Conjoint Analysis - Part 2/3
Download as PPTX, PDF9 likes8,333 views
The document outlines a class on conjoint analysis, including demonstrations of different conjoint methods, an exercise in conducting conjoint analysis in Excel, and using needs-based market segmentation from conjoint results. Examples are provided of rank-based, choice-based, and pair-wise conjoint, along with exercises in dummy coding attributes, estimating part-worths from regression, and interpreting conjoint output. The exercises help illustrate how conjoint analysis can be used to understand customer preferences and segment markets.
Conjoint Analysis - Part 2/3
- 1. Class Outline • Conjoint Demonstration • Conjoint Analysis using Excel • Needs-Based Market Segmentation
- 3. More examples • 1. Rank-based conjoint – http://www.sawtoothsoftware.com/products/ssiweb/ssiwebsamples • 2. ACA Choice Based conjoint – http://www.sawtoothsoftware.com/products/ssiweb/ssiwebsamples • 3. Pair-wise “choices” – http://www.dobney.com/Conjoint/CnjtDemo.htm Conjoint Demos
- 4. Conjoint Analysis using Excel
- 5. Group Exercise – Rating Based Conjoint • 10=Most Preferred; 1=Least Preferred Brand Memory Price Rating Apple 20 150 10 Apple 10 150 8 Apple 20 200 6 Samsung 20 150 7 Apple 10 200 4 Samsung 10 150 5 Samsung 20 200 3 Samsung 10 200 1
- 6. Group Exercise #1 - WTP • $50 = ( 4 ) units • Apple > Samsung ( 3 ) units = $( 37.5 ) • 20GB > 10GB ( 2 ) units = $( 25 ) • Most important attribute: ( Price ) • Least important attribute: ( Memory )
- 7. Group Exercise #2 - WTP • 10=Most Preferred; 1=Least Preferred Brand Camera Price Rating Samsung Yes 150 10 Samsung Yes 200 6 Samsung No 150 7 Samsung No 200 3 LG Yes 150 8 LG Yes 200 4 LG No 150 5 LG No 200 1
- 8. Group Exercise #2 - Question • $50 = ( • Brand: ( ) Rating )>( ) ( ) Util. = $( ) • Camera: ( )> ( ) ( ) Util. = $( • Most important attribute: ( • Least important attribute: ( ) ) )
- 9. Group Exercise #2 - Solution • $50 = ( 4 ) Rating • Brand: (Samsung) > (LG) ( 2 ) Util. = $( 25 ) • Camera: ( Yes )> ( No ) ( 3 ) Util. = $(37.5) • Most important attribute: (Price) • Least important attribute: (Brand)
- 12. Brand DumApple Memory Apple 1 20 Apple 1 10 Apple 1 20 Samsung 0 20 Apple 1 10 Samsung 0 10 Samsung 0 20 Samsung 0 10 DumM20 1 0 1 1 0 0 1 0 • DumApple = 1, Brand=Apple = 0, otherwise • DumM20 = 1, Memory = 20 = 0, otherwise • DumP150 = 1, Price = 150 = 0, otherwise Price 150 150 200 150 200 150 200 200 DumP150 1 1 0 1 0 1 0 0 Baseline category: Samsung Baseline category: 10 GB Baseline category: $200
- 13. Coeff Standard Error t Stat Intercept 1 1.677E-16 5.964E+15 DumApple 3 1.677E-16 1.789E+16 DumM20 2 1.677E-16 1.193E+16 DumP150 4 1.677E-16 2.385E+16 P-value L95% U95% 4.744E-63 1 1 5.857E-65 3 3 2.965E-64 2 2 1.853E-65 4 4 Rating = 1 + 3 * DumApple + 2 * DumM20 + 4 * DumP150 Respondent1’s Rating 10 8 6 7 4 5 3 1 Brand Apple Apple Apple Samsung Apple Samsung Samsung Samsung Memory 20 10 20 20 10 10 20 10 Price 150 150 200 150 200 150 200 200
- 14. Estimating Regression Coefficients Using “linest” function 1. Select 4 cells
- 15. Estimating Regression Coefficients Using “linest” function 2. =transpose((linest(B3:B10,C3:E10,1,0)) Dependent Independent Variable Variable 3. Ctrl + Shift + Enter (array formula)
- 16. Estimating Regression Coefficients Using “linest” function DumP150 DumM20 DumApple Intercept
- 17. Group Exercise #3 - Estimation • Download the excel file: “Conjoint1.xlsx” • Q1. Based on Brand, Memory, Price descriptions, finish dummy coding on “Sheet1.” • Q2. On “Sheet2”, estimate partworths of respondent 1 and respondent 2. • Q3. Make partworth tables for respondent 1 and 2. • Q4. Describe how these two individuals are different in their preference.
