2
MKT 315
Ce Liang
Dr.M.GailVermillion
5/35/2020
Assumptions in conjoint analysis
There are various assumptions in a conjoint analysis. The first assumption is that any product or service is a bundle of attributes. This means that depending on the product in question, it will be defined by several attributes, including its name, outlook, reliability among many other attributes (Rao, 2014). A product is also defined by other attributes such as aesthetics, physical and psychological. Delete the yellow. all you need to say is “products are a bundle of attributes”
The other assumption is that products variance is based on the attribute levels. This means that products with highly regarded attributes will have higher prices and vice versa. This depends on the value that consumers attach to the product. This assumption has nothing to do with price. All it says is “If an attribute level changes the product Changes” That is all you needed to say. Since it appears that you do not understand I am going to give you an example, This is just so you can learn – do not include this in your REDO tool summary 3. Example: If a pair of glasses are made from the attributes Color, Brand and Material. If the glasses change from being black glasses to white glasses the product changes. Do you understand? Reply in the email you send me.
The last assumption is that the preferences of consumers for the bundles of attributes differ. This means that the value attributed to a certain feature by one consumer may differ with the next consumer. Therefore, this can only be established by finding out the utility levels for each attribute. Delete the yellow
Steps in a conjoint analysis
To perform a conjoint analysis, several key steps have to be followed. These are explained below; I do not want you to give examples I want you to tell me the steps. This is where you should be writing the things I say over and over in the video. I try to say it in one to two sentences sometimes less. I want you to watch the video again and put it on slow motion or stop it. It there is ANYTHING you do not understand I want you to pause the tape and write down your question. You can text it to me or email me.
The first step is the selection of attributes and levels that a product is composed of. The focus groups, and what else is used? in this case, may include all people. These may range from aesthetic, performance, physical and psychological aspects (Rao, 2014). For instance, a house may be defined by aspects such as space, location, purpose and type. This means that we can have a three bedroomed house, located in town, suitable for young families and is a bungalow. This description will be different for different houses. The levels should also be kept as specific as possible to avoid any contradictions. Delete yellow
The second step is the selection of stimulus representation. This begs the question of the choices of the seller regarding the profile to be exhibited to consumers. This requires one to come up with actual products, make use of texts, pictures, prototypes and use of pictures and texts. Delete yellow I said several times what we would be using and gave the reason. Write what I said in the video here.
The third step involves designing the experiment. This involves establishing how many possible profiles can be found and whether the profilers should rate or rank each of the profiles. Delete yellow This requires the use of orthogonal experimental design.
The fourth step involves the regression analysis. This is followed by an interpretation of the conjoint output. The interpretation is made by use of a bar chat (Rao, 2014). Once this is done, it can enable one to understand the customer value structure better. It also enables the producer to use utility to find the best product. Besides, it is also possible to find the total attribute importance and an estimation of the relative market share. Delete yellow. Tell me what the three things we look at are on the Regression output and what they mean. One sentence for each. I will give you the first one Adjusted R2 should be at .3 or more and shows how much data we captured. Now you do Sig F and Coefficients.
I am a little concerned that you do not care if you pass this class because you did not even bother to email and ask for help or try to do what you know is 60% of the assignment. Even after I know you said you had been in contact with your group. You are encouraged to share data with your group. So do what you can then send your work by email to group and say “hey guys this is what I have done so far can you help or send me in right direction”
This is the last time I will allow a REDO on anything in the class. SO PLEASE ask for my help.
