Marketing
Research & Analytics
COURSE FACULTY : PREETI PRIYA
TEACHING ASSISTANCE: NISHMA SHAH & JIGAR PATEL
New Product Development Process
Concept Product
Idea Product
Screening/Idea Design & Product Launch
Generation Validation
Refinement Development
KANO TABLE
KANO ANALYSIS
New Product Development Process
Concept Product
Idea Product
Screening/Idea Design & Product Launch
Generation Validation
Refinement Development
Product Design Research
Marketing Experiments
Prod Mgmt Brainstorming
Consumer Clinics
Surveys
Quali Research
Idea Generation and Screening is followed by designing
and Crafting the product
◦ Combining art, science and technology to transform the concepts
into new products.
Product Design Research
AHP KANO Conjoint Analysis
• Multi attribute • Identifying • Multi attribute
utility/value categories of utility/value
measurement to features/attributes measurement to
decipher consumer that will drive decipher consumer
choices consumer value choices
Here…
Conjoint Analysis
Kano Analysis within Conjoint Analysis
Conjoint Analysis
Or
Conjoint Experimentation
Probably the most commonly used quantitative market research method.
Applied in 1970s by Paul Green in context of new product development process
Selecting Alternative Product Designs, Targeting and Pricing in context of NPD
Why consumers choose
one brand or product over another?
Which organisation
should I choose?
Consumers are usually unable to quantify how they would trade off one
feature for another, and when asked, many state that "everything is
important”.
Conjoint Analysis: Learning Goals for this
module
1. What is the basic underlying model in conjoint analysis?
2. What are the different stages in conjoint analytic New Product Design
Projects?
*************************************
1. How does conjoint work?
2. How do we analyze conjoint data using Excel and SPSS?
How should we develop physical well-
being of women in IRMA
A group of community women of IRMA come together and
discuss about what can be done about increasing their physical
fitness.
Yoga Classes
Yoga Classes
1. A Yoga Class in the morning hours with no fees
2. A Yoga Class in the evening hours with fees as per your wish
3. A Yoga Class in morning hours with a fixed monthly fees
4. A Yoga Class in the evening hours with no fees
5. A Yoga Class in the morning hours with fees as per your wish.
6. A Yoga Class in evening hours with a fixed monthly fees
When we spoke to a community member
regarding this…She suggests her preference in
the following order
1. A Yoga Class in the morning hours with no fees – 4th
2. A Yoga Class in the evening hours with fees as per your wish
– 1st
3. A Yoga Class in morning hours with a fixed monthly fees – 6th
4. A Yoga Class in the evening hours with no fees – 3rd
5. A Yoga Class in the morning hours with fees as per your wish.
– 5th
6. A Yoga Class in evening hours with a fixed monthly fees -2nd
Fees Structure
Rank
Voluntary Fixed No
Yoga Morning 5 6 4
Teacher Evening 1 2 3
With these ranks we assign utility points
(higher, the better)
Fees Structure
Score
Voluntary Fixed No
Yoga Morning 1 0 2
Teacher Evening 5 4 3
Fees Structure
Rank
Voluntary Fixed No
Yoga Morning 5 6 4
Teacher Evening 1 2 3
Utility assignment can yield average utility of
an attribute across all choices where it
appears
Fees Structure
Score
Voluntary Fixed No Average
Yoga Morning 1 0 2 1
Teacher Evening 5 4 3 4
Average 3 2 2.5
Yoga Teacher Fee Structure
Morning 1 Voluntary 3
Evening 4 Fixed 2
No Fees 2.5
This is the individual value system and it
recaptures the stated original ranking to a
large extent.
