MKT Tool 3
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