Conjoint Analysis - Management By The Numbers Analysis This module covers how to interpret the...

Conjoint AnalysisThis module covers how to interpret the results of a conjoint study, including the topics of attribute importance, willingness-to-pay, statistical validity, customer feature trade-offs, and market share prediction.

Author: Ronald T. WilcoxMarketing Metrics Reference: Chapter 4© 2010-14 Ronald T. Wilcox and Management by the Numbers, Inc.

MEASU

RIN

GC

ON

SUM

ERPR

EFEREN

CES

Measuring Consumer Preference

Consumers have complex preferences!

We want to be able to understand the preferences of our customers and potential customers for the products and services we are offering or may consider offering.

We can examine the choices people make to uncover how they value the different features of a product or service.

MBTN | Management by the Numbers 2

CO

NJO

INT

AN

ALYSIS-D

EFINITIO

NConjoint Analysis - Definition

We seek to determine how consumers value the different features that make up a product and the tradeoffs they are willing to make among the different features when they make product choices. Sometimes these hidden drivers of choice are not apparent to the consumers themselves.

DefinitionConjoint analysis is a statistical technique used in marketing research to measure how people value the different features or attributes that comprise a multi-attribute product or service.


WH

ENC

ON

JOIN

TA

NALYSIS

CAN

HELP

When Conjoint Analysis Can Help

Conjoint Analysis can help when:

1. The product or service being examined is comprised of multiple attributes or potential attributes.

• Example: When deciding on a new car a consumer may consider price, fuel economy, size of the engine, etc.

2. The various values or levels an attribute are well understood by consumers and can be described in unambiguous terms.

• Example: “24 Miles per Gallon” is unambiguous, “Great New Styling” is ambiguous.

3. It is reasonable to believe that consumers strongly consider these attributes when making product choices.


HO

WC

ON

JOIN

TA

NALYSIS

WO

RKS

How Conjoint Analysis Works

Show consumers a series of hypothetical products defined by their attributes.

• Ask the respondent to pick the product they like the best.• Some types of conjoint analysis ask consumers to rank

products according to their preferences rather than choosing a single most-preferred product.

Gather multiple observations per person and include many people in the study.

• Use responses to estimate the preferences for various features.• Marketing researchers generally used specialized software to

come up with these estimates.


EXPER

IMEN

TALD

ESIGNThe Experimental Design

Conjoint analysis begins with an experimental design. This design includes all attributes, and the values of the attributes, that will be tested. Conjoint analysis distinguishes between attributes and what are generally called “levels.”

An attribute is pretty self-explanatory. It could be something like price, color, horsepower, material used for the upholstery, or presence of sunroof.

A “level” is the specific value or realization of the attribute. For example, the attribute “color” may have the levels “red,” “blue,” and “yellow” while the attribute “presence of a sunroof” will have levels “yes” and “no.”

Before a researcher begins to collect data it is important to write down all of the levels of each attribute to be tested. Commercially available software packages will require that the user provide these as input.


EXPER

IMEN

TALD

ESIGNThe Experimental Design

Attribute: Price Brand Horsepower Upholstery Sunroof$23,000 Toyota 220 HP Cloth Yes

Levels: $25,000 Volkswagen 250 HP Leather No$27,000 Saturn 280 HP$29,000 Kia

It is important to note that any level of an attribute might be combined with any other level of another attribute during the experiment. For example, the consumer might be presented a hypothetical product that a Volkswagen, priced at $29,000, has 220 horsepower, a leather interior and a sunroof.

This is a very simple design that contains a total of 15 attribute levels. Real world designs often contain more attributes and levels than are presented here.

Staying with our car example, an experimental design might look like the following:


BASIC

CO

NJO

INT

AN

ALYSISO

UTPU

TBasic Conjoint Analysis Output

Attribute Level Utility (Part-worth) t-valuePrice $23,000 2.10 14.00

$25,000 1.15 7.67$27,000 -1.56 10.40$29,000 -1.69 11.27

Brand Toyota 0.75 5.00Volkswagen 0.65 4.33

Saturn -0.13 0.87Kia -1.27 8.47

Horsepower 220 HP -2.24 14.93250 HP 1.06 7.07280 HP 1.18 7.87

Upholstery Cloth -1.60 10.67Leather 1.60 10.67

Sunroof Yes 0.68 4.53No -0.68 4.53

The basic results of a conjoint analysis are the estimated attribute level utilities or part-worths. These correspond to the average consumer preference for the level of any given attribute.