Total Value
Value Value System Stated
System System Score Original
Product Score Score Rank Rank
Morning + Voluntary 1+3 4 4 5
Morning + Fixed 1+2 3 6 6
Morning + No Fees 1+2.5 3.5 5 4
Evening + Voluntary 4+3 7 1 1
Evening +Fixed 4+2 6 3 2
Evening + No Fees 4+2.5 6.5 2 3
Conjoint Analysis:
The Underlying Model
A Product is a “bundle” of attributes. Attributes can take multiple
forms
Consumers evaluate the alternatives in the marketplace by examining
how much they offer on the various attributes and how critical each
attribute is to them
Total Value of product = sum of sub-values of its attribute levels to
the individual
A consumer prefers the product that delivers the greatest Total Value
to him/her
Decompose the total product value into the value of each sub-part in
order to determine preference for the composed product/service
ATTRIBUTES are features and characteristics that define various
aspects of a product or service
Fees Structure
Rank
Voluntary Fixed No
Yoga Morning 5 6 4
Teacher Evening 1 2 3
LEVELS are the values that each attribute may take
Fees Structure
Rank
Voluntary Fixed No
Yoga Morning 5 6 4
Teacher Evening 1 2 3
PROFILES are hypothetical products or services described as bundles
of attributes that are each set at a particular level
1. A Yoga Class in the morning hours
with no fees
2. A Yoga Class in the evening hours
with fees as per your wish
Fees Structure 3. A Yoga Class in morning hours
Rank
Voluntary Fixed No with a fixed monthly fees
4. A Yoga Class in the evening hours
Yoga Morning Mor + Vol Mor + Fix Morn+ No with fees with no fees
Teacher Evening Eve+Vol Eve+Fix Eve+No
5. A Yoga Class in the morning hours
with fees as per your wish.
6. A Yoga Class in evening hours with
a fixed monthly fees
PARTWORTHS are the utility of each attribute level. They are called partworths because they capture
how much each PART of the product is WORTH to the consumer. UTILS are the units in which
PARTWORTHS are measured.
A morning yoga teacher has only 1util partworth,
whereas the evening teacher has 4 utils.
Fees Structure
Score
Voluntary Fixed No Average
Yoga Morning 1 0 2 1
Teacher Evening 5 4 3 4
Average 3 2 2.5
Yoga Teacher Fee Structure
Morning 1 Voluntary 3
Evening 4 Fixed 2
No Fees 2.5
UTILITY is a measure of the value provided to a consumer. For example, the utility of a specific profile
is the overall value that the bundle of attributes in the profile provides the consumer.
Total Value
Value Value System Stated
System System Score Original
Product Score Score Rank Rank
Morning + Voluntary 1+3 4 4 5
Morning + Fixed 1+2 3 6 6
Morning + No Fees 1+2.5 3.5 5 4
Evening + Voluntary 4+3 7 1 1
Evening +Fixed 4+2 6 2 2
Evening + No Fees 4+2.5 6.5 3 3
Conjoint Analysis
allows marketers to quantify consumer preferences for various
features and characteristics of products or services. It essentially
builds a mathematical model of consumer preferences, thus
allowing managers to predict how consumers would choose
between products and services that are or may become available in
the market.
Questions for which conjoint analysis is
relevant to address include:
Which features should we include in our products/services?
How many different products/services should we offer?
How much are consumers willing to pay for each feature of our
product/service? How much would they be willing to pay extra for an
improvement on an existing characteristic?
How should we price our products/services?
What relative market share/penetration should we expect to obtain if we
launch product/service X and not product/service Y?
How do consumers differ in their preferences? Which segment(s) of
consumers should we focus on serving? What product(s) would appeal to this
(these) segment(s) relative to existing alternatives in the marketplace?
Conducting a Conjoint
Experimentation
DECISION STAGES IN CONJOINT STUDY
Stages in Designing Conjoint Experiment
Estimate
Build Parameters
Model Partworths
Collect
Data
Create
Product
Choose
Profiles
relevant
Choose
levels
Attributes
U( Product) = u( A Level of Att1) + u(A Level Att2) + u(A Level Att3) + u(A Level Attn)
Stages in Designing Conjoint Experiment
.
• Use of Qualitative Research and Kano Analysis to identify the relevant
attributes
• “Table Stake”/Must Be attributes X Create
Choose Product
• Performance Attributes √ Profiles
relevant
• Delighter/Attractive Attributes √ Choose
levels
Attributes
• Avoid too many attributes . Keep it in the range of 4-6 attributes.
• Choice of attributes in a way that can help managers take decision on
product configuration and/or price. Attributes which can sway
consumer choice and also firm can control them!!
• Attributes with infeasible combinations –(Vegetarian(Y/N) and
Main Course (Chicken /Fish/Paneer/Green Veg)
• Attributes with ambiguous or subjective levels ( Levels like
Stylish/Aesthetic..)
