MARKETING MANAGEMENT - II

P G DM ’ 6 1 S ESSION 3

P ROF. KA R N I KA
Choosing Data Collection Methods
Data Collection methodology is determined by the following

◦ Sampling – e.g.: information given in sample list

◦ Type of Population

◦ Question Form – e.g.: close-ended for self-administered

◦ Question Content – e.g.: sensitive information

◦ Response Rate

◦ Budget, Resources, Duration

Population and Sample
Population consists of all objects that possess a set of characteristics pertaining to a market research
problem

CENSUS covers the entire population

◦ Generates Parameter

SAMPLE – subset representation of the Population

◦ Generates Statistic

Sample Statistic helps make inferences about Population Parameter

Census can be undertaken when: Sampling can be undertaken when:

◦ Population size is small ◦ Population size is large; census is not possible

◦ All individuals or objects must be studied ◦ Limited resources, like budget and time
◦ Cost of incorrect decision is high ◦ Higher response quality is required
◦ Sampling errors are high ◦ Population is homogeneous
 Errors in Data Collections
Sample
Population Frame Respondent Interviewer

Ambiguous questions
Ambiguous answers
Sampling Interviewer Error
Incorrect Error Inaccurate response
Definition Inability to respond
Unwillingness to respond
Non-Response
(refusal, unavailable)
Sampling Frame
A list of items in the population of interest, such as, individuals, households, organisations; and
uses information from databases
◦ Includes individuals in the population
◦ Includes information that will enable contact with them
◦ Includes quality sources of data
EMPLOYED RESIDENTS OF XYZ CITY
Convenience Sampling helps make this frame as you go… S.no. NAME AGE OCCUPATION ADDRESS

Duration or Time available affect the Sampling Frame 1. Aarti Kumar 25 Designer #32, Sector…
2. Bhavesh Singh 28 Architect L21/5 Prakriti…
3. Chitranjan Verma 30 Teacher A103 Diamond …
4. Gurpreet Gulati 32 Doctor #92/2 District …
5. Monica Sarabhai 29 Analyst #23 Main Street…
A Sample is then drawn from this sampling frame 6. Shruti Sharma 26 Trader Luxe Villa RR Road…
7. Vedika Gupta 27 Banker 93A/3 Park…
…
Sampling Error
Total Error – difference between true value of the variable (population parameter) and the observed
value (sample statistic)

Sampling Error – when the difference between population parameter and sample statistic is only
because of sampling
◦ Sampling Errors may be reduced by increasing Sample Size

Non-Sampling Error – when the error in data collection and analysis is not due to sampling, e.g.:
error due to measurement, data recording, data analysis for the population
Sampling Process
Step 1: Define the target population – sample elements or units, area of coverage

Step 2: Determine the Sampling Frame – list of all randomly selected population members used to
obtain the sample

Step 3: Resolve differences (if any) in Population and Sampling Frame

Step 4: Select a Sampling Procedure

Sampling Process
Step 5: Determine the right Sample Size

Step 6: Obtain Data

Step 7: Deal with Non-response

Step 8: Generate Information for decision making

 Sampling
Techniques

Probabilistic Non-Probabilistic

Simple
Stratified Judgemental Snowball
Random

Cluster Systematic Convenience Quota

Multistage
Simple Random Sampling
Each population member has an equal probability of being selected resulting in an unbiased
representation of the population – lottery or draw, dice, coin, spinning wheel

Achieved by assigning sequential values and randomly

selecting those values

The resulting value of the dependent variable has

an error of +/- variance
Stratified Random Sampling
Random Sampling across various homogenous strata of the
group (defined to increase efficiency)
◦ Population is divided into smaller groups (“strata”) based on
shared characteristics
◦ Elements are randomly chosen from strata keeping in mind the
proportion of group size vs population
◦ Every population element must belong to only one strata –
there must be no overlap (mutually exclusive) to avoid
skewness

Used when the population is breakable into subgroups

DISPROPORTIONATE STRATIFIED
SAMPLING
Cluster Sampling
The population is divided into subgroups, clusters

Randomly selected clusters are studied in entirety, i.e.,

every member of the subgroup is interviewed (reduces
precision)

Useful when subgroups that are representative of the

population can be identified

Multi-Stage Sampling – iterative cluster sampling

 STRATIFIED CLUSTER
Homogeneity within subgroup Homogeneity between subgroups
Heterogeneity between subgroups Heterogeneity within subgroup
All subgroups are included Random selection of subgroups
Improved Sampling Efficiency Improved Sampling Efficiency
Accuracy over Cost Cost over Accuracy
Systematic Sampling
Systematic selection of every ith member on the list

