias
 Questionnaire construction
ISSUE OF RELIABILITY AND VALIDITY

 Two extremely important aspects of a good research.

 How many of you had questions vis-à-vis reliability and validity built into your SIP questionnaires?
 What is Reliability?
 Validity?
 The degree to which
a measurement
instrument
Validity
accurately reflects
what it is designed to
measure.
 The degree to which a
measurement
Reliability instrument is
consistent in what it
measures.
VALID AND RELIABLE
VALID BUT NOT RELIABLE
NOT VALID BUT RELIABLE
NOT VALID AND NOT RELIABLE
A GOOD INSTRUMENT SHOULD HAVE:

 Objectivity: Definition of the task at hand – “Objective of playing darts is to hit the bulls eye”
 Have defined a ‘rubrics’ for your projects – sent it over email.
 “Siddarth’s ability to hit the bulls eye”
 What is being measured? Siddarth? Ability?

 Practicability: Online questionnaire may not be the right instrument for a study on Indian Kirana store shoppers.
 Reliability
 Validity
TRUE SCORE TEST THEORY Random sources
of error

X o  XT  X S  X R
Observed Score

True Score

Systematic sources
of error
TYPES OF RELIABILITY…

 Test-Retest: stability of the instrument – test it 2-3 times after a span of time.
 Alternate form reliability: in test-retest reliability, the same instrument is given the second time. In alternate form,
the questions may change.
 Split half reliability: if an instrument has six questions that measure ‘loyalty’, compare the results of first three with
the next three.
… TYPES OF RELIABILITY…

 Internal consistency – Several questions are often used to measure one construct.
E.g. ‘Price consciousness’ is measured using the statements below:
… TYPES OF RELIABILITY

 Internal consistency is said to be achieved when scores of items are highly correlated.
 Cronbach’s alpha is used to calculate internal consistency.
 Internal consistency is also called inter-item reliability.
 Inter rater reliability: more than two judges judging a boxing match.
 Three faculty members judging your SIP presentations.
TYPES OF VALIDITY…

 Face validity: There is prima facie evidence to say that the results are accurate. E.g. the study predicts a market
share of 45%. Secondary sources also confirm this.
…TYPES OF VALIDITY…

 Construct validity: Are we measuring the construct as it was defined?

 E.g. Brand Loyalty:
 Behavioral angle: A customer is loyal to a brand if (s)he buys the brand often.
 How often did you buy toothpaste in the last one year?
 Of these purchases, how often did you buy ‘Colgate’?

 Attitudinal angle: A customer is loyal to a brand if (s)he has a positive disposition towards it.
 Please rate the following brands on a scale 1 to 5 (I = dislike, 5= like)
 Colgate
 Pepsodent
 Close-up
 Etc.
…TYPES OF VALIDITY…

 Content validity: Are we measuring the construct fully?

 Trust has three dimensions: Ability, Benevolence and Integrity.
 Criterion validity: is the extent to which a measure is related to an outcome.
 Predictive validity:
 “high trust in a brand would lead to high purchase intention”.

 Concurrent validity
 Measuring attitudinal loyalty and behavioral loyalty should be highly correlated.

 Convergent validity: Two or more methods of measurement leading to similar results. E.g. market share being predicted by
analysis of scanner panel data and a survey method.
 Discriminant validity: can a set of questions discriminate between two constructs:
 …TYPES OF VALIDITY
 What do the following questions measure?
 …TYPES OF VALIDITY
 What do the following questions measure?
CLASSROOM EXERCISE…1

 Rishi measures attitudinal loyalty using a few statements constructed by him and observes low effect on
repurchase intention.
 Predictive validity?
 Construct validity?
 What needs to be done to rectify the situation?
CLASSROOM EXERCISE…2

 Chandan measures trust as an overall construct and also measures it on its dimensions of ability, benevolence and
integrity. He observes a difference between the mean scores of both measurements.
 Content validity?
 How can this be corrected?
CLASSROOM EXERCISE…3

 Jaysun measures trust using its three dimensions and gets the following Cronbach
alphas:

Dimension Cronbach alpha

Ability 0.85
Integrity 0.78
Benevolence 0.60

 How would you interpret these?

 What corrective actions would you take?
FRAMING OF THE QUESTIONS

 How important are the following attributes in an insecticide. Please give a rating
between 1 (=not important at all) and 10 (=Extremely important)
 Brand
 Safety
 Ease of application
 Action
 Effectiveness
 Smell
 Price
 As a manager in an insecticide company, can you use this information to design a product?
 Sir Humphrey: "You know what happens: nice young lady comes up to you.
Obviously you want to create a good impression, you don't want to look a
fool, do you? So she starts asking you some questions: Mr. Woolley, are you
worried about the number of young people without jobs?“
 Bernard Woolley: "Yes”
 SH: "Are you worried about the rise in crime among teenagers?“
 BW: "Yes”
 SH: "Do you think there is a lack of discipline in our Comprehensive
schools?“
 BW: "Yes”
 SH: "Do you think young people welcome some authority and leadership in
their lives?“
 BW: "Yes”
 SH: "Do you think they respond to a challenge?“
 BW: "Yes”
 SH: "Would you be in favor of reintroducing Compulsory army Service?“
 BW: "Oh...well, I suppose I might be.”
 SH: "Yes or no?“
 BW: "Yes"
 Sir Humphrey: Alternatively the young lady can get the opposite result.“

 Bernard Woolley: "How?“

 Sir Humphrey: "Mr. Woolley, are you worried about the danger of war?“

 Bernard Woolley: "Yes“

 Sir Humphrey: "Are you worried about the growth of armaments?“

 Bernard Woolley: "Yes“

 Sir Humphrey: "Do you think there is a danger in giving young people guns
and teaching them how to kill?“

 Bernard Woolley: "Yes“

 Sir Humphrey: "Do you think it is wrong to force people to take up arms
against their will?“

 Bernard Woolley: "Yes“

 Sir Humphrey: "Would you oppose the reintroduction of Compulsory army

Service?“

 Bernard Woolley: "Yes" !!!