References
Rao, V. R. (2014). Applied conjoint analysis (p. 389). New York: Springer. Do not use references it makes you try to write more and leads you off track
cereal ranking
PROFILE | Rank | X1 | X2 | X3 | X4 | X5 | X6 | X7 | X8 | ||
1 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | Original Data from Surveys put in format to run Regression Analysis in Excel - go to next sheet | |
2 | 5 | 0 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | ||
3 | 6 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | 0 | ||
4 | 3 | 0 | 1 | 0 | 0 | 0 | 1 | 1 | 0 | ||
5 | 7 | 0 | 1 | 0 | 1 | 1 | 0 | 0 | 0 | ||
6 | 4 | 0 | 1 | 1 | 0 | 0 | 0 | 0 | 1 | ||
7 | 2 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 1 | ||
8 | 8 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 0 | ||
9 | 9 | 1 | 0 | 1 | 0 | 0 | 1 | 0 | 0 | ||
1 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | ||
2 | 6 | 0 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | ||
3 | 9 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | 0 | ||
4 | 4 | 0 | 1 | 0 | 0 | 0 | 1 | 1 | 0 | ||
5 | 8 | 0 | 1 | 0 | 1 | 1 | 0 | 0 | 0 | ||
6 | 5 | 0 | 1 | 1 | 0 | 0 | 0 | 0 | 1 | ||
7 | 3 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 1 | ||
8 | 2 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 0 | ||
9 | 7 | 1 | 0 | 1 | 0 | 0 | 1 | 0 | 0 | ||
1 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | ||
2 | 4 | 0 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | ||
3 | 9 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | 0 | ||
4 | 2 | 0 | 1 | 0 | 0 | 0 | 1 | 1 | 0 | ||
5 | 8 | 0 | 1 | 0 | 1 | 1 | 0 | 0 | 0 | ||
6 | 7 | 0 | 1 | 1 | 0 | 0 | 0 | 0 | 1 | ||
7 | 5 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 1 | ||
8 | 3 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 0 | ||
9 | 6 | 1 | 0 | 1 | 0 | 0 | 1 | 0 | 0 | ||
1 | 4 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | ||
2 | 8 | 0 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | ||
3 | 9 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | 0 | ||
4 | 5 | 0 | 1 | 0 | 0 | 0 | 1 | 1 | 0 | ||
5 | 6 | 0 | 1 | 0 | 1 | 1 | 0 | 0 | 0 | ||
6 | 7 | 0 | 1 | 1 | 0 | 0 | 0 | 0 | 1 | ||
7 | 1 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 1 | ||
8 | 2 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 0 | ||
9 | 3 | 1 | 0 | 1 | 0 | 0 | 1 | 0 | 0 | ||
1 | 1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | ||
2 | 8 | 0 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | ||
3 | 3 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | 0 | ||
4 | 2 | 0 | 1 | 0 | 0 | 0 | 1 | 1 | 0 | ||
5 | 5 | 0 | 1 | 0 | 1 | 1 | 0 | 0 | 0 | ||
6 | 9 | 0 | 1 | 1 | 0 | 0 | 0 | 0 | 1 | ||
7 | 4 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 1 | ||
8 | 6 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 0 | ||
9 | 7 | 1 | 0 | 1 | 0 | 0 | 1 | 0 | 0 |
&"Helvetica Neue,Regular"&12&K000000&P
Regression Output
SUMMARY OUTPUT | ||||||||||
Regression Statistics | ||||||||||
Multiple R | 0.7402702209 | |||||||||
R Square | 0.548 | This sheet is the output you get after running a regression analysis for Conjoint Analysis data. | ||||||||
Adjusted R Square | 0.4475555556 | |||||||||
Standard Error | 1.9407902171 | AdjR2 shows how much data you captured we want it to be .30 or above | ||||||||
Observations | 45 | |||||||||
Significance F tell me my results are not due to luck if the number is below .05 | ||||||||||
ANOVA | ||||||||||
df | SS | MS | F | Significance F | Coefficients are the UTILITY numbers that will be rescaled to fall between 0-1 - | |||||
Regression | 8 | 164.4 | 20.55 | 5.4557522124 | 0.0001549253 | the utility numbers represent how happy a person is with each attribute level. | ||||
Residual | 36 | 135.6 | 3.7666666667 | The Intercept is included with UTILITY numbers, | ||||||
Total | 44 | 300 | Take note that the chart does not include the levels that were dropped. | |||||||
