A simple tutorial to show conjoint analysis and cluster analysis. please send your feedback, this version is still rough and I would like to iteratively improve it so it is useful for most.
Investment in The Coconut Industry by Nancy Cheruiyot
A Simple Tutorial on Conjoint and Cluster Analysis
1. Marketing Research
Rags Srinivasan
Customer Segmentation and
Market Share Estimation With
Conjoint Analysis
2. Marketing is about segmentation and
targeting
Nothing more strategic than segmentation
Cannot treat the Value proposition is Target them
differently – SKUs,
whole market as different for each
one segment messaging
Rags Srinivasan IterativePath.com
3. What defines a segment?
Internally homogenous, externally
heterogeneous
4. Is your segmentation valid?
Meaningful, relevant and
Not too small, Not too
intuitively identified by
large
constituent variables
Rags Srinivasan IterativePath.com
5. Conjoint analysis helps you with the
clustering
Premise: The whole is the sum of its parts.
We can infer the relative importance of
parts from the customer preference of the
whole.
Rags Srinivasan IterativePath.com
6. For Example
Assign a value between 1 and 100 to these options. 100 means most
likeable and 1 means least likeable
Price: $2499 Price: $799 Price: $1999
Screen: 50” Screen: 42” Screen: 42”
Display: LCD Display: Plasma Display: LCD
Rags Srinivasan IterativePath.com
8. … and relative importance of attributes
What is the utility value a customer assigns to each
attribute?
Rags Srinivasan IterativePath.com
9. But you cannot ask customers about every
combination
Use commercial software to generate a manageable set of profiles
Conjoin
Manageabl
t
Attributes Levels e set of Survey
Softwar
profiles
e
Rags Srinivasan IterativePath.com
10. Let Us Walk Through An Example: My Work
On Airline Unbundled Pricing
Questions: How much do airline customers value services like free-
baggage, free drinks etc? Are airlines better off increasing ticket price
instead of unbundling pricing?
SFO JFK
11. With Following Options …
3 Airlines 2 Price Extras for Baggage, Pillows and Soft-
levels: $275, drinks
$250
12. Created 8 Profiles For Measuring Customer
Utility
Brand: 3 levels A manageable set of 8
Price: 2 levels profiles that stand-in for
Software
Baggage Fees: 2 levels all variable
Pillow Fees: 2 levels combinations
Drink Fee: 2 levels
13. Survey customers to find their utility value
for each profile
Rate your likelihood of choosing the option on a scale of 1 – 10 ( 8 profiles)
14. Model: Utility = f(Brand, Price,Fees)
Write customer utility (their likelihood of picking the airline) as a linear
function of these variables
U = Constant + b1 * JetBlue + b2* Delta + b3* Price$275 +
b4* BaggageFee$20 + b5 * PillowFee$4 +b6 * DrinkFee$2
JetBlue and Delta are mutually exclusive – 1 or 0
AA is implicitly defined when both JetBlue and Delta are 0
Price$275 = 1 means price is $275 , if it is 0 the price is $250
So on and so forth
b1, b2, … are the regression coefficients that are the relative
utilities of attributes that we seek to find
15. Use SPSS to indentify clusters
This margin is too narrow to contain it. Stay tuned I will add a
Camtasia demo of using SPSS to do Cluster analysis and
Regression.
Rags Srinivasan IterativePath.com
16. Run multiple regression for each cluster to
find the coffecients
U = 8.36 + 0.88 * JetBlue – 0.06 * Delta – 1.9 * Price$275 –
If we did not
2.41 * BaggageFee$20 – 0.83 * PillowFee$4 – 0.79 *
cluster
DrinkFee$2
U = 7.9 + 1.28 * JetBlue – 0.16 * Delta – 2.34 * Price$275 –
Cluster 1
3.14 * BaggageFee$20 – 0.92* PillowFee$4 – 0.87 *
DrinkFee$2
U = 8.6 + 0.4 * JetBlue + 0.17 * Delta – 1.24 * Price$275 –
Cluster 2
1.68 * BaggageFee$20 – 0.63* PillowFee$4 – 0.58 *
DrinkFee$2
Rags Srinivasan IterativePath.com
17. You can see the difference between two
clusters
JetBlue, $250, Baggage Fee $20, Pillow Fee $4, Drink Fee $2
Cluster 1 Cluster 2
JetBlue 9.18 9
$250 0 0
Baggage Fee $20 -3.14 -1.68
Pillow Fee $4 -0.92 -0.63
Drink Fee $2 -0.87 -0.58
Total Utility 4.25 6.11
18. Compute market share from the
utility values of the brands
Market
Utility
Share
of Product i
of Product i
Ui
MS i
U1 U 2 U 3
Feb 11, 2009
19. The net of this is
When you want to segment customers and target them with
multiple SKUs you need to do cluster analysis
Conjoint analysis gets you there and more