ASSIGNMENT 1A

 Questionnaire Design: Imperatives

 Assignment: To be submitted group-wise.
 Choose a questionnaire that you used for your SIP (any one of the 6/7)
 Critically analyze the questionnaire, referring to the document (Questionnaire Design:
Imperatives)
 Submit a report and an improved version of the questionnaire.
 Make sure that the objective of the research is clearly mentioned on the top.
 Make sure the criticism is constructive – you should touch upon both the good and the not
so good aspects of the questionnaire.
 Make sure that everyone in the group is involved in the exercise.
 Submission to be made in hard copy form to the CR.
INFORMED CONSENT FORM
TYPES OF QUESTIONS

 Open-ended question presents no response options to the respondent

 Unprobed format seeks no additional information
 Advantage:
 Allows respondent to use his or her own words

 Disadvantages:
 Difficult to code and interpret
 Respondents may not give complete answers
 Closed-ended question provides options on the questionnaire that can be answered quickly and
easily.
SCALE CHARACTERISTICS/ LEVELS

 Nominal scales: those that use only labels

 Ordinal scales: those with which the researcher can rank-order the respondents or responses
 Interval scales: those in which the distance between each descriptor is equal
 Ratio scales: ones in which a true zero exists
PERMISSIBLE STATISTICS

Scale Descriptive statistics Inferential statistics

Nominal Percentages, mode Chi-square, binomial test
Ordinal Percentile, median Rank-order correlation,
Friedman ANOVA
Interval Range, mean, standard Correlations, t and z tests,
deviation ANOVA, regression, factor,
cluster and several others
Ratio Geometric mean, Coefficient of variation
harmonic mean
INCENTIVES

 How large should the incentives be?

MEASUREMENT OF CONSTRUCTS USING A MULTI-ITEM
SCALE
 Likert scale
 Semantic differential scale
 “When we talk about attitudes, we are talking about
constructs of the mind as they are expressed in response to
our questions. But usually all we really know are questions
we ask and the answers we get.”
 “What does it mean if a finding is significant or that the
ultimate in statistical analytical techniques have been applied, if
the data collection instrument generated invalid data at the
onset”
 Jacob Jacoby
MEASUREMENT IN BUSINESS ADMINISTRATION

 Constructs borrowed from other disciplines.

 Cognitive dissonance, attitude, utility
 Constructs that are trans-disciplinary
 Trust, fairness
 Constructs developed by management scientists
 Repurchase intention, Brand equity, customer loyalty, service quality, etc.
CHALLENGE OF MEASUREMENT IN
MANAGEMENT

 In physical sciences, measurement is almost always additive.

 Length: Ruler, cms.
 Weight: Scale, kgs.
 Time: Clock, minutes.
 In marketing, is the measurement always additive?
 Consider preference.
 Additive:

 Preference for a pair of jeans= Σ[(importance of attributes X desirability of its levels)]

 Preference for a pair of jeans= П[(importance of attributes X desirability of its levels)]
 The problem is complicated further because the various constructs interact with each other.
MULTI-ITEM SCALE
Instructions

Likert scale,
interval
scale

Scale items
MULTI-DIMENSIONAL SCALE

Ability

Benevolence

Integrity
 HOW DO YOU DECIDE ‘WHAT YOU ARE MEASURING
WAS THE CONSTRUCT OF INTEREST?

1. Domain of the construct.

 Define the construct very clearly
 “the buyer's cognitive state of being adequately or inadequately rewarded in a buying situation
for the sacrifice he has undergone. The adequacy is a consequence of matching actual past
purchase and consumption experience with the reward that was expected from the brand In
terms of its anticipated potential to satisfy the motives served by the particular product class.
It includes not only reward from consumption of the brand but any other reward received in
the purchasing and consuming process.”
 Howard and Sheth on ‘Satisfaction’

 Need for preferential treatment: “A customer’s inherent psychological need to be given a

preferential treatment that is not generally imparted to other customers and at other times.”
CREATING A RELIABLE AND VALID CONSTRUCT
Specify the domain of • Literature Search
the construct
Generate Sample of • Literature search, Experience survey, Insight simulating
Items examples, critical incidents, FGDs

Collect data

Purify measure

Collect data
Adopted from
Churchill GA Assess reliability • Coefficient alpha, split-half, etc.
(1979), “A Paradigm
for Developing a
Better Measure of
Marketing
Assess Validity • Criterion Validity, Discriminant Validity, etc.
Constructs”, JMR,
pp. 64-73
Develop norms • Methods of summarizing distribution of scores
FACTOR ANALYSIS
THE MATH BEHIND FACTOR ANALYSIS METHOD
FACTOR ANALYSIS

 A simple example
 3 courses – Finance, Marketing and Business Policy
 If 𝑋1 , 𝑋2 , 𝑋3 are marks (out of 10) obtained by the students,

Student Finance 𝑋1 Marketing 𝑋2 Buss. Policy 𝑋3

1 3 6 5

2 5 3 3

3 9 4 8

4 4 5 7

5 7 9 5
 These grades are functions of two underlying factors 𝐹1 [quantitative ability] and
𝐹2 [Verbal ability]
 𝑋𝐹𝑖𝑛 = 𝑙𝐹𝑖𝑛,𝑄𝐴 𝐹𝑄𝐴 + 𝑙𝐹𝑖𝑛,𝑉𝐴 𝐹𝑉𝐴 + 𝜀1

 𝑋𝑀𝑘𝑡 = 𝑙𝑀𝑘𝑡,𝑄𝐴 𝐹𝑄𝐴 + 𝑙𝑀𝑘𝑡,𝑉𝐴 𝐹𝑉𝐴 + 𝜀2

 𝑋𝐵𝑃 = 𝑙𝐵𝑃,𝑄𝐴 𝐹𝑄𝐴 + 𝑙𝐵𝑃,𝑉𝐴 𝐹𝑉𝐴 + 𝜀3

Variable 𝑋𝒊 Loading on
𝐹𝑄𝐴 , 𝑙𝑋,𝑄𝐴 𝐹𝑉𝐴 , 𝑙𝑋,𝑉𝐴
Finance 𝑋𝐹𝐼𝑁 + 0
Marketing 𝑋𝑀𝑘𝑡 0 +
Business Pol 𝑋𝐵𝑃 0 +
ROTATION
SOME TERMS