BASIC

CO

NJO

INT

AN

ALYSISO

UTPU


Note: Within a given attribute, the estimated utilities are generally scaled in such a way that they add up to zero. So, a negative number does not mean that the given level has “negative utility,” It just means that this level is on average less preferred than a level with an estimated utility that is positive.

MBTN | Management by the Numbers


$25,000 1.15 7.67$27,000 -1.56 10.40$29,000 -1.69 11.27


Saturn -0.13 0.87Kia -1.27 8.47



Sunroof Yes 0.68 4.53No -0.68 4.53

9

BASIC

CO

NJO

INT

AN

ALYSISO

UTPU


Conjoint analysis output is also often accompanied by t-values, a standard metric for evaluating statistical significance. Because of the way conjoint utilities are scaled, the standard interpretation of t-values can yield misleading results.

For example, the level “Saturn” of the attribute “Brand” has a t-value of 0.87. In general, a t-value of this magnitude would fail a test of statistical significance. However, this t-value is generated because within the attribute “Brand” the level “Saturn” has neither a very high nor very low relative preference. It is basically in the middle in terms of overall preference.


Attribute Level Utility (Part-worth) t-valueBrand Toyota 0.75 5.00

Volkswagen 0.65 4.33Saturn -0.13 0.87Kia -1.27 8.47

10

BASIC

CO

NJO

INT

AN

ALYSISO

UTPU


Because of the scaling, levels that have more moderate levels of preference within a given attribute are likely to have estimated utilities close to zero which will tend to produce very low t-values (recall that the t-test is measuring the probability that the true value of a parameter is not different from zero).



$25,000 1.15 7.67$27,000 -1.56 10.40$29,000 -1.69 11.27


Saturn -0.13 0.87Kia -1.27 8.47



Sunroof Yes 0.68 4.53No -0.68 4.53

11

BASIC

INTER

PRETATIO

NBasic Interpretation

The utility of any product we might be considering is simply the sum of the utilities of its attribute levels. For example, a Toyota with 280 horsepower, leather interior, no sunroof and a price of $23,000 has a utility of 0.75 + 1.18 + 1.60 - 0.68 + 2.10 = 4.95

As a natural consequence of this additive property, relatively large positive numbers can be interpreted as adding a lot to overall product utility. Negative utilities indicate less desired levels of attributes with the most negative being the least desired.

Utility estimates are directly comparable across attributes. For example, having a leather rather than cloth interior (1.60 versus -1.60) adds more to overall product utility than having a sunroof (0.68 versus -0.68).


INTER

PRETATIO

N: ATTR

IBUTE

IMPO

RTAN

CE

Interpretation: Attribute ImportanceIntuitively, the difference in the utilities between the most preferred andleast preferred level within an attribute tells you something about howimportant this attribute is in the consumer choice process.

In order to calculate the importance of any given attribute, you just takethe difference between the highest and lowest utility level of thatattribute and divide this by the sum of the differences between thehighest and lowest utility level for all attributes (including the one inquestion). The resulting number will always lie between zero and oneand is generally interpreted as the percent decision weight of anattribute in the overall choice process.


INTER

PRETATIO

N: ATTR

IBUTE

IMPO

RTAN

CE

Interpretation: Attribute Importance

Consider the attribute “horsepower.” What is its overall importance inthe choice process?

Insight:This means that 25% of the overall decision weight of consumers is on horsepower. The greater this percentage, the greater the decision weight. An analogous calculation for “Price” is about 27%, “Brand” about 15%, “Sunroof” about 10%, and “Upholstery” about 23%. These numbers provide a very intuitive metric for thinking about the importance of each attribute in the decision process.

For practice, you might want to calculate these other important values.

( ))68.068.0()60.160.1()24.218.1()27.175.0()69.110.2(24.218.1

++++++++++

=HorsepowerI =.25


INTER

PRETATIO

N: WILLIN

GN

ESS-TO-PAYInterpretation: Willingness-to-pay

Question 1: Consider a Toyota at a price of $23,000 with 280 horsepower, leather interior and no sunroof. How much would the average consumer be willing-to-pay if a sunroof were added?

