The power of conjoint analysis

ThePowerofConjointAnalysis
Insightintopreference,choiceandtrade-off
DavidMurray

Downloadfreebooksat


David Murray

The Power of Conjoint Analysis
Insight into preference, choice and trade-off

2
Download free eBooks at bookboon.com


The Power of Conjoint Analysis : Insight into preference, choice and trade-off
© 2012 David Murray & bookboon.com
ISBN 978-87-403-0214-1

3
Download free eBooks at bookboon.com


Contents

The Power of Conjoint Analysis

Contents
1Introduction

5



The Role of Statistics in Market Research

5

2

An Outline of Conjoint Analysis

13



Sample Size Considerations:

15

3

Case Studies

28



Public Sector Case Study: Exercising Choice for Specialist Healthcare Treatment

35

4

Generic Conjoint Techniques

5

Reflection and the Way Forward

Bibliography

360°
thinking

.

Biography

360°
thinking

.

40
49
54
55

360°

thinking

.

Discover the truth at www.deloitte.ca/careers

© Deloitte & Touche LLP and affiliated entities.

Discover the truth at www.deloitte.ca/careers

Touche LLP and affiliated entities.

© Deloitte & Touche LLP and affiliated entities.

Discover the truth 4at www.deloitte.ca/careers
Click on the ad to read more
Download free eBooks at bookboon.com
© Deloitte & Touche LLP and affiliated entities.

Disc


Introduction

The Power of Conjoint Analysis

1Introduction
The Role of Statistics in Market Research
Wikipedia defines statistics as ‘The study of the collection, organisation, analysis and interpretation of
data’. Wordwebonline goes on to add the fact that data can be represented in numerical format, which

will inform the basis of this particular insight.
Statistics is a branch of applied mathematics concerned with the collection, interpretation and to an
extent, the manipulation of quantitative data. These data can be compiled from a number of sources:
• Secondary data – these are data that already exist in hard or soft documents where the
analyst has the appropriate authority to extract and interpret. Examples of these documents,
freely available in the public domain, are Census Statistics, Regional Trends, Social Trends,
Euromonitors and data available upon subscription from Mintel, the Target Group Index or
National Shoppers’ Survey.
• Primary data – these are data necessitating fieldwork to derive meaningful information in
the form of Market Research Surveys where data can be compiled through observation or
communication using:
-- Self-completion questionnaires
-- Face-to-face interviews door-to-door or on the street
-- Telephone interviews
-- Online questionnaires completed upon an e-mail request or by an invite on a website
The completed questionnaires will contain response codes to enable the analyst to input or data capture
responses in numerical format. For example a response to a dichotomous will create three codes:
RESPONSE

CODE

YES

1

NO

2

DON’T KNOW

3
Fig. 1 Dichotomous coding frame

or a response to a semantic scale will create either three codes:
RESPONSE

CODE

AGREE

1

NEITHER AGREE OR DISAGREE

2

Fig. 2 Three scale semantic coding frame

5
Download free eBooks at bookboon.com


Introduction

The Power of Conjoint Analysis

or five codes:
RESPONSE

CODE

STRONGLY AGREE

1

AGREE

2

NEITHER AGREE OR DISAGREE

3

DISAGREE

4

STRONGLY DISAGREE

5

Fig. 3 Five scale semantic coding frame

Qualitative responses can also be formatted into a coding structure to permit quantitative data analysis,
for example in a customer satisfaction survey, the question asks:
“What one thing would improve the service Company X provides its customers?”
A randomly selected ten questionnaires from the hundred carried out reported the following answers:
• “Answer the telephone within five rings”
• “Staff to be polite and courteous”

• “Quicker response to my telephone call”
• “Replenish stock levels”
• “Shelves run out of stock very quickly”
• “More staff to be available to answer queries”
• “When I phone up, not to be passed from one staff member to another”
• “More accurate invoices”
• “Phone me back quicker with an answer”
• “Invoice not clear”
Within these open-ended responses there are some key words:
• Tele(phone)
• Stock
• Staff – polite
• Staff – available
• Invoice
If codes were to be affixed to these key words or statements, we can now quantify open-ended responses
to this very important question and analyse the data:

6
Download free eBooks at bookboon.com


Introduction

The Power of Conjoint Analysis

KEYWORD

CODE

RESPONSE(S)

% RESPONSE

Tele(phone)

1

4

40

Stock

2

2

20

Staff – polite

3

1

10

Staff – available

4

2

20

Invoice

5

2

20

Base

10
Fig. 4 Hard coding open-ended analysis

Please note one respondent referred to two service aspects, (telephone and staff available)
These codes have been set up based on a random selection of open-ended responses from every nth, or
in this case, 10th questionnaire. These codes together with their interpretation, or coding frame, can now
be used to code all one hundred responses with a default ‘others’ for any responses which do not apply
to the descriptions within the coding frame.
You will have noticed a simple analysis of responses in this table - this is what is known as Univariate
Analysis. Univariate analysis is carried out using a single variable, in this case percentage response to
the question, “What one thing would improve the service Company X provides its customers?”
The analysis depicted is a frequency analysis of the distribution of responses. So we can conclude that
40%, or 4 in every 10 responses made a comment pertaining to the company’s telephone service. The
standard output of Univariate Analysis is a frequency distribution table and for a more visual impact,
a chart or a graph.

Let us now assume that the analyst has coded all one hundred responses to this open-ended question.
The data table now is as follows:
KEYWORD/PHRASE

RESPONSES

TELE(PHONE)

37

STOCK

26

STAFF – POLITE

12

STAFF – AVAILABLE

18

INVOICE

11

Fig. 5 Integrating hard coded open-ended questions into the coding frame
(Some respondents will have given more than one answer).

7

Download free eBooks at bookboon.com


Introduction

The Power of Conjoint Analysis

Most market research questions ask a number of classification questions. This enables the analyst to find
out more about the respondents in terms of:
• Gender
• Age
• Income
• Social grade
• Household size
• Employment status
Let us now have a look at the gender distribution of these responses:
KEYWORD/PHRASE

RESPONSES

MALES

FEMALES

TELE(PHONE)

37

9

28

STOCK

26

6

20

STAFF – POLITE

12

8

4

STAFF – AVAILABLE

18

12

6

INVOICE

11

9

2

Fig. 6 Gender distribution of hard coded responses

We are now in a position to add more detail and insight into the analysis by carrying out Bivariate
Analysis. Bivariate Analysis determines the relationship between these two variables, improvement and
gender. Bivariate Analysis can be helpful in testing simple hypotheses of association, for example:
• Males are more likely to query invoices
• Females are more likely to complain about items being out of stock
Let us now look at testing these hypotheses:
• Query Invoices:
RESPONSES

PROFILE %

9

81.8

FEMALES

2

18.2

TOTAL

11

100

MALES

Fig. 7 Gender profile of hard coded responses

8
Download free eBooks at bookboon.com


Introduction

The Power of Conjoint Analysis

• One can see that males are more likely than females to focus on invoice accuracy.
Furthermore, let us assume that the overall sample of one hundred respondents was split
fifty males and fifty females. We can now add value to this analysis and state that of all
males who took part in the survey, 18% or almost 1 in 5, are more likely to query an invoice
whereas only 4% or 1 in 25 females were likely to query invoice accuracy.
Therefore males are 4½ times more likely to query an invoice. The hypothesis ‘Males are
more likely to query invoices’ has been tested and proved accurate.
• Out of stock:
SAMPLE SIZE