- 18. Group Exercise #3 - Solution • Q1. Refer to the coding in “Sheet2” • Q2. Partworth estimation Variable Respondent 1 Variable Respondent 1 DumP150 3.75 DumP150 2.00 DumM20 0.75 DumM20 2.00 DumApple 3.25 DumApple 2.00 Intercept 0.00 Intercept 0.00
- 19. Group Exercise #3 - Solution • Q3. Partworth tables Respondent 1) Brand Memory Price Apple 3.25 10 GB 0.0 $200 0.0 Samsung 0.0 20 GB 0.75 $150 3.75 Respondent 2) Brand Memory Price Apple 10 GB 0.0 $200 0.0 Samsung • 2.00 0.0 20 GB 2.00 $150 2.00 Q4) • Respondent 1: Price-sensitive, brand-loyal, does not value memory • Respondent 2: equally values three attributes; Less brand loyal than respondent 1, Values memory more than respondent 1
- 20. Group Exercise #4: Partworth Table Interpretation • From regression analysis of rating-based conjoint, we get a “partworth” for each level of each attribute: User interface Brand Price Touch screen 9.2 Apple 59.5 CAD199 40.1 Keyboard 0.0 Blackberry 20.5 CAD299 30.1 Samsung 30.5 CAD399 20.1 Motorola 0.0 CAD499 0.0 • Q1) Calculate utility of the following profile. User interface Profile A Brand Price Utility Touch screen Blackberry CAD299 ?
- 21. Group Exercise #4: Partworth Table Interpretation • From regression analysis of rating-based conjoint, we get a “partworth” for each level of each attribute: User interface Brand Price Touch screen 9.2 Apple 59.5 CAD199 40.1 Keyboard 0.0 Blackberry 20.5 CAD299 30.1 Samsung 30.5 CAD399 20.1 Motorola 0.0 CAD499 0.0 • Q1) Calculate utility of the following profile. User interface Profile A Brand Price Utility Touch screen Blackberry CAD299 ?
- 22. Group Exercise #4 – Q1 Solution • “Utility” is a measure of preference. Higher the utility, greater the likelihood of purchase. • Given a product “profile” described in terms of the chosen attributes and attribute levels, we can compute its utility simply by adding up the partworths for the various elements of the profile • Solution User interface Profile A Brand Price Utility Touch screen Blackberry CAD299 ? – Utility of Profile A = 9.2 + 20.5 + 30.1 = 59.8 utils
- 23. Group Exercise #4: Q2 Choice Prediction • From regression analysis of rating-based conjoint, we get a “partworth” for each level of each attribute: User interface Brand Price Touch screen Apple 59.5 CAD199 40.1 Keyboard 0.0 Blackberry 20.5 CAD299 30.1 Samsung 30.5 CAD399 20.1 Motorola • 9.2 0.0 CAD499 0.0 Q2) Suppose you are given the following three profiles. Given the three alternatives, which product is the consumer most likely to choose? User interface Brand Price Profile C Touch screen Blackberry CAD299 Profile D Keyboard Apple CAD399 Profile E Touch screen Samsung CAD199 Utility
- 24. Group Exercise #4: Q2 Solution • Suppose you are given the following three profiles. Given the three alternatives, which product is the consumer most likely to choose? User interface Price Utility Profile C Touch screen Blackberry CAD299 59.8 Profile D Keyboard Apple CAD399 79.6 Profile E • Brand Touch screen Samsung CAD199 79.8 Profile E has the highest utility and so is the most likely to be chosen. Profile D is the next most likely and Profile C is the least likely.