 Bartlett’s test of sphericity:

 H0: The population correlation matrix is an identity matrix.
 Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy:
 Value should be > 0.5
 Eigen value: total variance explained by each factor.
ROTATION

 Varimax – searches for a rotation of the original factors such that the variance of the
loadings is maximized.
 Most popularly used method of rotation.
 Qurtimax – minimizes the number of factors needed to explain each variable.
 Equimax – compromise between Varimax and Quartimax
 While all the above are orthogonal rotations, sometimes, oblique rotation provides a
better explanation of the factors. But this method is rarely used. They may be used
when one expects the factors to be slightly correlated.
BEHAVIORAL RESEARCH: THROUGH EXPERIMENTATION
SESSION 7
 Experiments marked the emergence of modern science in the 16th and
17th century.
 ‘Bodies That Stay Atop Water or Move Within It.’ (Galileo, 1612)
 ‘On loadstone and magnetic bodies.’ (William Gilbert, 1600)
 An experiment is a way of systematically observing a phenomenon.
AN EXPERIMENT IS DIFFERENT FROM MERE OBSERVATION

 Early scientists like Aristotle based their scientific findings

merely by observation.
 Francis Bacon the creator of empiricism. ‘Baconian methods’
(scientific methods)
 “He taught that not only should we observe nature in the raw,
but also ‘twist the lion’s tale’, that is, manipulate our world in
order to learn its secrets”
 Hacking on Francis Bacon (1983)

 Modern scientists emphasize the need to ‘control’ for all

extraneous influences that might bias or limit observation.
EXPERIMENTS AND CAUSATION IN MANAGEMENT

 Used to study an extremely well defined ‘focused’ phenomenon.

 Do consumers prefer concept A/B/C
 What is more attractive – 10% off or 10% more
 What is the psychological impact of pricing a product as Rs. 79.00 vs. Rs. 80.00?
 Used to study cause and effect.
EFFECT

 Counterfactual – David Hume (1711-1776)

 In an experiment we observe what did happen when treatment was given.
 Counterfactual is the knowledge of what would have happened if the treatment was not given.
 Effect = difference between what happened as a result of the treatment and what would have happened
without the treatment (counterfactual).
 “Control condition”
10% OFF
 Does the demand for Colgate increase if it is at a discount of 10%?
 Have observations (Sales) for non-discount periods.
 Take observations for discount period.
 Can the increased sales be attributable to the discount?
 What about extraneous factors?
 Are the consumers same in both the instances?

 It is not possible to simultaneously give and not give discounts to the same consumers.
 Cause probing research tries to create reasonable approximations to physically
impossible counterfactual.
 Make probabilistic inferences.
 1st video: Versace
CAUSAL RELATIONSHIP

 A causal relationship occurs only when:

 The cause preceded the effect
 The cause was related to the effect
 We can find no other plausible explanation for the effect other than the cause.
 In an experiment:
 We manipulate the presumed cause
 We see whether variation in the cause is related to variation in the effect.
 We use various methods to reduce the plausibility of other explanations for
the effect.
 Video 2: Does size of a plate make you obese? [1:16 – 4:38; 17:36].
CORRELATION DOES NOT PROVE CAUSATION

 Income and education are correlated.

 Do you need to have high income to have good education or do you need to have good
education to have high income?
 There may be a correlation, but no causation
 There may be a unidirectional causation –
 High income –> high education OR
 High education –> high income .

 There may be a bi directional causation

 High income –> high education –> high income

 The relationship may not be causal at all, but due to a third variable (often called a Confound).
 Wisdom causes bot income and education to be high.
MANIPULABLE AND NON- MANIPULABLE CAUSES

 Manipulable
 Dosage, extent of discount, format of discount, color and type of packaging, etc.
 Non – manipulable
 Age, gender, number of times they have bought a particular brand, etc.
AFTER AN EXPERIMENT
B

A C
 Casual Description
 A affects C.
 B also affects C.
 A affects C by a larger amount as compared to B.
 B moderates the relationship between A and C [A*B have a significant effect on C]
 Or B mediates the relation between A and C
 Casual explanation
SOME TERMS AND DEFINITIONS

 Experiment: a study in which an intervention is deliberately introduced to observe its

effects.
 Randomized experiment: an experiment in which units are assigned to receive the
treatment or an alternative condition by random process.
 Quasi – experiment: an experiment where units are not assigned randomly.
 Natural experiment: is not really an experiment because the cause cannot be
manipulated.
 Field experiment: experiment done in the field (say organization)
 Correlational study is usually synonymous with non-experimental or observational
study; a study that observes the size and direction of relationship among variables.
 Independent variable – cause that is manipulated.
 Dependent variable – effect of interest.
 Video 3: Do creative people buy ‘Apple’ products?
 Does owning an ‘Apple’ product make you creative?
MOST EXPERIMENTS ARE HIGHLY LOCAL BUT HAVE
GENERAL ASPIRATIONS

 Experiments are localized and particularistic

 Study the affect of one or two causes on a particular variable of interest.
 The aim is to make general statements about these localized phenomenon.
 While experiments are used in practical applications, they are also widely used in
theory construction.
 Video 4: Asch Experiments
ASCH EXPERIMENTS

 Two types of respondents:

 Genuinely believe the group is right. (referent informational influence)
 ‘Go along’ with the group to avoid the discomfort of not complying with the group. (normative influence)
 What happened when the ‘naïve’ respondent got an ally in the group?
 How did Asch prove normative influence?
UTOS: COMPONENTS OF EXPERIMENT

 Units
 Treatments
 Observations
 Settings in which the study is conducted.
GENERALIZATIONS

 Construct validity generalizations - inferences about constructs that research operations presents.
 External validity generalizations - inferences about whether casual relationship holds over variations in persons,
settings, treatment and measurement variables.
VIDEOS

 Video 1: Does brand have an effect on evaluation of a product?