RESPONSES

PROFILE %

PENETRATION%

MALES

50

6

23.1

12.0

FEMALES

50

20

76.9

40.0

100

26

100.0

Fig. 8 Comparative profile and penetration of hard coded responses

Increase your impact with MSM Executive Education

For almost 60 years Maastricht School of Management has been enhancing the management capacity
of professionals and organizations around the world through state-of-the-art management education.
Our broad range of Open Enrollment Executive Programs offers you a unique interactive, stimulating and
multicultural learning experience.
Be prepared for tomorrow’s management challenges and apply today.
For more information, visit www.msm.nl or contact us at +31 43 38 70 808 or via
For more information, visit www.msm.nl or contact us at +31 43 38 70 808
the
globally networked management school
or via
Executive Education-170x115-B2.indd 1

18-08-11 15:13

9
Download free eBooks at bookboon.com

Click on the ad to read more


Introduction

The Power of Conjoint Analysis

By applying the same procedure as in the previous analyses, three in every four complainants about outof-stock items are female. Furthermore, four in every ten females voiced their opinion about this problem
compared with only one in eight males. In terms of the hypotheses, we can now accept the hypothesis
that females are more likely to complain about items being out-of-stock than males.
Let us now assume that the questionnaire contained five point semantic scales regarding their satisfaction
with the company’s service. The questionnaire asks the respondents to either give a score of between 1

and 5, with 5 being very satisfied and 1 being not at all satisfied, or just a tick box under the appropriate
satisfaction heading for each of these service aspects:
Aspect of service provided

Very Satisfied Satisfied Neither...nor Dissatisfied
(5)

(4)

(3)

Very Dissatisfied

(2)

(1)

1. Convenient opening hours

1

2

3

4

5

2. The company are market leaders

1

2

3

4

5

3. Company X is very professional

1

2

3

4

5

4. Their product range is extensive

1

2

3

4

5

5. When I ring the telephone is answered quickly

1

2

3

4

5

6. Staff are approachable

1

2

3

4

5

7. Staff are polite

1

2

3

4

5

8. All things considered, I am happy with the

1

2

3

4

5

service provided
Fig. 9 Five scale very satisfied to very dissatisfied semantic scale pertaining to key business drivers

Based on the answers given by the one hundred respondents, we can now correlate the overall satisfaction,
(all things considered, I am happy with the service), scores with each of the seven aspects of the service
delivery. Correlation analysis is the statistical relationship, in this instance, between two variables, the
dependent variable (all things considered), and each of the independent variables (e.g. opening hours).

Output is expressed as the ‘coefficient of correlation’ or the strength of the association between the
dependent and independent variables. The coefficient is always 1.0 or less with 1.0 equating to perfect
correlation. The closer the coefficient is to 1.0, the better the correlation.
Should three independent variables result in similar coefficients of correlation of 0.7 or greater, then we
need to test variations of these or combinations of these three variables using a multivariate statistical
technique called Regression Analysis.

10
Download free eBooks at bookboon.com


Introduction

The Power of Conjoint Analysis

Multivariate Analysis relates to the relationship between a number of variables, usually three or more,
and can be applied for not just explaining current the current relationship between variables but can
also be used for modelling purposes to predict possible outcomes or ‘What if ’ scenarios by measuring
the impact on a dependent variable by changing the values of independent variables.
Here is a list of multivariate analysis techniques to explain multidimensional data relationships:
• Regression Analysis – the focus is on the relationship between a dependent variable, e.g.
overall satisfaction, and one or more independent variables e.g. polite staff, telephone
answered quickly and market leaders. Having established the coefficient of correlation at a
level of say 0.67, where only two-thirds of the relationship is explained, one can vary any
of the independent variables to measure the impact on the coefficient of correlation so
as to increase the outcome from 0.67 to 0.8 or beyond. There are a number of regression
techniques available, but coverage of these is beyond the scope of this book.

GOT-THE-ENERGY-TO-LEAD.COM
We believe that energy suppliers should be renewable, too. We are therefore looking for enthusiastic

new colleagues with plenty of ideas who want to join RWE in changing the world. Visit us online to find
out what we are offering and how we are working together to ensure the energy of the future.