I will add the dropped attribute levels back as 0 and then rescale them with the rest of the numbers - next sheet | ||||||||||
Coefficients | Standard Error | t Stat | P-value | Lower 95% | Upper 95% | Lower 95.0% | Upper 95.0% | |||
Intercept | 1.6 | 0.8679477711 | 1.8434288943 | 0.0735111575 | -0.1602796677 | 3.3602796677 | -0.1602796677 | 3.3602796677 | ||
X1 | -0.4666666667 | 0.7086763875 | -0.6585046079 | 0.5144036435 | -1.9039289968 | 0.9705956635 | -1.9039289968 | 0.9705956635 | ||
X2 | 0.4666666667 | 0.7086763875 | 0.6585046079 | 0.5144036435 | -0.9705956635 | 1.9039289968 | -0.9705956635 | 1.9039289968 | ||
X3 | 4.0666666667 | 0.7086763875 | 5.7383972971 | 0.0000015546 | 2.6294043365 | 5.5039289968 | 2.6294043365 | 5.5039289968 | ||
X4 | 3.1333333333 | 0.7086763875 | 4.4213880814 | 0.0000866603 | 1.6960710032 | 4.5705956635 | 1.6960710032 | 4.5705956635 | ||
X5 | 1.6 | 0.7086763875 | 2.2577300841 | 0.0301228241 | 0.1627376698 | 3.0372623302 | 0.1627376698 | 3.0372623302 | ||
X6 | 1.2 | 0.7086763875 | 1.6932975631 | 0.0990363362 | -0.2372623302 | 2.6372623302 | -0.2372623302 | 2.6372623302 | ||
X7 | -0.0666666667 | 0.7086763875 | -0.0940720868 | 0.9255735344 | -1.5039289968 | 1.3705956635 | -1.5039289968 | 1.3705956635 | ||
X8 | 0.2666666667 | 0.7086763875 | 0.3762883474 | 0.7089111697 | -1.1705956635 | 1.7039289968 | -1.1705956635 | 1.7039289968 |
Rescale
STEP 1: Find the biggest number under coefficients including Intercept-sometimes the biggest number is the Intercept | ||||||
Step 2: Find the smallest number | ||||||
Step 3: Calculate the Range by subtracting the smallest number from the largest number. | ||||||
Step 4: Rescale using the following formula - =(Coeffient Number - Lowest Number)/Range | ||||||
Step 5: Round off to 2 digits and then transfer to the next page | ||||||
Coefficients | RESCALED | Rounded Off | ||||
Intercept | 1.6 | 0.4569536424 | 0.46 | |||
X1 | -0.4666666667 | 0.0007358352 | 0 | |||
X2 | 0.4666666667 | 0.2067696836 | 0.21 | |||
X3 | 4.0666666667 | 1.0014716703 | 1 | |||
X4 | 3.1333333333 | 0.7954378219 | 0.79 | |||
X5 | 1.6 | 0.4569536424 | 0.46 | |||
X6 | 1.2 | 0.3686534216 | 0.37 | |||
X7 | -0.0666666667 | 0.0890360559 | 0.09 | |||
X8 | 0.2666666667 | 0.1626195732 | 0.16 | |||
soggy | 0 | 0.1037527594 | 0.1 | |||
Bad | 0 | 0.1037527594 | 0.1 | |||
Salty | 0 | 0.1037527594 | 0.1 | |||
Grandpa | 0 | 0.1037527594 | 0.1 | |||
Biggest Number | 4.06 | |||||
Smallest Number | -0.47 | |||||
Range | 4.53 |
Final Graph Data
X1 | Crunchy | |||||||
X2 | Chewy | Texture | ||||||
Dropped | Soggy | |||||||
This shows the attributes I dropped from each category-soggy , Bad, Salty, Grandpa | ||||||||
X3 | Good | |||||||
X4 | Edible | Taste | ||||||
Dropped | Bad | |||||||
X5 | Healthy | |||||||
X6 | Sugary | Nutrition | ||||||
Dropped | Salty | |||||||
X7 | Children | |||||||
X8 | College | Age | ||||||
Dropped | Grandpa | |||||||
use yellow box to create a column graph - only left X1-X8 so you could see where to put data from now on simply call by the attribute level | ||||||||
FROM PREVIOUS SHEET | X1 | Crunchy | 0 | |||||
Coeffiecients | Rounded Off | X2 | Chewy | 0.21 | ||||
Intercept | 0.46 | Soggy | 0.1 | |||||
X1 | 0 | |||||||
X2 | 0.21 | X3 | Good | 1 | ||||
X3 | 1 | X4 | Edible | 0.79 | ||||
X4 | 0.79 | Bad | 0.1 | |||||
X5 | 0.46 | |||||||
X6 | 0.37 | X5 | Healthy | 0.46 | ||||
X7 | 0.09 | X6 | Sugary | 0.37 | ||||
X8 | 0.16 | Salty | 0.1 | |||||
soggy | 0.1 | |||||||
Bad | 0.1 | X7 | Children | 0.09 | ||||
Salty | 0.1 | X8 | College | 0.16 | ||||
Grandpa | 0.1 | Grandpa | 0.1 | |||||
Intercept | 0.46 | |||||||
Understanding Consumer Preferences with Conjoint Analysis
Overview of Today’s Class
- Understanding conjoint analysis
- The procedure for conjoint analysis
- Interpreting conjoint output
- Creating and using choice simulators
- Running conjoint analysis using Excel
So, What Is Conjoint Analysis?