 Video 2: Does size of a plate make you fat?
 1:16 – 4:38; 17:36 – onwards
 Video 3: Creative and innovative people buy ‘Apple’. But does buying the brand ‘Apple’ make you creative?
 Video 4: Asch experiments.
CAN YOU CREATE EXPERIMENTS FOR THE FOLLOWING RESEARCH
OBJECTIVES

 Effect of music on customer’s evaluation of a product.

 Effect of celebrity endorser on customer’s evaluation of a product.
 Congruency between the personality of a product and brand personality.
 Effect on consumer (un)fairness wrt. Point of comparison (stranger vs. friend).
 Human brains are unable to multi-task (A reason why you shouldn’t talk over mobile phone while driving 🙂).
 Effect of colour of packaging on consumer’s willingness to pay for a cup of yogurt.
 Effect of country of origin on consumer’s willingness to pay for a cup of yogurt.
MORE THAN ONE CAUSES

 A 2-cause case: does color of

packing and ‘country of origin’
have an effect on customer
willingness to pay for yogurt?
 Black vs. white packing
 Country not mentioned vs.
Switzerland (or ‘Swiss’)
mentioned on the pack
 SIMONA BOTTI, ANN L. MCGILL (2010) THE
LOCUS OF CHOICE: PERSONAL
CAUSALITY AND SATISFACTION WITH
HEDONIC AND UTILITARIAN DECISIONS,
AN EXAMPLE: JOURNAL OF CONSUMER RESEARCH,VOL.
37, PG 1065 – 1078.
 A museum director wonders whether to plan the consumers’ visits or let them choose on their own.
 A spa wants to offer a discount coupon for its treatments. It is wondering whether to give a coupon that can be
redeemed on any of the treatments or to give one that can be redeemed on a specific treatment?
 Consumers may consume the same products or
services with different goals, for example, for their
own pleasure—a hedonic goal—or to achieve some  Goal: Hedonic vs. Utilitarian
higher level purpose—a utilitarian goal. This study
investigates whether this difference in goals  Choice: Self-chosen vs. Externally determined
influences satisfaction with an outcome that was
either self-chosen or externally determined.
 CONSUMPTION GOALS
Hedonic
 Fun, sensorial, spontaneous
 Eating out for pleasure
 Visiting a museum for fun
Utilitarian
 Functional. Sensible and useful
 Eating out to better understand local culture
 Visiting the museum for collecting material for class
project.
LOCUS OF CHOICE

 Internal – when a consumer chooses something herself.

 individuals perceive themselves to be meaningful agents in what they will experience and attribute outcomes to their own
actions.
 External – choice made by fate or third party.
 individuals perceive their experience as being determined by causal forces beyond their control
 Four groups were created – respondents were randomly allocated among these four groups.
STUDY 1: 2 BY 2 DESIGN
A moderately large U.S. city is planning to
Choice, Choice,
open a new photography museum. The
Utilitarian Hedonic
curators are testing different formats.You
have been selected for the test. The
curators have asked you to imagine
No No being a antourist visitingvisiting
art student the museum
the
choice, choice, just for fun.
museum toYou will be
collect presented
material forwith
Utilitarian Hedonic some options about
your bachelor’s how You
thesis. to structure
will be a
virtual
presentedmuseum visit options about how
with some
The
You curators
to structure
can choose have
a virtual already
onemuseum chosen
of the visit
options
one
You of
The
keeping
can the
curators
in options,
choose
mindhave one
thatkeeping
already
ofyour in mind
the goal
chosen
options
is to
that
keeping
one oftheir
maximize ingoal
the is that
options,
mind
the fun toofkeeping
maximize
your
the goal your
in mind
visit. is to
fun during
maximize
that the visit.
their goal
your islearning
to maximize
duringyourthe
visit.
learning during the visit.
 Subsequently, the participants witnessed a virtual visit.
 People who were in the ‘choice’ cell got to choose and plan their visit, while the computer randomly chose for those in the
‘no-choice’ cell.
 Dependent variable: satisfaction with the outcome
 “How much did you like to visit the museum?
STUDY 1 RESULTS
 Choice enhanced the extent to which participants liked the museum visit when their decisions were led by the
hedonic goal of having fun.
 However, when participants were driven by the utilitarian goal of writing a thesis, they liked the personally chosen
museum visit only as much as the curator-chosen one.
ALTERNATIVE EXPLANATIONS

 In study 1 the no choice condition involved an expert making the choice on behalf of the consumer. The
mitigation in satisfaction observed in the utilitarian condition could then be explained with participants’ belief that
the curator was at least as qualified as they were in selecting the most pedagogically effective alternative.
 The aim of the study was to tie the results to personal causation and not to expertise.
STUDY 1: 2 BY 2 DESIGN
Imagine that you have been working suffering mild
Choice, Choice,
Utilitarian Hedonic really
back pains
hard and
during
a general
the semester
soreness.andYour
doctor tells
achieved important
you thatacademic
these aresuccesses.
just
You
symptoms
think thatof overall
a professional
body fatigue
massageand at a
recommends
local spa renowned
a professional
for the quality
massage ofat a
No No local spa
their massage
renowned
treatments
for thewould
quality of
choice, choice, represent
their massagea great
treatments.
way to reward
Imagineyourself.
having
Utilitarian Hedonic a gift certificate
Imagine having ato giftspend
certificate
at theto spaspend
that
at
offered
the spafourthat
types of massage
offered treatment.
four types of
massage
The are
You treatment
certificate
treatment.
allowedisto valid onlyany
select forof
The
massage
the
You certificate
four
are #3 (of is
massages
allowed the valid
to four only
select forof
massages)
any
massage
the #3 (of the four massages)
four massages.
STUDY 2 RESULTS
SETTING UP AN EXPERIMENT
PRESTUDY