The car without the sunroof has a utility of 0.75 + 1.18 + 1.60 - 0.68 + 2.10 = 4.95. If we add the sunroof this increases the overall utility to 0.75 + 1.18 + 1.60 + 0.68 + 2.10 = 6.31 for a difference of 6.31 – 4.95 = 1.36.

This implies that we can reduce the utility of price by 1.36 and the average consumer would be just as happy as before the sunroof was installed. By carefully examining the estimated utilities for the different levels of price we can solve for this new price!


Answer: To find out how much the price can be raised we must convert the change in utility to a change in price. We do this by first noting how much the original car costs $23,000 and the utility associate with that figure, 2.10. We know that we can reduce the price utility by 1.36. This is equivalent to saying that we can reduce the price utility to 2.10 – 1.36 = 0.74. By referring to the utility estimates we can immediately see that this implies a price between $25,000 and $27,000 because -1.56 < 0.74 < 1.15. In fact, if we assume a linear relationship between price and utility in the range between $25,000 and $27,000 we can solve for the exact price by performing a linear interpolation within this range.

WILLIN

GN

ESS-TO-PAY(C

ON

TINU

ED)Willingness-to-pay (continued)

58.302,25$000,2$*)56.1(15.1

74.015.1000,25$ =−−−

+

Utility spread between the two tested price

points ($25K and $27K).

Utility spread between $25K and

the target utility.


WILLIN

GN

ESS-TO-PAY(C

ON

TINU

ED)Willingness-to-pay (continued)


This implies that if the sunroof is added and the price of the vehicle is raised from $23,000 to about $25,300 the average consumer would be indifferent between the two vehicles. Managerially, it provides the maximum amount the price can be raised if the new feature is included without compromising market share. In this case the maximum is $2,300 ($25,300 - $23,000).

58.302,25$000,2$*)56.1(15.1

74.015.1000,25$ =−−−

+

Utility spread between the two tested price

points ($25K and $27K).

Utility spread between $25K and

the target utility.

17

INTER

PRETATIO

N: TR

ADE-O

FFAN

ALYSISInterpretation: Trade-off analysis

Calculating willingness-to-pay for a feature upgrade is an example of one particular type of trade-off analysis. A consumer is trading off money (price) for something better on another dimension of the product.


Insight: More generally, we can use conjoint analysis any trade-off a consumer would be willing to make among the attributes that are tested in the experiment. These estimated trade-offs become important determinants of design when a company is developing new products or services.

18

INTER

PRETATIO

N: TR

ADE-O

FFAN


Question 2: Suppose Volkswagen is currently selling a car for $25,000 that comes standard with a 250 HP engine, cloth interior and no sunroof. If Volkswagen upgrades this vehicle to have a leather interior, how much power (horsepower) would the average consumer be willing to give up to get this upgrade?

This analysis proceeds much like the willingness-to-pay analysis. First, we must determine how much additional utility the average consumer gets from the upgrade to a leather interior.


INTER

PRETATIO

N: TR

ADE-O

FFAN



Answer:The additional utility for the leather interior is the difference between the utility for the leather and cloth interior, 1.6 + 1.6 = 3.2. Converting this to horsepower, exactly as we did in the willingness-to-pay problem, yields:

The average consumer would be willing to reduce overall horsepower from 250HP to 221HP, a reduction of 29HP, to get a leather interior.

HPHPHP 221)220250(*)24.2(06.1

)6.1(6.1250 =−−−−−

−

Utility spread between the two relevant levels (250HP and 220HP).

Utility spread between leather and cloth

interiors.

20

INTER

PRETATIO

N: MAR

KETS

HAR

EF

OR

ECASTIN

GInterpretation: Market Share Forecasting

Another common application is forecasting market share. In order to use conjoint analysis output for this kind of prediction two conditions must be satisfied.

1. The company must know the other products, besides their own offering, that a consumer is likely to consider when making a selection in the relevant category.

2. Each of these competitive products’ important features must be included in the experimental design. In other words, you must be able to calculate the utility of not only your own product offering but that of the competitive products as well.


INTER

PRETATIO

N: MAR

KETS

HAR

EF

OR

ECASTIN

GInterpretation: Market Share Forecasting

Market share prediction relies on the use of a multinomial logit model. The basic form of the logit model is:

∑ =

= n

j

U

U

i j

i

eeShare

1

where

• Ui is the estimated utility of product i (sum of attribute utilities)• Uj is the estimated utility of product j• n is the total number of products in the competitive

set, including product i.