- Methodology used to decompose an individual’s value system for a product from overall judgment of the product
- Decomposition of the value system allows researcher to understand the value/utility of each product attribute at each attribute level.
- That’s right!
- For each attribute
- For each attribute level
When in CA Used?
- Very useful to make feature and feature-level trade-offs in new product design
- Calculate market share
- Determine market entry barriers
- Simulate market activity
3 Assumptions of Conjoint Analysis
- Every product/service is a “bundle” of attributes
- e.g. Image, brand name, reliability, etc. For this class we will be hired by a brand so you may not use brand as an attribute!!
- Physical, psychological, and aesthetic attributes
- Products differ via varying levels of attributes provided
- e.g. Quiet; Subdued; $1,200
3 Assumptions of Conjoint Analysis
- Consumer preferences for these “bundles” differ
- And hence, overall consumer preferences for these bundles can be decomposed into basic building blocks
- Utilities for each attribute and their levels
Procedure for Conjoint Analysis
Designing and Conducting the Experiment
Selecting attributes and levels that form
the product
Choosing stimulus representation and
Response People
Interpreting conjoint output
Data Collection &
Data Analysis via Multiple Regression
Step 1. Selecting Attributes and Levels That Form the Product – Focus Groups
- Can include all Peoples - physical, performance, psychological, aesthetic
- e.g. let’s assume 4 attributes of a Car
1. Image: Family, Sporty, Prestigious
2. Sound:Quiet, Subdued, Loud
3. People: 2 people, 4 people, 5 people
4. Service: Easy, Difficult, Impossible
Keep the levels specific
- Avoid words like “high, low, medium” or “average”
- Avoid any words that are not actionable
Step 1. Selecting Attributes and Levels That Form the Product
- Key to selecting attributes:
- Use focus groups, managers’ inputs, and competitive analyses
- Most relevant - 3 to 7 attributes
- Match number of levels - 3 to 4 levels each
- In our Car example: How many versions/combinations are possible?
- For our example? 3x3x3x3 = 81 possible profiles
- Do consumers have to rank all versions?
Step 2. Choosing Stimulus Representation
- What are your choices in exhibiting the “profiles” to customers?
- Create actual products
- Use prototypes
- Use pictures
- Use text
- Use pictures and text
Step 2. Choosing Stimulus Representation
- Full profile
- all attributes included in each profile
Profile 1
Impossible
Quiet
Sporty
4 people
Profile 2
Service: Easy
Sound: Subdued
Image: Sporty
People: 5 people
Step 2. Choosing Response People
- Choosing customer response People
- ranking profiles – less than 10 profiles
- rating profiles – 10 OR more than 10 profiles
- choice based
- Pause….
- Remember decompositional technique?
- What are you “decomposing”?
Step 2. Choosing Response People
- Generally, ranking and rating data provide similar results - hence choose based on
- number of profiles
- potential respondent fatigue
Step 3. Designing the Experiment
- How many profiles possible?
- Multiplicative product of number of levels across all attributes
- For our example? 3x3x3x3 = 81 possible profiles
- Should respondents rank/rate each profile?
- Tiring! Fatigue = Source of error
- Use orthogonal experimental design
Step 3. Orthogonal Experimental Designs
- Limited number of profiles
- However, limited enough such that RELIABLE estimation of all utilities is possible
- So, how many profiles?
- At least = # of utilities estimated
- # of utilities estimated =
Sum across all attributes (# of levels for each attribute - 1)
Step 3. Orthogonal Experimental Designs
- So for our example…
- # of levels for Image = 3
- # of levels for Sound = 3
- # of levels for People = 3
- # of levels of Service = 3
- Hence, # of utilities estimated =
- (3-1) + (3-1) + (3-1) + (3-1) = 8
- Hence, # of profiles = 8 = FLOOR
- WOW! Ranking/Rating a minimum of 8 carefully selected profiles will enable us to RELIABLY estimate utilities for 81 possible profiles
- Efficient & Reliable
- If Orthogonal Design book does not have design with 8 profiles go to the next level
Step 3. The Design Part A The Code-sheet
- Think multiple regression
Y = a + b1X1 + b2X2 + b3X3 + b4 X4 + b5X5 + b6X6 + b7X7 + b8X8
- Image
- X1 = 0,1 (0 = not Prestigious, 1 = Prestigious)
- X2 = 0,1 (0 = not Sporty, 1 = Sporty)
- Sound
- X3 = 0,1 (0= not Quiet, 1 = Quiet)
- X4 = 0,1 (0=not Subdued, 1 = Subdued)
- People
- X5 = 0,1 (0 = not 5 people, 1 = 5 people)
- X6 = 0,1 (0 = not 4 people, 1 = 4 people)
- Service
- X7 = 0,1 (0= not Easy, 1 = Easy)
- X8 = 0,1 (0=not Difficult, 1 = Difficult)
- What about Family, Loud, 2 people, & Impossible?