The first selection is made randomly, and every subsequent ith member is included – this can
lead to a biased sample (accuracy reduces if the sample list is ordered – example increasing
revenue, days of the week)
Non-Probability Sampling Techniques
Judgement Sampling – expert judgement is used to identify sample based on biases (less costly,
small sample size, useful in cases such as early adopters for innovation product testing)

Snowball Sampling – each respondent is asked to identify one or more respondents

Convenience Sampling – selecting sampling units that are convenient to approach (often used for
pretesting)

Quota Sampling – Judgement sampling with additional selection constraints eliminating certain
biases (e.g.: 50% respondents must belong to xyz category)
Designing Survey Questionnaires
Translate research questions into information that is required

E.g., Research Question: What is the probable demand for a proposed product?
◦ Information Sought:

◦ Attitude and Awareness of the category

◦ Attitudes and Awareness of existing options in the market

◦ Benefits or Value sought from the category

Open-Ended Questions
Enable a wide range of answers with minimal influence from the researcher

Respondents appreciate the freedom, esp. in cases where given options do not reflect their attitude
Close-Ended Questions
5 or 7 response categories are ideal for discrimination among respondents

Certain numerical values are best grouped together, for instance, income, purchase frequency in a
given period of time

Open-ended questions followed by a close-ended question is a Probe

Scales
Nominal Scale – mutually exclusive labelled categories; no specified ordering or spacing (count is
analysed)

Ordinal Scale – categories are ranked or arranged in order; does not imply difference between levels
(middle and most frequent value is analysed)

Interval Scale – equal increment in attribute; location of zero is not fixed (mean is analysed)

Ratio Scale – values are relative to zero where zero means absence of variable
 Types of Scales
TYPE EXAMPLES CODING RULES APPLICATION ANALYSIS
“Yes” or “No”;
NOMINAL Identical or Classification (location,
“Employed” or Percentage; Chi-Square (mode)
(Dichotomous) Different objects social class)
“Unemployed”
ORDINAL Comparative Objects are greater Percentile; Median; ANOVA
Preference or Ranking
(Rank Order) rank order or smaller (difference in treatments)
Intervals are equal
Likert; Stapel; Mean; Standard Deviation; t-test;
– allows for Index number,
INTERVAL Semantic- ANOVA; Regression;
comparing Attitudes, Temperature
Differential Factor Analysis
difference in values
Meaningful zero –
Constant sum Sales; Incomes; Geometric and Harmonic Mean;
RATIO allows absolute
scale Units Produced; Costs Coefficient of Variation
comparison
Display Quality – Stapel Scale (Interval Scale)

-5 -4 -3 -2 -1 0 1 2 3 4 5

Likert Scale (Interval Scale)

Semantic-Differential Scale (Interval Scale)
Constant Sum Scale (Ratio Scale)
Framing the Questions
Simple and direct vocabulary that is familiar and comprehensible for the respondents

Avoid vague and ambiguous wording

Avoid double-barrelled questions – two different questions combined to form one question
◦ E.g.:
“ARE YOU SATISFIED WITH THE SERVICE AND PRICE?”;
“ARE YOU CURRENTLY WORKING OR LOOKING FOR A JOB?”
Framing the Questions
Avoid questions that suggest an answer or the researcher’s opinion
◦ May include possible answers; bias; negative or positive sentiments towards a subject

Frame straight-forward and easy-to-follow instructions

Questions must be framed in a manner such that they are applicable to all respondents
◦ Use Filter questions if needed

Questions must be short to avoid respondent fatigue

◦ General rule is to frame questions with 20 words or less
 The Right Question Order
ORDER TYPE PURPOSE EXAMPLE

Starting Qs Broad and General Break ice and establish rapport “Do you use social media apps?”

Reassure that the survey is easy “Which Social Media apps do you
Next Few Simple and Direct
to answer use frequently?”
“What makes you spend more time
Relate more to the subject and
Up to 1/3rd Focused on one social media app compared
convey more information
to the others?”
“Rank in order of importance the
Focused – some may Obtain most of the subject
Major Part social media characteristics that are
be hard to answer information
important to you.”
Personal Questions Obtain Classification and
Closing Qs “What is your household income?”
that may be sensitive Demographic information
Asking Sensitive Questions
Casual Approach

“Everybody” Approach – “As you know many people…”

“Other People” Approach – “Do you know someone who…”

Anonymous Responses

Kinsey Technique – straightforward and clear language

Asking Sensitive Questions
Randomised Response
Approach

◦ Respondents choose one of two

random questions (one sensitive,
the other not) with a coin flip; while
the researcher is unaware of which
question was answered
◦ The researcher would then estimate
the number of respondents that
answered the primary question
 Randomised Response Approach
Consider that the interviewer asked the following questions, and the respondent flips a coin (or uses the
last digit of their birth data being odd or even) to select which question to answer:

Q1: HAVE YOU EVER STOLEN SOMETHING WORTH INR 4000/- OR MORE? (SENSITIVE QUESTION)
Q2: IS THE LAST DIGIT OF YOUR BIRTH DATE 5? (NON-SENSITIVE OR INNOCUOUS QUESTION)
◦ After flipping a coin, 300 (out of 1000 total respondents) answered “Yes” – the researcher does not know
which question was answered and will determine the responses using proportions

◦ Prob (Choosing Q1) : Prob (Choosing Q2) = 50:50

◦ Therefore, 500 respondents (50%) answered Q1 and 500 answered Q2
◦ Only 10% of those who answered Q2 can answer “Yes”; i.e. 50 respondents answered “Yes” to Q2
◦ Thus, the remaining 250 responded “Yes” to Question 1; i.e., 50% respondents

p(yes) = p(Q1 is Chosen) * p(Yes to Q1) + p(Q2 is Chosen) * p(Yes to Q2)

0.30 = 0.50 * x + 0.50 * 0.10
Questionnaire Errors
Inaccurate response - lack of information, memory recall
◦ Aided-recall technique

Unwilling to respond accurately – privacy concerns, time, fatigue, social desirability

Interviewer error – questioning, probing, rapport

Fruitzone (B): Designing the Questionnaire
Projective Techniques – bring out unconscious personality dimensions (true feelings, beliefs,
attitudes, desires, motives, fears) with simple unstructured tasks
Mall intercept random sampling

Associations; Inkblot; Construction or Completion; Arrangement; Play

Top-of-the-mind recall is difficult to obtain via FGDs – respondents may

be unwilling to discuss choices in presence of other respondents

Such techniques can be used till no new information is obtained

WORD ASSOCIATION SENTENCE COMPLETION

HEALTHY DRINK… A person who drinks juice is…

FRESH… When I want a healthy drink, I buy…

TASTY… I think sugarcane juice is good because…

FRUIT… Quality juice can be bought at…

JUICE… Juice should be drunk by…

Results from the Projective Techniques
➢ Majority of the participants associate fruit juice with health
➢ Juice is preferred over fruits, for convenience
➢ 60% respondents associated packaged fruit juice with preservatives
➢ Juice consumption happens in retail outlets and juice shops
➢ Freshness is associated with fruits and juice
➢ Juice is a predominant form of consuming sugarcane – associations “tasty” and “energising”
are top-of-the-mind attributes
➢ Tropiviva is second amongst participants for brand recall
Fruitzone: Consultants’ Conclusions
Customer Classification must be based on Consumer Behaviour rather than only Demographics –
also describing product attributes that will attract the customer justifying product purchase decisions
◦ Based on Secondary Research: Information Conscious (IC); Health Conscious (HC); Price Conscious (PC);
Convenience Conscious (CC); and Novelty Tryers (NT)

Defines attributes that the customers values

Why customers buy a particular product

Difficult to measure; however, more powerful in determining purchase intent

Fruitzone (B): Designing the Questionnaire
CHECK: A) OBJECTIVE OF THE QUESTION; B) NEED FOR THE QUESTION; C) THE SENTENCE FRAMING; D) SCALES
USED; AND E) PLACEMENT OF THE QUESTION IN THE QUESTIONNAIRE

Question 5: Most Favourite Brand – “None” would make the list exhaustive

Question 6: Asks a Rating on the scale while the next question asks for Ranks – may lead to confused
and erroneous responses

Question 2: A respondent compared purchase locations like Railway Stations and Airports to retail
stores; and that they do not reflect day-to-day activities. Should this have ben dropped?
Thank You
APPENDIX
Research Things to Know…
Choice Modelling – constructing models that represent the process of consumer choices
Alpha Testing – done within the organisation; functionality and usability are tested
Beta Testing – done in the user’s environment; focuses on feedback, improvements, and
customer satisfaction
 Analysis Techniques

Chi-Square determine difference between expected and observed value

Median middle number separating the upper and lower half
ANOVA (difference in
compare means of different groups
treatments)
Standard Deviation extent to which the values differ from average
t-test compare means or proportions of two groups
Regression relationship between two or more variables
Factor Analysis reduces a set of variables into a smaller subset
Geometric Mean central tendency
Harmonic Mean reciprocal of average of reciprocals (balances weightage)
Coefficient of Variation extent of variability relative to the mean (ratio of SD:Mean)