Yes, that e you see really is the one you sort of remember from high school, the base of the natural logarithm (approximately 2.718).


EXAM

PLE: MAR

KETS

HAR

EF

OR

ECASTIN

GExample: Market Share Forecasting

Question 3: Suppose we are considering marketing a car with the following profile: Saturn; $23,000; 220 HP; Cloth interior; No sunroof. We believe that when consumers consider our car they will also consider purchasing cars that are currently on the market with the following profiles:

• $27,000; Toyota; 250 HP; Cloth Interior; No Sunroof• $29,000; Volkswagen; 280 HP; Leather Interior; No Sunroof• $23,000; Kia; 220 HP; Cloth Interior; No Sunroof

Relative to the other cars consumers will be considering, what market share will this car achieve?

We need to calculate the overall utility of each vehicle and then use the market share formula.


EXAM

PLE: MAR

KETS

HAR

EF

OR

ECASTIN



Answer:For the Saturn and its associated product profile the estimated utility is 2.10 – 0.13 – 2.24 – 1.60 – 0.68 = – 2.55. Similarly, the utilities of the three competing products can be calculated:

(Toyota) = – 1.56 + 0.75 + 1.06 – 1.60 – 0.68 = – 2.03 (Volkswagen) = – 1.69 + 0.65 + 1.18 + 1.60 – 0.68 = + 1.06(Kia) = + 2.10 – 1.27 – 2.24 – 1.60 – 0.68 = – 3.69

Now, substitute these values into the market share formula…

∑ =

= n

j

U

U

ij

i

eeShare

1

24

EXAM

PLE: MAR

KETS

HAR

EF

OR

ECASTIN



Answer:

Which after substituting values gives us:

Conjoint analysis predicts a market share of 2.5% relative to this set of competitors.

=+++

= −−−

−

69.306.103.255.2

55.2

eeeeeShareSaturn .025 or 2.5%

∑ =

= n

j

U

U

i j

i

eeShare

1

25

MAR

KETS

HAR

EF

OR

ECASTIN

G–

KEEP

INM

IND

Market Share Forecasting – Keep in Mind

Things to keep in mind with using conjoint analysis for estimating potential market share:1. There are many non-attribute drivers of market share. These

include distribution (ACV), channel promotion and display, retail price variation, awareness of a new brand or awareness of the enhanced features on existing brand, etc.

2. There may be other product attributes or product choices that are not included in experimental design.

3. Competitive response is not considered in conjoint. The dynamic nature of choice is not captured unless you also consider “what if” scenarios.

4. Validity of the logit model for market share estimation.


CO

NJO

INT

AN

SWER

SQ

UESTIO

NS

Questions Conjoint Analysis Answered

• Determining which attributes are the primary drivers of product choice (attribute importance).

• Determining consumers’ willingness-to-pay for a proposed new product or product redesign.

• Quantifying the trade-offs customers or potential customers are willing to make among the various attributes or features that are under consideration in the new product design process.

• Predicting the market share of a proposed new product given the current offerings of competitors.

We have now worked through examples for each of the four questions we set out to explore using conjoint analysis:


FIN

ALC

ON

SIDER

ATION

SFinal Considerations

• There are many types of conjoint analysis. This presentation has explored one popular type, choice-based conjoint analysis. For some applications, a different type might be appropriate.

• It is common in modern forms of conjoint analysis to run separate analyses for different groups of consumers instead of running a single analysis for all consumers. In some cases, it is also possible to estimate separate conjoint analysis utilities for individual participants.

• Estimating the utilities in a conjoint analysis is not easy. Specialized software has automated this process, but if you are engaged in a business application of conjoint analysis it is important that you understand how these estimates are being produced.


FU

RTH

ERR

EFEREN

CE

Further Reference


Marketing Metrics by Farris, Bendle, Pfeifer and Reibstein, 2nd edition, chapter 4.A good reference for many of the specific questions you may have about conducting a conjoint analysis can be found at: http://www.sawtoothsoftware.com/conjoint-analysis-software which is the website of a company (Sawtooth Software) that markets conjoint analysis software.

29

http://www.sawtoothsoftware.com/conjoint-analysis-software

Conjoint Analysis - Management By The Numbers Analysis This module covers how to interpret the...

Documents

Transcript of Conjoint Analysis - Management By The Numbers Analysis This module covers how to interpret the...