Step 3. The Design – Part B
- Remember, for our example we need at least 8 profiles
- KEY: Each row above is a profile to be ranked/rated
- Put profiles in words – there are 9 rows, hence 9 profiles above
PROFILE | Rank | X1 | X2 | X3 | X4 | X5 | X6 | X7 | X8 |
1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | |
2 | 0 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | |
3 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | 0 | |
4 | 0 | 1 | 0 | 0 | 0 | 1 | 1 | 0 | |
5 | 0 | 1 | 0 | 1 | 1 | 0 | 0 | 0 | |
6 | 0 | 1 | 1 | 0 | 0 | 0 | 0 | 1 | |
7 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 1 | |
8 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 0 | |
9 | 1 | 0 | 1 | 0 | 0 | 1 | 0 | 0 |
Step 3. Conducting the Experiment
Use Design Part A and Design Part B together
- Profile 1 results from row 1 in Design Part B (ignore the label row), and from the code-sheet that you created in Design Part A
- Profile 2 results from combining row 2 in Design Part B and the code-sheet in Design Part A
- And so on….
Step 3. Combining Design Parts A & B
Profile 1
Image: Family
Sound: Loud
People: 2 people
Service: Impossible
Profile 2
Image: Family
Sound: Subdued
People: 4 people
Brand name: Difficult
Profile 3
Image: Family
Sound: Quiet
People: 5 people
Brand name: Easy
Profile 4
Image: Sporty
Sound: Loud
People: 4 people
Service: Easy
Profile 5
Image: Sporty
Sound: Subdued
People: 5 people
Service: Impossible
Profile 6
Image: Sporty
Sound: Quiet
People: 2 people
Service: Difficult
Profile 7
Image: Prestigious
Sound: Loud
People: 5 people
Service: Difficult
Profile 8
Image: Prestigious
Sound: Subdued
People: 2 people
Service: Easy
Profile 9
Image: Prestigious
Sound: Quiet
People: 4 people
Service: Impossible
Step 3. Presenting the Profiles
- Few Rules:
- Make the profiles uncluttered – not too wordy
- Mention both the feature name and the feature level in each profile
- Put a rating or a ranking option below each profiles
- Let the respondents clearly know what the scale or ranking means
- Watch for signs of confusion and fatigue – pre-test, pre-test
Step 3. Presenting the Profiles – I have template for you!!!
