07.07.

10

Conjoint Analysis
Basic Principle

Klaus Goepel 7.7.2010

Conjoint Analysis
Basic Principle

The presentation explains the principle, using a simple example. It shows, how to calculate the part-worth utilities and how to derive the relative preferences from individual attributes from there. A full factorial and a fractional factorial design is used. An Excel template for this example is available from the author.

Conjoint analysis or Stated preference analysis is a statistical technique that originated in mathematical psychology.

Conjoint Analysis 7.7.2010 Gp

07.07.10

Conjoint Analysis
Basic Principle

Keywords conjoint analysis, stated preference analysis, linear regression, product management, marketing, part-worth, utilities, relative preference, statistics, analytic hierarchy process, AHP

Today it is used in many of the social sciences and applied sciences including - Marketing, - Product management, - Operations research.

Conjoint Analysis
Basic Principle Buying a smart phone, MP3 player a b Your Preference The preference for a combination of (conjoint) attributes will reveal the partworth of individual attributes. Attribute 1: Memory Attribute 2: Delivery Attribute 1: Memory Attribute 2: Delivery

Conjoint Analysis 7.7.2010 Gp

07.07.10

Conjoint Analysis
Basic Principle Buying a smart phone, MP3 player

b The preference for a combination of (conjoint) attributes will reveal the partworth of individual attributes.

Higher emphasis on short delivery time.

Higher emphasis on large memory size

Conjoint Analysis
Basic Principle Buying a smart phone, MP3 player

Part-worth utilities of individual attributes are calculated based on the ranking of a defined set of combinations of attribute values.

Conjoint Analysis 7.7.2010 Gp

07.07.10

Conjoint Analysis
Basic Principle Buying a smart phone, MP3 player

Attributes:

Color Memory

green, red 16 MB, 64 MB 1 day, 1 week 5 6 7 8 Green, 16 MB, 1 day Red, 16 MB, 1 day Green, 64 MB, 1 day Red, 64 MB, 1 day
Attribute values are coded with -1 and +1 Attribute 1: Color Attribute 2: Memory Attribute 2: Delivery

Levels: -1; +1

Example Buying a smart phone

Models: 1 2 3 4

Delivery

Green, 16 MB, 1 week Red, 16 MB, 1 week Green, 64 MB, 1 week Red, 64 MB, 1 week

Sequence

Delivery

Memory

Conjoint Analysis
Basic Principle

Conjoint Attributes Green, 16 MB, 1 week Red, 16 MB, 1 week Green, 64 MB, 1 week Red, 64 MB, 1 week Green, 16 MB, 1 day Red, 16 MB, 1 day Green, 64 MB, 1 day Red, 64 MB, 1 day

1 2 3 4 5 6 7 8

-1 1 -1 1 -1 1 -1 1

Color

-1 -1 1 1 -1 -1 1 1

-1 -1 -1 -1 1 1 1 1 Attribute values are coded with -1 and +1 Attribute 1: Color Attribute 2: Memory Attribute 2: Delivery Example Buying a smart phone

Conjoint Analysis 7.7.2010 Gp

07.07.10

Conjoint Analysis
Basic Principle

Full factorial design

Design Matrix -1 1 -1 -1 -1 1 1 -1 -1 1 1 -1 -1 -1 -1 1 1 1 1 Design Matrix k attributes: 2k possible combinations Full factorial Design

k attributes: 2k combinations

1 -1 1 -1 1

Conjoint Analysis
Basic Principle X1 7 5 delivery 6 8 1 2 3
or em y

X2 -1 -1 1 1 -1 -1 1 1

X3 -1 -1 -1 -1 1 1 1 1 X1: Color = (+1,-1) Graphical Representation of Combinations Full factorial Design

-1 1 -1 1 -1 1 -1 1

3 2

4 5 6 7 8

1 X2 X1 X3 color

X2: Memory = (+1,-1) X3: Delivery = (+1,-1)

Conjoint Analysis 7.7.2010 Gp

07.07.10

Sequence

Conjoint Analysis
Basic Principle Conjoint Attributes Green, 16 MB, 1 week Red, 16 MB, 1 week Green, 64 MB, 1 week Red, 64 MB, 1 week Green, 16 MB, 1 day Red, 16 MB, 1 day Green, 64 MB, 1 day Red, 64 MB, 1 day X1 -1 1 -1 1 -1 1 -1 1 X2 -1 -1 1 1 -1 -1 1 1 X3 -1 -1 -1 -1 1 1 1 1 Rank 8 7 4 3 6 5 2 1 X1: Color = (+1,-1) X2: Memory = (+1,-1) X3: Delivery = (+1,-1) Ranking of combinations