 A prestudy is done before the study when the researcher is not sure of the extent to which the stimulus’
manipulations.
 E.g. high vs. low discount – how ‘high’ is ‘high’?
 Treatment: The differential stimulus applied to different groups.
 Pretest: information gathered before the treatment.
 Posttest: information gathered after the treatment.
 Control: the group not given a treatment or the group against which the dependent variable is studied.
EXPERIMENTAL DESIGNS

 One group Posttest-only design

 X O1
 Students pass calculus exams (at levels substantially over chance) after taking a course on calculus.
 one group post test only design with multiple substantive post-tests.
 X {O1, O2, O3…}
 Detectives investigating a murder.
 One group pre-test post-test design.
 O1X O2
 Weight - exercise program – weight
 The removed treatment design
 O1X O2
 O1’ Xrem O2’
 Hawthorne experiments
 Post-test only design with nonequivalent groups
 X O1
 . O2
RANDOM ASSIGNMENT

 Random assignment of subjects in different treatment groups ascertains that only the treatment affects the
dependent variable and nothing else.
MANIPULATION CHECKS

 Some questions included in the questionnaire that measure the ability of the treatment given to manipulate the
independent variable.
 Manipulation checks are defined as a process that is used to verify if the experiment worked or works.
 Where should the manipulation check questions be placed in an experiment?
ANALYSIS OF EXPERIMENTAL DATA

 T-test
 ANOVA
TWO SAMPLE COMPARISON

 An analyst at a department store wants to evaluate a recent credit card promotion. To this end,
500 cardholders were randomly selected. Half received an ad promoting a reduced interest rate on
purchases made over the next three months, and half received a standard seasonal ad.
 Compare the spending of the two groups.
 H0: mprom = mno_prom
PAIRED SAMPLE COMPARISON

 A physician is evaluating a new diet for her patients with a family history of heart disease.To test
the effectiveness of this diet, 16 patients are placed on the diet for 6 months.Their weights and
triglyceride levels are measured before and after the study, and the physician wants to know if
either set of measurements has changed.
 Determine whether there is a statistically significant difference between the pre- and post-diet
weights and triglyceride levels of these patients.
 H0W: mold_wt = mnew_wt
 H0T: mold_tri = mnew_tri
 ONE-WAY ANOVA EXERCISE-1

 A sales manager wishes to determine the optimal number of product training days needed for new
employees. He has performance scores for three groups: employees with one, two, or three days of
training.
 H0: m1 = m2 = m3
 H0: s1 = s2 = s3
 In response to customer requests, an electronics firm is
developing a new DVD player. Using a prototype, the marketing
team has collected data. ANOVA is being used to discover if
consumers of various ages rated the design differently.
 Looking at the DVD data, market researchers ask:
 Are the groups different from each other by age?
 H0: m18-24 = m25-31 = …=m53-59

 Are the groups between 32 to 45 different from the others?

 H0: m32-45 = m<32 & >45

 Are the two groups between the ages of 32 and 45 really different?
 H0: m32-38 = m39-45

 Can participants under 32 and over 45 be considered statistically

equivalent?
 H0: m<32 = m>45
 A sales manager wishes to determine the optimal
number of product training days needed for new
employees. He has performance scores for three
groups: employees with one, two, or three days of
training.
 H0: m1 = m2 = m3

 While significant group differences were found, there

are no prior hypotheses about how the three groups
should differ.Therefore, he has decided to simply
compare every group to every other group.
 H0: m1 = m2
 H0: m1 = m3
 H0: m2 = m3
SUMMARY STEPS FOR ANOVA

 Validate the assumption of variance equality

 Obtain the ANOVA table and results
 Visually inspect the group means
 Perform custom contrasts, tailored to your specific hypotheses
 Compare each mean to every other mean, assuming variance equality or not.
REGRESSION ANALYSIS
 Necessary Reading: Pilgrim Bank A
REGRESSION ANALYSIS

 Includes any techniques for modelling and analyzing

several variables, when the focus is on the relationship
between a dependent variable and one or more
independent variables.
 Helps understand how the typical value of the
dependent variable changes when any one of the
independent variables is varied, while the other
independent variables are held fixed
 Regression models predict a value of the Y variable given
known values of the X variables
 Prediction within the range of values in the dataset used for
model-fitting is known informally as interpolation
 Prediction outside this range of the data is known as
extrapolation
 Regression models involve the following variables:

 The unknown parameters denoted as β

 The independent variables, X

 The dependent variable, Y

 A regression model relates Y to a function of X and β

Y1      1X 1   2 X 2
LINE FITTING IN REGRESSION
 We believe/ hypothesize: Revenues are related to
economy
 => Revenues (R ) are related to income (Y), interest
rates (I), prices (P), time (T)
 => R = a + b*Y + c*I + d*P + e*T
 Assumptions:
 All influences that are not included, have no effect
 All influences that are included have a precise effect
DATA TYPES

 Crosssectional data (i.e., data collected at one point in time)

 Data Census
 Time series data (i.e., data collected over a period of time)
 Stock prices, weather reports
GOODNESS OF FIT & STATISTICAL SIGNIFICANCE

 Goodness of fit include the

 R-squared: %age of variation in dependent variable as explained
by the independents
 Statistical significance can be checked by an
 F-test of the overall fit: dependent variable explained as a
combination of the independents
 t-tests of individual parameters: dependent variable explained by
the independent variable
PRACTICE SET – 1

 How effective is my advertising program?

 Dataset - 01 reg adexp.txt
 Sales in million units
 Ad expenditure in million $.
 Is ad expenditure the only explanatory variable?
 Is the model significant?
 Does advertising expenditure significantly impact sales?
 What is the sales likely to be if ad expenditure is Zero?
 What is the per mill $ impact of ad expenditure on sales?
PRACTICE SET – 2

 What impacts the sales of chicken in the United States?