Stages in Designing Conjoint Experiment
• Conjoint
. can accommodate any practical number of levels for an
attribute
• Best Practice is to keep approximately the same number of levels and
to limit the number of levels to 2-4 so as not to overload participants
Create
in the study. Product
Choose
• Be realistic in choosing levels. relevant Profiles
Choose
• Meaningful to consumers and controllable for the firm. Attributes levels
• Avoid “confounded” levels. ( Round or Square)
• Ensure that you are not comparing apple and oranges
• Avoid “Non- Mutually” Exclusive levels. ( Rs 150 per month/Rs 1200
per year)
• Distinguish between categorical and continuous attribute. ( Price,
Mileage, etc. – continuous attribute)
Lets talk about something
that is currently a concern
for us
Preference Filling for SIS
project
Designing SIS for oneself!
Attribute 1 Attribute 2 Attribute 3 Attribute 4
Team Functional Area Base Location Stipend
Solo Marketing Home Town Rs 10,000 per month
Team of 2 Social/Development No Home Town Rs 20,000 per month
Team of 3 or more Others
Stages in Designing Conjoint Experiment
.
• construct product profiles that consumers will
evaluate. It is important that these product
profiles represent offerings that may be Create
Product
available on the market or that don't seem Choose
relevant Profiles
Choose
infeasible or unrealistic to offer. Attributes levels
• The number of attributes and levels chosen
determines how many possible profiles can be
constructed.
Selecting the Best Profiles
Create
Choose Product
relevant Profiles
Choose
Attributes levels
Max No. of Profiles = No. of LevelsAttribute 1 * No. of LevelsAttribute 2 *... * No. of LevelsAttribute N
Designing SIS
In a full design we will have
3X3X2X2=36 profiles
Data Collection - preference measurement from
participants on 36 profiles is cumbersome!
We need to cut down on no. of profiles to be evaluated by
consumers…. But HOW?
Selecting the Best Profiles
It is important to note that using any random set of profiles
may lead to unusable data.
Selecting the Best Profiles
The idea is to select profiles in which the attributes have
enough variations to be able to quantify the link between
attributes and preferences and therefore capture
important tradeoffs.
Selecting the Best Profiles
The idea is to select profiles in which the attributes have
enough variations to be able to quantify the link between
attributes and preferences and capture important
tradeoffs.
Orthogonal Designs
In an orthogonal design, the levels of the features are
chosen such that, for each pair of features, say a and b, the
high level a appears equally often in profiles that have a
high level b as in profiles that have a low level of b, and vice
versa.
Profiles that get created are uncorrelated with each other.
Programmes available to create such an orthogonal design –
SPSS has its own procedure.
Create DIS Profiles
Stages in Designing Conjoint Experiment
• Preference/Purchase Intention rating is the most
preferred when OLS regression based model is
used. Estimate
• Rating on a 5 point or a 7 point scale
Parameters
Build Model
Partworths
Collect
• Preference ordinal ranking is also used for OLS Create
Product Data
Choose
regression Choose
relevant
levels
Profiles
Attributes
• Ranking in steps – most preferred, in-between
and least preferred
• Ranking within these categories
Illustration of a Conjoint Questionnaire
Definitely
1 2 3 4 5 Definitely
not apply
apply
Project #1
No of Students: 1 Your Rating
Functional Area: Marketing
Base Location : Anand
Stipend: Rs 20000 per month
Stages in Designing Conjoint Experiment
Estimate
Build Parameters
Model Partworths
Collect
Data
Create
Product • Build The model using Dummy
Choose Variable Regression Modeling in
Profiles
relevant Excel or in SPSS using SPSS Conjoint
Choose
levels Syntax
Attributes
Model Building and Calibration
For ratings data, simple regression can be used to compute the part-worths for
the attribute levels.