11
Download free eBooks at bookboon.com

Click
Click on
on the
the ad
ad to
to read
read more
more


Introduction

The Power of Conjoint Analysis

• Factor Analysis – this is a method of data reduction by grouping together those variables
which are closely correlated. This grouping together of closely associated variables forms
new variables called factors or principle components. For example, in a survey on ‘Motorists’
Opinions’, the following scale variables from a range of fifty original variables, made up
Factor 1, (Moy 2001)
-- Good petrol consumption
-- Low maintenance costs
-- Keeps its value
-- Good value for money
We might interpret and label factor 1 as the ‘Economy’ factor.

• Cluster Analysis – assigns a set of objects into groups called clusters, so that the objects in
the same cluster are more similar to each other than those in other clusters.
• Correspondence Analysis – provides a means of displaying or summarising a set of data in
two-dimensional graphic form.
• Discriminate Analysis – identifying exploitable subgroups of consumers to design products
with benefits. The essence of this is to match goods and services to consumer requirements.
The method of maximising identification and discrimination of key variables with a data set
is Automatic Interaction Detector, a sub technique of Cluster Analysis.
• Conjoint Analysis – or trade-off analysis is a collection of standard statistical techniques that
provides objective insights into consumer preferences. This multivariate statistical analysis
technique and its related models now form the basis of this book.

12
Download free eBooks at bookboon.com


An Outline of Conjoint Analysis

The Power of Conjoint Analysis

2 An Outline of Conjoint Analysis
It has been said that the term ‘Conjoint’ was derived from two words: ‘considered jointly’. I cannot confirm
that this is true but it does illustrate the fundamental idea behind Conjoint Analysis.
Since its introduction in 1971, conjoint analysis has been one of the fastest growing marketing research
techniques, particularly in the USA. A large number of blue-chip companies have adopted the technique,
namely The Boston Consulting Group, Hewlet Packard, Levi Strauss, McKinsey and Company, Proctor
and Gamble, Smith, Kline and French, Xerox and South-Western Bell. This appears to be due to the
actionability of its results and the predictive power it has displayed (Wilderstrom, 1994).
In conjoint analysis, consumers are asked to react to a number of hypothetical concepts or service
descriptions. When consumers are offered a wide range of choices, they would ideally like to have

everything at the easiest level of access, but will inevitably have to ask themselves:
“What am I willing to give up?” As we can’t have everything, trade-offs are inevitable.
Let’s look at a simple scenario – booking a week’s holiday in Spain. Upon evaluating a holiday brochure,
the customer notices holidays are available in February, Easter, May and August. There are four different
types of accommodation – camping in a tent, an apartment, hotel or a villa. Amenities available include
en-suite bathroom facilities, located near to the beach, located near to a golf course or located near to a
bullring. The prices listed in the brochure are holidays from £200 - £1,500 with two intermediary price
points of £500 and £1,000. Ideally, as human nature prevails, we would love to select a villa by the beach
in August for £200, but that offer will never arise. So, depending on one’s personal budget, the time of
year one can go on holiday subject to work constraints and personal holiday pursuits, one has to make
trade-offs:
“OK, I can’t afford a villa, but maybe an apartment for £500 but at least it’s near the beach and August
is too hot”
or
“I’ve only got £200 so I’ll have to go out of season and can’t expect anything more than camping in a tent
near the bullring”
This gives us a rather simplistic view of trade-offs, or a snapshot of consumer behaviour upon evaluating
choice. When evaluating choice on a much larger scale to determine the optimal balance of trade-offs
is where the statistics play an important role.