 Data set – 02 reg chicken.txt
 YEAR = Year
 Y = Per Capita Consumption of Chickens, Pounds
 X1 = Real Disposable Income Per Capita, $
 X2 = Real Retail Price of Chicken Per Pound, Cents
 X3 = Real Retail Price of Pork Per Pound, Cents
 X4 = Real Retail Price of Beef Per Pound, Cents
 X5 = Composite Real Price of Chicken Substitutes Per Pound, Cents
 Have we taken all possible variables in the analysis?
 Is the model significant?
 Are all the variables significant?
 How much consumption of chicken is invariant of all other factors?
 Do all variables impact all consumption similarly?
DATASET ‘03 BOOKS.TXT’

 Variable definitions
 Grade: Grade obtained
 Books: number of books owned
 Attend: Number of classes attended
 Model 1: Regress grade against books
 Model 2: Regress grade against books with intercept term=0.
 What happens to R-sq? is this a better model in terms of prediction?
σ 𝑌𝑖 −𝑌෠ 2
 Calculate mean square error for both the models 𝑀𝑆𝐸 = 𝑁
 Which is the better model in terms of prediction?
DATA SET BOOKS.TXT: WHAT IS THE R-SQ IN EACH MODEL.

Books Attend Books Attend

Grade
Grade Grade

Calculate the correlation between all the three variables.

MULTIPLE REGRESSION, ‘04 CAR.TXT’

 Variable definitions:
 Car: number of minutes per week spent on taking care of one’s car
 Sex: Gender, 0=female 1=male
 Age: age in years
 Extro: projection of self-image through objects (in this case the car) they own.
 Model 1: Run regression with all the variables.
 Model 2: Run regression with all the variables but use the ‘stepwise’ feature.
CATEGORICAL INDEPENDENT VARIABLES

 05 stress.txt,Variables:
 Stress: measured on a scale
 Job: 1 = primary school teacher; 2 = Secondary school teacher; 3 = College lecturer; 4 = old university lecturer; 5 = new
university lecturer.
 Dummy variable coding
 Effect coding
MODERATION (INTERACTION)

 06 test.txt;Variables:
 LEARN: learning environment
 TEST: test environment
 0 = dry; 1 = wet

 Score: final performance

 Research question: Does a person score more when (s)he prepares and gives the test in the same environment?
MODERATION (INTERACTION WHEN ONE VARIABLE IS CATEGORICAL AND
OTHER IS CONTINUOUS)

 07 stress1.txt:Variable definitions:
 STRESS: the amount of stress in a person’s life.
 EVENTS: the number and severity of events that cause stress (e.g. changing job, shifting house, etc.)
 STATUS: 0 = single, 1 = married
 Explore if:
 Higher life event scores lead to higher levels of stress.
 Married people feel less stress.
 Having a partner reduces the impact of stressful events. Events cause lesser stress in case of married people.
 Standardize the continuous variables. (why?)
MODERATION WHEN BOTH INDEPENDENT VARIABLES ARE
CONTINUOUS

 03 books.txt
 Grade: Grade obtained
 Books: number of books owned
 Attend: Number of classes attended
 Is it important to read a number of books and attend number of classes together to score better?
 Do not forget to normalize the two continuous variables.
MEDIATION - COMPLETE

Awareness Image Brand Equity

Perception of Fair Repurchase

Customer trust
treatment intention
MEDIATION - PARTIAL

Awareness Image Brand Equity

Perception of Fair Repurchase

Customer trust
treatment intention
RULES TO ESTABLISH MEDIATION*
1. Show that X is a significant predictor
of Y.
 𝑌 = 𝑎 + 𝑏𝑋; b is significant. M
2. Show that X is a significant predictor
of M.
 𝑀 = 𝑐 + 𝑑𝑋; d is significant.
3. Show that M is a significant predictor
of Y when we control for X. X Y
 𝑌 = 𝑝 + 𝑞𝑋 + 𝑟𝑀; r is significant
 If q is not significant, it is complete
mediation,
 If q is significant, it is partial
mediation.

*Baron and Kenny, 1986

MEDIATION: DATA SET – ‘08 BOOKS’

 Variable definition
 enjoy: degree of joy derived
Buy
from reading a book.
 buy: number of books
bought
 read: number of books Enjoy Read
read.
 Does buying books mediate
(also tell fully or partially)
between propensity to
enjoy books and the
number of books read?
 PILGRIM BANK
CASE 1
 Research Question:
 Would encouraging transaction migration to lower cost channels improve customer profitability?
 Research Objective:
 Is there a difference between the profitability of on-line and off-line customers?
 Secondary objective:
 Is profitability different for different types of people?
DUMMY VARIABLES

Income (Rs.) D1 D2 D3
10000 and less 0 0 0
10000-20000 1 0 0
20000-30000 0 1 0
30000 and more 0 0 1
DUMMY VARIABLES

 In excel, use if() function:

 E.g. IF(G2=1200,1,0) and IF(G2=1300,1,0)
REGRESSION ANALYSIS

 Used for prediction and forecasting

 Used to understand which among the independent

variables are related to the dependent variable, and to

explore the forms of these relationships.

 Can be used to infer causal relationships between the

independent and dependent variables.

MULTIDIMENSIONAL SCALING: PERCEPTUAL MAPPING
POSITIONING
WHY IS IT USED?

 Positioning studies
 Positioning- repositioning decisions
 Product attribute finalization
 Advertising decisions
MDS MAP (MENTAL MAP)

Outside Oriented
Going to church
Doing voluntary
service

Going to a beauty
parlor

Family Sunbathing
Powdering the baby Personal
Baking a cake
Relaxing in a bath
Using a room
Smoothening on a
freshener Wearing soft clothes hand lotion

Using fabric softener Smelling flowers

Moisturizing face with
a cream
Windows to air out
the room
Cooking dinner

Home Oriented
TWO PROCESSES

 Proxcal
 1= similar, 10 = dissimilar. ‘dissimilar’
 Alscal
 0= same pile, 1 = different piles
TYPICAL MDS QUESTIONNAIRE - PROXCAL
Using fabric Smelling a Going to a Etc.
softener flower parlor
Using fabric Small rating large rating
softener for small for large
perceived perceived
distance distance
(dissimilar)
Smelling a
flower
Going to a
parlor
Etc.
EXERCISE

 Eg. Toothpastes
 Germ killing
 Fresh breath
 Calcium
 Strong teeth
 White teeth
 Colgate
 Close up
 Pepsodent
 Brands may also form a part of the attributes.
QUESTIONNAIRE FOR ALSCAL

IIM- IIM -
 Cards Raipur Ranchi
 Instructions given to
respondent:
 Put the B-schools in IIM IIM
different piles, keeping
similar ones in one pile and
Kashipur Rohtak
dissimilar ones in different.
 You can make as many piles
IIM
as you want.
IIM Trichy
Udaipur
HOW DOES ROTATION OF AXIS HELP IN MDS
 MDS by itself is not enough. Why?
 What are the pros and cons of both the methods – Proxcal and Alscal?
 Using factor analysis (factor scores) for positioning studies
CONJOINT ANALYSIS
SOME QUESTIONS THAT CONJOINT ANSWERS

 Why does a consumer buy a product A rather than a product B?