- 25. Group Exercise #4: Q3 Willingness-to-pay • From regression analysis of rating-based conjoint, we get a “partworth” for each level of each attribute: User interface Brand Price Touch screen 9.2 Apple 59.5 CAD199 40.1 Keyboard 0.0 Blackberry 20.5 CAD299 30.1 Samsung 30.5 CAD399 20.1 Motorola 0.0 CAD499 0.0 • The WTP for a certain attribute value change (within a certain attribute) is defined as the maximum dollar amount that the customer will pay for the change (provided the change represents an improvement). • Q3) what is the WTP for the change from Brand=Samsung to Brand=Apple?
- 26. Group Exercise #4: Q3 Solution • The change represents an incremental value of 59.5-30.5=29.0 utils WTP = 29 x Dollar value of 1 util • The dollar value per util can be inferred by looking at the extremes of the partworths for the price attribute: (40.1 – 0.0) utils = $499 - $199 1 util = $300/40.1 = $7.481297 • Putting it all together WTP = 29 x Dollar value of 1 util = 29 x $7.481297 = $216.9576 • If the profile with the more preferred attribute level is priced exactly WTP amount dollars higher, then the customer is exactly indifferent about the change in that attribute level.
- 27. Group Exercise #4: Q4 Assessing Attribute Importance • From regression analysis of rating-based conjoint, we get a “partworth” for each level of each attribute: User interface Brand Price Touch screen 9.2 Apple 59.5 CAD199 40.1 Keyboard 0.0 Blackberry 20.5 CAD299 30.1 Samsung 30.5 CAD399 20.1 Motorola 0.0 CAD499 0.0 • Q4) Calculate relative importance of each attribute in %.
- 28. Group Exercise #4: Q4 Solution • In conjoint analysis, the “importance” of each attribute is computed via the following 2 steps. • Step 1: For each attribute, look at the partworths for all the attribute levels within that attribute and compute the difference between the largest and the smallest. This difference is the “partworth range” for that attribute. • In our example, the partworth ranges are: Attribute Partworth Range User interface 9.2 – 0.0 = 9.2 Brand 59.5 – 0.0 = 59.5 Price 40.1 – 0.0 = 40.1
- 29. Group Exercise #4: Q4 Solution • Step 2: The importance (weights) are just partworth ranges normalized to add up to 1. To normalize, compute the sum of the partworth ranges. Divide each partworth ranges by the sum. • In our example, the sum of the partworth ranges is: 9.2 + 59.5 + 40.1 = 108.8. • So the importance for each attribute is: Attribute Importance User interface 9.2/108.8 =0.0846 Brand 59.5/108.8 =0.5469 Price 40.1/108.8 =0.3685
- 30. Segmentation Using Conjoint Analysis
- 31. Segmentation Using Conjoint Analysis • A market segment is a sub-set of a market made up of people with one or more characteristics that cause them to demand similar product or services • Why?
- 32. Consumer A • $50 = 4 Units $12.5 = 1 Unit • Apple > Samsung 4 Units = $50 • 20GB > 10GB 1 Units = $12.5 Consumer Consumer E F Consumer Consumer C D Consumer B = 2 Units • $502 Units • $502 Units • $502 Units = = • $50 =Units • $50 = 2 $251 Unit $251 Unit Unit = =1 $251 Unit = 1 Unit $25 Samsung > Applg = = $25 Samsung > Apple • • • SamsungApple • SamsungApple > > Applg • Samsung > 0.5 Units$12.0 0.5 Units$12.0 = = $12.2 0.5 Units$12.5 0.5 Units$12.510GB = = • 20GB10GB 0.5 Units = > > • 20GB10GB • 20GB10GB • 20GB10GB > > • 20GB > 4 Units$100 4 Units$100$100 = = 4 Units$100 4 Units$100 = = 4 Units =
- 33. $ Value of Apple Segment 1 Segment 2 $ Value of Memory