Step 4. Obtain Rankings or Ratings
- Higher the ranking/rating means higher the number
- 9 means most preferred
- 1 means least preferred
Rank | X1 | X2 | X3 | X4 | X5 | X6 | X7 | X8 |
1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
2 | 0 | 0 | 0 | 1 | 0 | 1 | 0 | 1 |
5 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | 0 |
6 | 0 | 1 | 0 | 0 | 0 | 1 | 1 | 0 |
7 | 0 | 1 | 0 | 1 | 1 | 0 | 0 | 0 |
4 | 0 | 1 | 1 | 0 | 0 | 0 | 0 | 1 |
9 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 1 |
3 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 0 |
8 | 1 | 0 | 1 | 0 | 0 | 1 | 0 | 0 |
1 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
4 | 0 | 0 | 0 | 1 | 0 | 1 | 0 | 1 |
5 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | 0 |
2 | 0 | 1 | 0 | 0 | 0 | 1 | 1 | 0 |
6 | 0 | 1 | 0 | 1 | 1 | 0 | 0 | 0 |
3 | 0 | 1 | 1 | 0 | 0 | 0 | 0 | 1 |
7 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 1 |
9 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 0 |
8 | 1 | 0 | 1 | 0 | 0 | 1 | 0 | 0 |
2 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
1 | 0 | 0 | 0 | 1 | 0 | 1 | 0 | 1 |
4 | 0 | 0 | 1 | 0 | 1 | 0 | 1 | 0 |
6 | 0 | 1 | 0 | 0 | 0 | 1 | 1 | 0 |
5 | 0 | 1 | 0 | 1 | 1 | 0 | 0 | 0 |
8 | 0 | 1 | 1 | 0 | 0 | 0 | 0 | 1 |
3 | 1 | 0 | 0 | 0 | 1 | 0 | 0 | 1 |
7 | 1 | 0 | 0 | 1 | 0 | 0 | 1 | 0 |
9 | 1 | 0 | 1 | 0 | 0 | 1 | 0 | 0 |
Step 4. The Regression Analysis
- Use Excel for analysis, multiple regression
Step 4. Re-scaling Utilities - Utilities are re-scaled to fit between 0 and 1
Utility | Rescaled | Rescale Formula= U-L/Range | |
X3 = Quiet | 2.91 | 1 | (2.91-(-.58))/3.493 |
X1= Prestigious | 2.83 | 0.98 | (2.83-(-.58))/3.493 |
X5= 5 people | 2.25 | 0.81 | |
X4= Subdued | 2.08 | 0.76 | |
X2 = Sporty | 1.66 | 0.64 | |
X6= 4 people | 1.25 | 0.52 | |
Intercept | 1 | 0.45 | (1-(-.58))/3.493 |
X9= Family | 0 | 0.17 | (0-(-.58))/3.493 |
X10= Loud | 0 | 0.17 | |
X11= 2 people | 0 | 0.17 | |
X12= Impossible | 0 | 0.17 | |
X7= Easy | -0.41 | 0.05 | |
X8 = Difficult | -0.58 | 0 | |
Range=Highest-Lowest U=Utility Number | |||
Prestigious
Sporty
Family
Loud
Subdued
Quiet
2 people
4 people
5 people
Easy
Difficult
Impossible
Intercept
Chart1
Category 1 | Category 1 | Category 1 |
Category 2 | Category 2 | Category 2 |
Category 3 | Category 3 | Category 3 |
Category 4 | Category 4 | Category 4 |
0.45 |
Sheet1
series 1 | series 2 | series 3 | |
Category 1 | 0.98 | 0.64 | 0.17 |
Category 2 | 0.17 | 0.76 | 1 |
Category 3 | 0.17 | 0.52 | 0.81 |
Category 4 | 0.05 | 0 | 0.17 |
0.45 |
6 Outputs of Conjoint Analysis
- Once you’ve created a bar chart using the rescaled attribute level utilities, you can
Get a deeper understanding of customer value structure
Find the best product based on total utility
Determine overall attribute importance
Estimate relative market share
Anticipate how a change in one attribute will impact total utility and hence market share, and what value-neutral tradeoffs can be made – also called simulating the market
Identify the minimum acceptable product
Interpreting Output 1 – Develop Better Understanding of Customer Value Structure
- Making trade-offs between various levels of Image, Sound, People, & Service
- Understand drop in utilities between levels
- Find “sweet spots” if they exist
- Get a very good idea of customers’ value structure
Prestigious
Sporty
Family
Loud
Subdued
Quiet
2 people
4 people
5 people
Easy
Difficult
Impossible
Intercept
Linear or Non-Linear – MUST KNOW COST TO DETERMINE –in this class only use for price
Chart1
Category 1 | Category 1 | Category 1 |
Category 2 | Category 2 | Category 2 |
Category 3 | Category 3 | Category 3 |
Category 4 | Category 4 | Category 4 |
0.45 |
Sheet1
series 1 | series 2 | series 3 | |
Category 1 | 0.98 | 0.64 | 0.17 |
Category 2 | 0.17 | 0.76 | 1 |
Category 3 | 0.17 | 0.52 | 0.81 |
Category 4 | 0.05 | 0 | 0.17 |
0.45 |
$500
$800
$1200
MADE UP
SWEET SPOTS these have nothing to do with previous chart
$500
$800
$1200
Chart1
Category 1 | Category 1 | Category 1 |
Category 2 | Category 2 | Category 2 |
Sheet1
Series 1 | Series 2 | Series 3 | |
Category 1 | 0.45 | 0.25 | 0.45 |
Category 2 | 0.25 | 0.45 | 0.25 |
To resize chart data range, drag lower right corner of range. |
Interpreting Output 2 – Optimal Product
- How many profiles did customers rank for the Car example?