1 2 3 4 5 6 7 8

Conjoint Analysis
Basic Principle Linear model function with part-worth utilities

Ranking = part-worth of attribute 1 * attribute 1 level + part-worth of attribute 2 * attribute 2 level + part-worth of attribute 3 * attribute 3 level + baseline preference

Linear Model Function

The system of linear equations can be solved with linear regression

Y=

color*X1

memory*X2

delivery*X3 +

X1: Color = (+1,-1) X2: Memory = (+1,-1) X3: Delivery = (+1,-1)

Conjoint Analysis 7.7.2010 Gp

07.07.10

Conjoint Analysis
Rank Basic Principle Part-worth utilities = -1 1 -1 1 -1 1 -1 1 * + -1 -1 1 1 -1 -1 1 1 * + -1 -1 -1 -1 1 1 1 1 * + Linear Model Function
The system of linear equations can be solved with linear regression

8 7 4 3 6 5 2 1

Color

Memory

Delivery

X1 7 5 6 8 1 2 3 3 1 2 4 4 5 6 7 8 -1 1 -1 1 -1 1 -1 1

X2 -1 -1 1 1 -1 -1 1 1

X3 -1 -1 -1 -1 1 1 1 1

Rank 8 7 4 3 6 5 2 1 Graphical Representation of Combinations

Conjoint Analysis 7.7.2010 Gp

Conjoint Analysis
Basic Principle

07.07.10

Conjoint Analysis
Basic Principle X1 7 5 6 8 1 2 3 3 1 X1 Main effect X1
Col=

X2 -1 -1 1 1 -1 -1 1 1

X3 -1 -1 -1 -1 1 1 1 1

-1 1 -1 1 -1 1 -1 1

Rank 8 7 4 3 6 5 2 1
Color

Graphical Representation of Combinations Calculating Part-worth Utilities = - 0.5

4 2

4 5 6 7 8

[16 16 - 20 20] 2 = -0.5

Conjoint Analysis
Basic Principle X1 7 5 6 8 1 2 3 3 1 X2 Main effect X2
Mem=

X2 -1 -1 1 1 -1 -1 1 1

X3 -1 -1 -1 -1 1 1 1 1

-1 1 -1 1 -1 1 -1 1

Rank 8 7 4 3 6 5 2 1
Color Memory

Graphical Representation of Combinations Calculating Part-worth Utilities = - 0.5 = -2

4 2

4 5 6 7

[10 10 - 26 26] 2 =

-2

Conjoint Analysis 7.7.2010 Gp

07.07.10

Conjoint Analysis
Basic Principle X1 7 5 6 8 1 2 3 3 1 X3 Main effect X3
Del=

X2 -1 -1 1 1 -1 -1 1 1

X3 -1 -1 -1 -1 1 1 1 1

-1 1 -1 1 -1 1 -1 1

Rank 8 7 4 3 6 5 2 1
Color Memory

Graphical Representation of Combinations Calculating Part-worth Utilities = - 0.5 = -2 = -1

4 2

4 5 6 7

Delivery

[14 14 - 22 22] 2 =

-1

Conjoint Analysis
Basic Principle

7 5 6

Part-worth utilities = = = Graphical Representation of Combinations Calculating Part-worth Utilities = - 0.5 = -2 = -1

Color

-0.5 -2 -1

3 1 2

Mem

Del

Color Memory Delivery

Conjoint Analysis 7.7.2010 Gp

07.07.10

Conjoint Analysis
Basic Principle Actual ranking and description with linear model function
9 8 7 6 5 4 3 2 1 0 1 week 16 MB green 1 week 16 MB red 1 week 64 MB green 1 week 64 MB red 1 day 16 MB green 1 day 16 MB red 1 day 64 MB green 1 day 64 MB red Rank Model

Ranking and Model Function

Color Memory Delivery

= - 0.5 = -2 = -1

Conjoint Analysis
Basic Principle Actual ranking and description with linear model function
9 8 7 6 5 4 3 2 1 0 1 week 16 MB green 1 week 16 MB red 1 week 64 MB green 1 week 64 MB red 1 day 16 MB green 1 day 16 MB red 1 day 64 MB green 1 day 64 MB red Rank Model