 What features/aspects of the product are valuable to customers?
 What are the trade-offs that a customer make while buying a product?
 How much is a consumer willing to pay for a new feature of my product?
 What is the set of attributes that will maximize consumer choice?
…
…

 What are the benefit segments in the market

place?
 What should the product line be to satisfy the
desires of the benefit segment?
 If my firm offers a product A and prices it Rs. 20.00
and the competitor offers a product B and prices
it Rs. 15.00, what is the likely market reaction?
 Does the new product cannibalize the sales of my
existing product?
PREFERENCE Emotional

Rational

Bundle Preference
of due to
Attributes Attributes

Preference
Brand due to
Brand
 Is the basis of differentiation (rational vs. emotional) dependent on the type of product category?
 What products are more likely to be sold on
 Rational basis (Attributes)?
 Emotional basis (Brand)?
CUSTOMER FOCUSED PRODUCT PLANNING
Our marketing actions
Product, price

Customers Customer preferences Customer Choices

Multi-attributed view Market shares

of products Sales
Competitor’s actions
(Attribute bundle) Profits
Product, price
MULTI-ATTRIBUTED VIEW OF PRODUCTS
Products

Attribute 1 Attribute 2 Attribute 3

Level 1 Level 2 Level 3

MULTI-ATTRIBUTED VIEW OF PRODUCTS
Preference for
Products

Attribute 1 Attribute 2 Attribute 3

Preference for Preference for Preference for

Level of attribute 1 Level of attribute 2 Level of attribute 3
 Overall preference for a product= preference for attribute 1 + preference for attribute 2 + preference for
attribute 3…n
 Overall worth = sum of part-worths
 Conjoint: considered jointly
 Part-worths correspond to a single respondent.
PART-WORTH FUNCTION MODEL
8
6
4
2
0
-2 Rs. 10 Rs. 20 Rs. 40
-4
-6
-8
-10
-12
PART-WORTH FUNCTION MODEL
15
10
5
0
-5 0.5 min 2 min 10 min

-10
-15
-20
WHICH OF THESE PRODUCTS WOULD BE MORE
PREFERRED?
Product A Product B A UA B UB

Rs. 20 4 Rs. 40 -11

Rs. 20 Rs. 40
2 min 4 0.5 12
min
2 min 0.5 min
Total 8 Total 1

More preferred
 Typically many factors or attributes are included.
 Factors/ Attributes often include price and brand name.
DECIDING ATTRIBUTES

 Related to consumer choice (based on qualitative market research like FGDs or Interviews)
 Distinguish different product offerings.
 Managerially relevant.
DECIDING LEVELS OF ATTRIBUTES

 Distinct, mutually exclusive, and collectively exhaustive.

 Cover a range of interest (no extrapolation)
 Try to keep the range of levels small (say 3 per factor) and equal across factors
MO-BIKE PUMP
 Factor Levels Explanation
Price Rs. 500, Rs. 1000 or Actual price paid by you
Rs. 2000
Size: fits in a Pocket, Fanny pack or The smallest space into which the
a book bag product would fit comfortably. The
pocket, fanny pack or a book bag
would have space for other items as
well.
Total time 0.5 min, 2 min, or 10 Time to set up and connect the pump
min to a flat tire, inflate the tire with
average effort to full pressure and
put the pump back
Ease of Easy, medium or Easy = even a 6-year-old can use
inflation difficult the product
Medium = even a 12 year old can
use the product
Difficult = requires an athletic adult
DESIGNING CONJOINT CARDS

 4 attributes, 3 levels each. 3 x 3x 3 x 3 = 81 possible cards (too many)

 Choose a representative subset of say 18 cards.
 Fractional Factorial design (conjoint designer software given to you)
 Arrange the 16 cards given to you in order of your decreasing preference. Keeping the most preferred card on
the top.
 Also arrange the 4 shaded cards in the order of decreasing preference.
 How did you do it?

Cards
NEXT CLASS

 Case: Sunbeam
CONSUMER CHOICE

Solibloc Airstick

 Choice Share Fits in a Fits in a

 Market Share pocket book bag
0.5 min 10 min
Easy to Difficult to
inflate inflate
Rs. 2000 Rs. 500
AIRSTICK MAKES A CHANGE

Solibloc Airstick

Fits in a Fits in a
pocket book bag
0.5 min 10 min
Easy to Medium to
inflate inflate
Rs. 2000 Rs. 500
WHAT WOULD HAPPEN IF A NEW COMPANY
RELISTICK WERE TO INTRODUCE A PRODUCT…
Solibloc Airstick Relistick

Fits in a Fits in a Fits in a

pocket book bag fannypack
0.5 min 10 min 2 min
Easy to Medium to medium to
inflate inflate inflate
Rs. 2000 Rs. 500 Rs. 1000
CONJOINT PLAN=‘C:\Conjoint_pump_design.sav'
/DATA='C:\CONJOINT_DATA.sav'
/SEQUENCE=PREF1 TO PREF20
/SUBJECT=ID
/FACTORS= time (DISCRETE)
ease (DISCRETE)
price (ANTIIDEAL)
size (DISCRETE)
/PRINT=SUMMARYONLY.
CASE 1: SUNBEAM
 Brand Line
HLL
Soaps Shampoos Conditioners Body
Lotions
Dove Dove Dove Dove
Lifebuoy Sunsilk Fair and
Lifebuoy Breadth Lovely
plus

Depth
Hamam Clinic Plus Ponds
Clinic All
Clear
Clinic Plus
Ayurvedic
Lirl
Product Line Brand Family Brand Portfolio
REDESIGNING OF PRODUCT LINES

• Mature
• Redesign the product lines to make them more responsive to needs and dictates of its markets.
• To optimize market share of each product category.
• Goals:
– What models should be in the line?
– What should their physical appearance be?
– What should their performance characteristics be?
PROCESS

 Consumer usage and attitude survey.