- How many Car combinations were possible?
- Can test all possible combinations
- Even if customers did not see all combinations
- WHY??
- The efficiency and reliability of CA!!
- In short, create optimal products
Interpreting Output 2 – Optimal Product
- The concept of TPU – Total Product Utility
- Best possible Car? – Look at the bar chart!!
Prestigious .98
Quiet 1.00
5 people .81
Impossible .17
TPU = 2.96
however, can we afford to offer this combination?
- Worst possible product?
- 2nd best?
- 15th from the top?...and so on
Interpreting Output 2 – Optimal Product
- To create an optimal product, a company MUST
- Provide customers with the highest possible TPU
- AND simultaneously make a profit!
- Rank order all products by TPU and by costs – then supply the one with the highest TPU and the maximum profit – we can’t do in class because we do not have costs for creating each attribute level
Interpreting Output 3:
Calculating Overall Feature Importance
- We know what the utility of each level of each feature is (the BAR CHART, OF COURSE!)
- What about the overall features?
- Image? Sound? People? Service?
- Calculate range for each feature
- Highest utility value within a feature minus the lowest utility value with a feature
Interpreting Output 3:
Calculating Overall Feature Importance
- Look at the bar chart!
- Image Range = 0.98 – 0.17 = 0.81
- Sound Range = 1.00 – 0.17 = 0.83
- People Range = 0.81 – 0.17 = 0.64
- Service Range = 0.17 – 0.00 = 0.17
- Sum of ranges =
- 0.81+0.83+0.64+0.17 = 2.45
Interpreting Output 3:
Calculating Overall Feature Importance
- Importance of each feature = Feature range divided by sum of all features’ ranges
- Importance of Image = 0.81/2.45 = 33.06%
- Importance of Sound = 0.83/2.45 = 33.87%
- Importance of People = 0.64/2.45 = 26.12%
- Importance of Service = 0.17/2.45 = 6.93%
Interpreting Output 3:
Calculating Overall Feature Importance
Chart1
Image |
Sound |
People |
Service |
Sheet1
Image | Sound | People | Service | |
33.08% | 33.87% | 26.12% | 6.93% |
Interpreting Output 3: Most Important Attributes/ “hot buttons”
- So what are the “hot buttons” or important attributes in our PC example?
- Image
- Sound
- People
- Service
- So this gives you an investment priority
- How is this similar to perceptual mapping?
Interpreting Output 4:
The Intercept
- The re-scaled intercept suggests a market entry barrier as perceived by the target market
- Minimum acceptable product for it to be part of target market’s consideration set
- Any product must total up to be greater than the intercept
- What does a large intercept value typically indicate?
Interpreting Output 5:
Calculating Current Market Share
- Remember from output 2, we could have potentially calculated TPU values for all possible combinations
- Somewhere among those TPU values are all our competitors
- Use the appropriate feature/level combinations and create the entire market’s profiles that you compete with
- You can create a realistic marketplace
- 10-12 competitors including yourself
Interpreting Output 5:
Calculating Current Market Share
- Create the entire market’s profiles and calculate each profile’s utility
- Market share =
exp (Utility of Us)/sum exp (Utility of Us; Utility of them)
Interpreting Output 5:Calculating
Current Market Share (Stage 1)
- Assume 3 product market–
Capri Prelude BMW
Family .17 Sporty .64 Prestigious .98
Loud .17 Subdued .76 Subdued .76
2 people .17 4 people .52 5 people .81
Easy .05 Difficult .00 Impossible .17
TPU .56 1.92 2.72
Stage 1: Current Market Share
Chart1
Capri |
Prelude |
BMW |
Sheet1
Market Share | |
Capri | 7.30% |
Prelude | 28.70% |
BMW | 64% |
To resize chart data range, drag lower right corner of range. |
Interpreting Output 5:Calculating Current Market Share (Stage 1)
- Market share calculations:
- Capri = exp(.56)/[exp(.56) + exp(1.92) + exp(2.72)]
- = 1.75/(1.75 + 6.82 + 15.18)
- = 7.3%
- Prelude = exp(1.92)/[exp(1.92) + exp(.56) + exp(2.72)]
- = 6.82/(6.82 + 1.75 + 15.18)
- = 28.7%
- BMW = exp(2.72)/[exp(2.57) + exp(.56) + exp(1.92)]
- = 15.18/(15.18 + 1.75 + 6.82)
- = 64%
Interpreting Output 5:
Simulating the Market – Potential Market Share
- Can test reaction to competitors’ actions
- Simulate the market
- Test impact of feature changes on market share
- What if Capri tries to catch-up to Prelude?