Ranking and Model Function

Color Memory Delivery

= - 0.5 = -2 = -1

Y = 4.5 -0.5 Xcolor -2 XMemory -1 XDelivery

Conjoint Analysis 7.7.2010 Gp

10

07.07.10

Conjoint Analysis
Basic Principle

7 5 6

Variations for Xi=1 1 1 1 = 0.5 = 1 = = 2 = 4 1 = 2 7 Calculating relative preferences

Col

3 1 Color Memory Delivery 17 = 47 = 27 = 2

Mem

Del

14% 57% 29%

Conjoint Analysis
Basic Principle Conjoint Analysis Attributes Part-worth Utilities Levels Ranking AHP Criteria, Sub-criteria - comparison Weights: Principal Eigenvector Ratio Scale, relative Scale Evaluation of Alternatives Conjoint Analysis & Analytic Hierarchy Process AHP

2k

Combinations

k2 k 2

Comparisons

k=4: 16 possible combinations

k=4: 6 pair-wise comparisons

Conjoint Analysis 7.7.2010 Gp

11

07.07.10

Conjoint Analysis
Basic Principle X1 7 5 6 8 1 2 3 3 1 2 4 4 5 6 7 8 -1 1 -1 1 -1 1 -1 1 X2 -1 -1 1 1 -1 -1 1 1 X3 -1 -1 -1 -1 1 1 1 1 Rank 8 7 4 3 6 5 2 1
Fractional Design

Conjoint Analysis
Basic Principle X1 7 5 6 8 1 2 3 3 1 2 4 4 5 6 7 8 -1 1 -1 1 -1 1 -1 1 X2 -1 -1 1 1 -1 -1 1 1 X3 -1 -1 -1 -1 1 1 1 1 Rank 8 7 4 3 6 5 2 1
Fractional Design

Graphical Representation

Conjoint Analysis 7.7.2010 Gp

12

07.07.10

Conjoint Analysis
Basic Principle X1 7 5 6 8 2 3 3 1 2 4 5 -1 -1 1 6 1 -1 -1 1 -1 -1 7 4
Fractional Design

X2

X3

Rank Graphical Representation

23-1

Fractional design

23-1
III

3 = 12

Conjoint Analysis
Basic Principle X1 4 3 1 2 3 2 1 4 1 -1 -1 1 X2 -1 1 -1 1 X3 -1 -1 1 1 Rank 7 4 6 1
Fractional Design

Graphical Representation

23-1

Fractional design

Conjoint Analysis 7.7.2010 Gp

13

07.07.10

Conjoint Analysis
Basic Principle Fractional factorial design 2k-p 4 3 1 2 3 2 1 4 X1 1 -1 -1 1 X2 -1 1 -1 1 X3 -1 -1 1 1 Rank 7 4 6 1 Calculating Part-worth Utilities
Color

Fractional Design

Graphical Representation

X1 Main effect X1

= - 0.5

[(7+1)(4+6)] 2 = -0.5

Conjoint Analysis
Basic Principle Fractional factorial design 2k-p 4 3 1 2 3 2 1 4 X1 1 -1 -1 1 X2 -1 1 -1 1 X3 -1 -1 1 1 Rank 7 4 6 1 Calculating Part-worth Utilities
Color

Fractional Design

Graphical Representation

X2 Main effect X2

Memory

= - 0.5 = -2

[(4+1)(7+6)] 2 = -2

Conjoint Analysis 7.7.2010 Gp

14

07.07.10

Conjoint Analysis
Basic Principle Fractional factorial design 2k-p 4 3 1 2 3 2 1 4 X1 1 -1 -1 1 X2 -1 1 -1 1 X3 -1 -1 1 1 Rank 7 4 6 1 Calculating Part-worth Utilities
Color

Fractional Design

Graphical Representation

X3 Main effect X3

Memory Delivery

= - 0.5 = -2 = -1

[(6+1)(7+4)] 2 = -1

Conjoint Analysis
Basic Principle

Fractional Design

Using a fractional factorial design the number of attribute combinations can be reduced.

Conjoint Analysis 7.7.2010 Gp

15

07.07.10

Conjoint Analysis
Basic Principle

Simple conjoint analysis can be done with linear regression, but more sophisticated statistical models and solutions are available.

Conjoint Analysis 7.7.2010 Gp

16