 Consumer attribute and benefit survey.
 A conjoint analysis survey.
 Product line simulations.
 What were the different attributes of a food processor that were considered?
 What were the levels of these attributes?
 How many distinct processor combinations are possible with these attributes and levels?
 How many were chosen for the study?
 Mathematics behind Conjoint – Linear Programming.
 Was there a qualitative study that preceded the conjoint study? What was the major objective for this study?
 How were the huge number of configurations reduced to 27?
 What data was collected?
 What was the relevance of each data item to the study?
 How was the data collected?
 Notice how the job of sorting was made easy for the respondents.
 What were the important attributes?
 What is the benefit of having importance data?
 What were the segmentation results?
 How do you think they were found out?
 Simulations?
 What makes you say that the results of the exercise were valid (or not valid)?
 What is base case simulation?
 Assumptions during simulations.
 Simulating “what if” scenarios.
 What is the benefit of simulations?
 How reliable are these simulations?
HOW DOES THE SOFTWARE
ESTIMATE THIS?

• Uses linear programming

• E.g. Card 11 > Card 7
• S(bookbag) + T(10 min) + E(Medium) +
P(500) > S(bookbag) + T(0.5 min) +
E(Difficult) + P(2000)
• 16 equations

• S(bookbag) + T(10 min) + E(Medium) +

P(500) = S(bookbag) + T(0.5 min) +
E(Difficult) + P(2000) + Z1
• Min(Z1+Z2+…Zn)
DECIDING ATTRIBUTES

 Related to consumer choice (based on qualitative market research like FGDs or Interviews)
 Distinguish different product offerings.
 Managerially relevant.
DECIDING LEVELS OF ATTRIBUTES

 Distinct, mutually exclusive, and collectively exhaustive.

 Cover a range of interest (no extrapolation)
 Try to keep the range of levels small (say 3 per factor) and equal across factors
DESIGNING CONJOINT CARDS

 4 attributes, 3 levels each. 3 x 3x 3 x 3 = 81 possible cards (too many)

 Choose a representative subset of say 18 cards.
 Fractional Factorial design (conjoint designer software given to you)
 Assignment
DESIGNING CARDS

design.exe

Reorder.exe
ESTIMATING PART-WORTH UTILITIES

 Using Linmap
 Estimated for each respondent
 Enter data in a notepad in the following format
 RES1 6 4 15 9 5 8 16 10 13 2 3 11 7 14 12 1 20 19 18 17
 Change the extension from .txt to .dat

Pump.dat Checkdat.exe
ASSIGNMENT 2

 Rearrange the cards and rank order them.

 Input them into a text file and change its extension from .txt to .dat
 Check for mistakes using the program checkdat.exe
 What is the factor that you laid most importance on? Is it consistent with your
findings from the analysis?
 Use Linmap.exe to get utility curves and misorders
 Input the data of your friends.
 Use linmap.exe to get group utility curves.
CAUTION…

 Adding (or subtracting) the same constant to all partworths of an attribute does not
alter the relative values of overall preference.
 The fact that part-worth for 0.5 min is positive and 10 min is negative does not mean
that the respondent ‘likes’ 0.5 min and ‘dislikes’ 10 min. zero point on the partworth
scale is arbitrary.
 It is not meaningful to compare the part-worth of the level of one factor with some
partworth of some other factor.
 Units of measurements are arbitrary. Transformations can be done. But they should
be done to all partworths of all attributes.
OTHER SIMULATIONS AND ANALYSES

 Sensitivity of market share to other product attributes.

 “Optimizing” product bundle
 Competitive price reactions
 Price equilibrium calculations
 Product line simulations
 Competitive product improvements
 Entry of new competitive products
 Cross tabulation of results against different segments
WORD OF CAUTION

 Market Share vs. Profits

 Run simulation for different concepts.
 Estimate costs of tooling/ manufacturing each concept.
 Compare profitability of alternative concept.
 LIMITATION OF CONJOINT

 Attribute list is not complete.

 Aesthetics, smell, taste, etc difficult to capture.
 Measurement, sampling errors are present.
 Does not include the effect of an innovative attribute or level.
 How much does a consumer really value an attribute?
 May vary from time to time and situation to situation.
 Cannot be used when the product category is so radically new that customers
don’t even know how they may use the product.
 Cannot be used in image driven, holistic products (e.g. perfumes, paintings)
 Full profile conjoint does not take into effect the interaction effect between
attributes.
 Gilligan Island Syndrome
PRIMARY USES OF CONJOINT

 Product line planning

 New product evaluation (concept testing)
 Competitive analysis
 Pricing
 Benefit segmentation
NEW DEVELOPMENTS IN CONJOINT

 Full profile conjoint

 Regression based conjoint
 Trade – off tables
 Paired comparison
 Self explicated conjoint
 Hybrid methods
 Choice based conjoint
 Adaptive conjoint
 HB Conjoint
BRAND EQUITY

 Preference for the attribute “Brand”

SOME EXAMPLES TO WORK ON

 Choosing products (laptop, mobile, etc.)

 Choosing a subject in the second year
 Choosing a B-school in the first year
 Choosing a career
 Choosing a place of work
 Choosing a club/ committee
 Choosing mess food
 Choosing a hostel room
 Etc.
CLUSTER ANALYSIS
 Hierarchical Cluster Analysis
 Two step cluster analysis
 Cluster analysis using Linmap
HIERARCHICAL CLUSTER ANALYSIS – READING A DENDOGRAM