- Loud to Quiet
- What happens?
Interpreting Output 5:Simulating the Market – Potential Market Share (Stage 2)
- Assume same 3 product market – hypothetical
- But with the changes from the previous slide
Capri Prelude BMW
Quiet 1.00 Subdued .76 Subdued .76
Family .17 Sporty .64 Prestigious .98
2 people .17 4 people .52 5 people .81
Easy .05 Difficult .00 Impossible .17
TPU 1.39 1.92 2.72
*
Stage 2: Changed Most Important Attribute
UP 8.1%
Chart1
Capri |
Prelude |
BMW |
Sheet1
Market Share | |
Capri | 15.40% |
Prelude | 26% |
BMW | 58.40% |
To resize chart data range, drag lower right corner of range. |
Interpreting Output 5: Simulating the Market – Potential Market Share (Stage 2)
- Market share calculations:
- Capri = exp(1.39)/[exp(1.39) + exp(1.92) + exp(2.72)]
- = 4.01/(4.01+ 6.82 + 15.18)
- = 15.4% up 8.1% (WOO HOO!)
- Prelude = exp(1.92)/[exp(1.92) + exp(1.39) + exp(2.72)]
- = 6.82/(6.82 + 4.01 + 15.18)
- = 26.2% down 2.5%
- BMW = exp(2.72)/[exp(2.72) + 39 + exp(1.92)]
- = 15.18/(15.18 + 4.01 + 6.82)
- = 58.4% down 5.6%
Interpreting Output 5:Simulating the Market – Competition reacts (Stage 3)
BMW will not sit still when they lose over 10% market share
Capri Prelude BMW
Quiet 1.00 Subdued .76 Subdued .76
Family .64 Sporty .64 Prestigious .98
2 people .17 4 people .52 5 people .81
Easy .05 Difficult .00 Easy .05
TPU 1.39 1.92 2.60
Stage 3: Potential Market Share
up 1.1% (WOO HOO!) from beginning up 9.2%
Chart1
Capri |
Prelude |
BMW |
Sheet1
Column1 | |
Capri | 16.50% |
Prelude | 28.00% |
BMW | 56% |
To resize chart data range, drag lower right corner of range. |
Interpreting Output 5:Simulating the Market – Potential Market Share (Stage 3)
- Market share calculations:
- Capri = exp(1.39)/[exp(1.39) + exp(1.92) + exp(2.60)]
- = 4.01/(4.01 + 6.82 + 13.46)
- = 16.5% up 1.1% (WOO HOO!) from beginning up 9.2%
- Prelude = exp(1.92)/[exp(1.92) + exp(1.39) + exp(2.60)]
- = 6.82/(6.82 + 4.01 + 13.46)
- = 28% up 1.8%
- BMW = exp(2.60)/[exp(2.60) + exp(1.39) + exp(1.92)]
- = 13.46/(13.46 + 4.01 + 6.82)
- = 55.5% down 2.9%
Interpreting Output 5:
Simulating the Market – Potential Market Share
- Can test many such rounds of our firm’s actions and competitive reactions
- Determine the most appropriate feature and level combination based on these simulations
Summary
- The value of conjoint analysis
- Using conjoint analysis
- Interpreting and leveraging conjoint analysis
7.30%28.70%64%CapriPreludeBMW
15.40%26%58.40%CapriPreludeBMW
16.50%28.00%56%CapriPreludeBMW
Tool Summary 3
1. Define CA/ Assumptions (20%)
2. EXPLAIN Steps for CA (20%)
3. Explain Example(60%)
Draw column chart
· Show linearities, non-linearities, sweet spots – only with numerical values
· Best product show product and TPU Worst product show product and TPU Is it viable? Why?
· Intercept – what does it mean? What is it in your example?
· Overall Feature Importance – Pie Chart
· Simulate Market – 3 products – 3 stages STAGE 1 :-Now
STAGE 2- I change my product HOT BUTTON BETTER CHANGE!! ONLY ONE ATTRIBUTE CHANGES
STAGE 3-competition reacts Do not let them change a hot button! Only ONE competitor reacts and they change only one attribute level