Professional Documents
Culture Documents
SESSION 4
How to discover consumer
preferences?
Dr. Xi Chen
18 Nov. 2014
From last lecture: dummy coding
coding is important!
m-1 dummy variables for a nominal variable with
m levels
interpretation is tricky!
always relative to the baseline level
From last lecture: moderation
X Y
From last lecture: mediation
1 = + +
must be significant
2 = + +
must be significant
3 = + + + must be insignificant or
less in magnitude
M
X Y
X
Y
Todays Agenda
Conjoint Analysis
1 hour 45 mins
what, why and how
Coffee Break
15 mins
SPSS Illustration
40 mins
A MOTIVATING EXAMPLE
Considering yourself purchasing a laptop
Tradeoff between options
Option A Option B
3.5 GHz Processor 2.5 GHz Processor
2 GB RAM 4 GB RAM
A product is a combination of attributes
Profile 1 Profile 2
Manager = Chef
Attributes = Ingredients
Question: HOW?
Breaking down the problem
Intuitive way:
Ask about different attributes one-by-one
How much would you pay for ?
Rate the importance of brand: ____
Profile 1 Profile 2
Conjoint Strategy
Analysis Insights
Attributes Products
CONJOINT ANALYSIS:
Some Terminologies
Utility
Two attributes:
1 : speed of processor
2 : size of memory
: overall utility
= 1 1 + 2 2
Option A Option B
: part-worth or importance
3.5 GHz Processor 2.5 GHz Processor
2 GB RAM 4 GB RAM
Revealed & stated preference
Revealed preference:
consumers are rational by
choosing an option with higher
utility
Stated preference:
consumer can take a good
guess of utilities offered by
Option A Option B options
3.5 GHz Processor 2.5 GHz Processor
2 GB RAM 4 GB RAM
CONJOINT ANALYSIS:
A CASE STUDY
Some background info
Respondents:
Shown these 32 profiles
Asked to rate these profiles using 0-100 point scale
Preference ratings from one respondent
IVs
DV
Dummy variable regression
Dummy coding
Estimation of part-worth & importance
Utility contribution
of different levels
Relative importance
of attributes
Takeaways from the case
Attribute 1 Attribute 2
1 1
1 2
1 3 Attribute 1 Attribute 2
2 1 1 3
2 2 2 2
2 3 3 1
3 1
3 2
3 3
2 constructing profiles
Final note
The set of cards is not always the same. Even though the
set of cards is different, the estimation results will be
similar. So it doesnt matter which set of cards you use!
3 data analysis
3 data analysis
SUBJECT PROFILE ATT#1 ATT#2 ATT#3 RATING
1 1 3 3 1 5
1 2 1 2 3 7
1 3 3 1 3 4
1 4 1 3 2 6
1 5 2 3 3 2
1 6 3 2 2 6
1 7 2 2 1 6
1 8 2 1 2 2
1 9 1 1 1 5
2 1 3 3 1 3
2 2 1 2 3 6
2 3 3 1 3 4
2 4 1 3 2 7
2 5 2 3 3 10
2 6 3 2 2 orthogonal
9
2 7 2 2 1 4
2 8 2 1 2 design
4
2 9 1 1 1 10
3
3
1
2
3
1
3
2
1
3
generated
3
5
3 3 3 1 3 5
3 4 1 3 2 2
3 5 2 3 3 3
3 6 3 2 2 2
3 7 2 2 1 3
3 8 2 1 2 2
3 9 1 1 1 6
4 1 3 3 1 5
3 data analysis
SUBJECT PROFILE ATT#1 ATT#2 ATT#3 RATING
1 1 3 3 1 5
1 2 1 2 3 7
1 3 3 1 3 4
1 4 1 3 2 6
1 5 2 3 3 2
1 6 3 2 2 6
1 7 2 2 1 6
1 8 2 1 2 2
1 9 1 1 1 5
2 1 3 3 1 3
2 2 1 2 3 6
2 3 3 1 3 4
2 4 1 3 2 7
2 5 2 3 3 10 one
2 6 3 2 2 9
2 7 2 2 1 subject
4
2 8 2 1 2 4
2
3
9
1
1
3
1
3
1
1
will rate all
10
3
3
3
2
3
1
3
2
1
3
3
9 profiles!