Regression Modeling on Conjoint Data
Attribute levels – Independent Variables
Preference/Purchase Intention Rating - Dependent Variable
Independent Variable in the form of Dummy Variables
Data Organisation for Regression
Modeling
Dummy Coding
◦ 1 indicates presence of feature
◦ 0 indicates absence of feature
To resolve the problem of linear dependency in regression
modeling, we create dummy variables one less than
number of levels
One can drop any one level from each of the attributes- no
rules for that
Dummy Coding
Number of Dummy Variables for an attribute=n-1
◦ Members 3 levels 2 variables
◦ Func. Area 3 levels 2 variables
◦ Base Loc. 2 levels 1 variable
◦ Stipend 2 levels 1 variable
Regression Modeling on Conjoint Data –
Create a “baseline” profile
Combination of dropped Card ATT 1a ATT 1b ATT 2a ATT 2b ATT 3 ATT 4 Willingness to Apply
variables form the baseline 1 0 0 0 1 0 0 2
profile 2 0 1 0 0 1 0 2
3 1 0 0 1 1 1 5
◦ e.g., “Team of 3 or more,” 4 1 0 0 0 0 0 2
“Others,” “No Home Town,” 5 0 0 0 0 1 1 2
“Rs 10 K” 6 1 0 1 0 1 0 5
7 0 1 0 1 1 0 4
◦ Partworths for these levels 8 0 0 1 0 1 0 4
set to 0 9 0 1 1 0 0 1 4
◦ Partworths of other levels =
deviations from this baseline
profile
What is the total utility of a project in domain other than
Marketing/Social, requires 3 people to work outside home
location with a stipend of Rs 10,000 pm?
Coefficients
Intercept 0.50
ATT 1a 1.33
ATT 1b 0.67
ATT 2a 2.33
ATT 2b 1.67
ATT 3 1.00
ATT 4 0.50
Simulations – Team of 2 + Social + Home Town +
Rs 20 k ? Will he/she apply?
What do you make of this?.............
Partworth Plots What is this participants ideal/dream SIS Project ?
We can predict choice - SIMULATIONS
A- Team of 2 | Social | Home Town | RS 20 k
B- Solo | Marketing | Non Home Town | Rs 10 K
How willing he/she would be to apply for these projects?
Which project will be preferred by him/her more?
◦ Max Utility Method
◦ Logit Preference Share
How important is each attribute
For each attribute:
◦ Range of an attribute = max partworth – min partworth
◦ Importance of an attribute = Range / (sum of ranges across all attributes)
◦ Increasing the width between the levels will tend to increase the importance of the attribute
Attribute Importance
Team 25.8
Domain 45.2
Location 19.4
Stipend 9.7
CONJOINT ANALYSIS
USING SPSS
Advantages over traditional excel based
CA
Individual and aggregate analysis possible
Design creation more robust
More outputs
Outputs unravel more insights
Simulations evaluation
Validity of Model can be Checked
CONJOINT PROCEDURE in SPSS
Writing & Running
Orthogonal
Data Organisation CONJOINT
Design
SYNTAX
Generating an Orthogonal Design
More accurate procedure than what we saw in case of
experiment design builder using excel.
Allows for inclusion of hold out cases to check validity and
reliability of the model
Also additionally one can include simulation profiles
Displaying an Orthogonal design
Listing for Experimenters
Profiles for Respondents
Organising Data
Depending upon type of data collected – RANK or RATING
Ranking ~ RANK
◦ Datapoint – rank for a profile
Rating ~ SCORE
◦ Data point – Purchase Intention score/rating for a profile
Organising Data – Effect Coding in SPSS (
variation of Dummy Coding
Var1 Var2
Solo 1 0
Team 2 0 1
Team3 -1 -1
Property of effect coding is
sum of B coefficients of all categories =0
So, if there are three categories – Solo, Team 2, Team 3
We create 2 variables – solo and team 2
Sum of B coefficients = 0
Hence one can derive the individual part worth of the base levels as well.
Running Conjoint Analysis
SPSS Syntax- RUN THE [Link] SYNTAX
Data Requirements
◦ File with design – PLAN file ([Link])
◦ Consumer Preference/ranking/rating file – DATA File ([Link])
Running Conjoint Analysis - Syntax
Specify the PLAN and DATA File
Specify the way data was recorded
Specify the Subject
Specify the Factors ( Discrete/Linear)
Specify the Outputs
PRINT | Plot | Utility
Outputs
Individual subject wise
Aggregated – Overall
Partworths
Partworth Importance
Correlations
Check the file DIS SPSS Output. spv
Simulations – Preference Shares
Simulations- Preference Share
Logit Utility
Simulations- Preference Share
BTL
determines the probability as the ratio of a profile’s utility to
that for all simulation profiles, averaged across all
respondents.