5
5
3 4 1 3 2 2
3 5 2 3 3 3
3 6 3 2 2 2
3 7 2 2 1 3
3 8 2 1 2 2
3 9 1 1 1 6
4 1 3 3 1 5
3 data analysis
SUBJECT PROFILE ATT#1 ATT#2 ATT#3 RATING
1 1 3 3 1 5
1 2 1 2 3 7
1 3 3 1 3 4
1 4 1 3 2 6
1 5 2 3 3 2
1 6 3 2 2 6
1 7 2 2 1 6
1 8 2 1 2 2
1 9 1 1 1 5
2 1 3 3 1 3
2 2 1 2 3 6
2 3 3 1 3 4
2 4 1 3 2 total # obs
7
2 5 2 3 3 10
2 6 3 2 2 9 =
2 7 2 2 1 4
2 8 2 1 2 # subjects
4
2 9 1 1 1 10
3
3
1
2
3
1
3
2
1
3
3
5
x
3
3
3
4
3
1
1
3
3
2
# profiles
5
2
3 5 2 3 3 3
3 6 3 2 2 2
3 7 2 2 1 3
3 8 2 1 2 2
3 9 1 1 1 6
4 1 3 3 1 5
3 data analysis
: overall utility/rating of
profile k by respondent p = +
=1 =1
Variable Part-worth
(Constant) 4.222
Information: Personal Advisor 1.000
Information: Internet -.333
Information: Brick & Mortar .000
Acquisition: Personal Advisor 1.000 Most preferred
Acquisition: Internet .667
Acquisition: Brick & Mortar .000
Price: Low 2.333
Price: Recommended 1.333
Price: Above .000
4 making use of results
Importance of
attributes Importance of an attribute is defined as the
range of part-worth across the its levels
Variable Part-worth
(Constant) 4.222 = ( )
Information: Personal Advisor 1.000
Information: Internet -.333
Information: Brick & Mortar .000
Information = 1 .333 = 1.333
Acquisition: Personal Advisor 1.000
Acquisition: Internet .667
Acquisition = 1 0 = 1
Acquisition: Brick & Mortar .000 Price = 2.333 0 = 2.333
Price: Low 2.333
Price: Recommended 1.333
Price: Above .000
4 making use of results
Importance of attributes
Variable Part- Importance
worth normalized by sum of importance of all
Information: 1.000 attributes
Personal Advisor
= , = 1
=1
Information: Internet -.333 1.333
=1
Information: .000
Brick & Mortar
1.333
Acquisition: 1.000 Information = = .286
Personal Advisor 1.333 + 1 + 2.333
Acquisition: Internet .667 1.000 1
Acquistion = = .214
Acquisition: .000 1.333 + 1 + 2.333
Brick & Mortar 2.333
Price = = .500
Price: Low 2.333 1.333 + 1 + 2.333
Price: Recommended 1.333 2.333
Price: Above .000
4 making use of results
information sources
1. through the internet (companys website)
2. from an expert / advisor (in home)
3. in brick & mortar stores
purchase product
1. through the internet (companys website)
2. from an expert / advisor (in home)
3. in brick & mortar stores
price
1. below recommended price
2. recommended price ( 3,000)
3. above recommended price
1 determining attributes & levels
2 constructing profiles
3 data analysis
conjointdata.sav
a respondent features in
multiple rows because s/he
evaluates multiple stimuli,
and rates each stimulus
3 data analysis
Coding Dummies
DV > rating
IVs > dummy variables
(for an attribute with K levels, you make K-1 dummies)
3 data analysis
run a regression with the rating as DV (I am using fire blanket
ratings) and the 6 dummies as IVs
3 data analysis
model fit!
model
significant!
no effect of
information
channel!
3 data analysis
4 making use of results