13
Download free eBooks at bookboon.com


An Outline of Conjoint Analysis

The Power of Conjoint Analysis

Keys to evaluating choice are:
• The product/service/proposition attributes

• Levels within these attributes
In the aforementioned example there are four key attributes, namely:
• Price per week
• Time of year
• Accommodation type
• Amenities
Within each of these attributes there are four levels each namely:
Price per week
• £200
• £500
• £1,000
• £1,500
Time of year
• February
• Easter
• May
• August
Accommodation
• Tent
• Apartment
• Hotel
• Villa
Amenities
• En-suite bathroom facilities
• Located near to the beach
• Located near to a golf course
• Located near to the bullring

14
Download free eBooks at bookboon.com

An Outline of Conjoint Analysis

The Power of Conjoint Analysis

Within this holiday offer or proposition, there are 4x4x4x4=256 permutations, clearly too much for
a consumer to evaluate. To simplify choice procedures, a subset of these, called a fractional factorial
design (Kuhfeld 1994), is presented to the consumer. The selection of permutations is randomised and
the application software (SPSS from IBM) extracts the selection automatically. In this instance, the
number of permutations randomly selected, are 16 i.e. (4+4+4+4). This is considered sufficient to model
the remaining 240 permutations under the assumption that the sample of respondents interviewed, is
statistically robust.

Sample Size Considerations:
We would recommend a sample size of at least 385 respondents to derive statistically valid
conclusions. The reason for this is that, at this particular level, the margin of sampling error starts
to decrease at a diminishing rate. The statistically validity based on a sample size of say 200, may
be questionable because the margin of sampling error is too high to measure statistically significant
differences between two data observations. For example, based on a 50%/50% response to a Yes/No
question, the margin of sample error for 200 respondents would be ±7%, meaning that the amount
of dispersion around the 50% proportion, results in a range of 43% - 57% answering either ‘Yes’ or
‘No’ to the question. As can be seen from this graph overleaf and data table, there is a level where
the margin of sampling error starts to decrease at a diminishing rate. We are again assuming a
50%/50% response:

With us you can
shape the future.
Every single day.
For more information go to:

www.eon-career.com

Your energy shapes the future.

15
Download free eBooks at bookboon.com

Click
Click on
on the
the ad
ad to
to read
read more
more


An Outline of Conjoint Analysis

The Power of Conjoint Analysis

Sample Size

Margin of Sampling Error (±)

100

9.8

200

6.9

300

5.7

400

4.9

500

4.4

600

4.0

700

3.7

800

3.5

900

3.3

1,000

3.1

Fig. 10 Margin of sampling errors for given sample sizes

The margin of sampling error is derived from the formula:
߲݊ ൌ

ඥሺ’ሺ’ െ ͳሻሻȀ[

Where ߲݊ ൌ margin of sampling error
P = proportion expressed as a decimal
n = sample size
1.96 = standard unit for 95% level of confidence

0$5*,12)(5525
























6$03/(6,=(6

Fig. 11 Geographical distribution of margin of sampling errors for a range of sample sizes

16
Download free eBooks at bookboon.com






An Outline of Conjoint Analysis

The Power of Conjoint Analysis

This means that as the sample size increases over 385 correspondents, the reduction in the margin of
sampling error starts to plateau for every incremental increase. So, we hereby recommend a sample of at

least 385 respondents. Obviously the greater the sample size, the greater the statistical validity, but there
is the consideration of increased fieldwork costs to offset against this increased validity.
Let’s now return to our example of holidaymakers’ evaluating a range of options. When formulating a
questionnaire to accommodate this evaluation, one must ensure that the travel company can offer every
single level, because if the results indicate preference for purchasing a holiday package comprising an
element the company cannot offer, the exercise becomes purely academic.
SPSS has now made the randomised selection of levels within attributes to enable the modelling of all
possible outcomes, four examples of which could be:
Example 1:
Price:£500
Time of Year:

August

Accommodation:Hotel
Amenities:

Located near a bullring

Example 2:
Price:£1,000
Time of Year:

May

Accommodation:Villa
Amenities:

Located near a golf-course

Example 3:
Price:£200
Time of Year:

February

Accommodation:Apartment
Amenities:

Bathroom facilities

17
Download free eBooks at bookboon.com


An Outline of Conjoint Analysis

The Power of Conjoint Analysis

Example 4:
Price:£1,500
Time of Year:

Easter

Accommodation:Tent
Amenities:

Located by the beach

Fig. 12 Randomised selection of levels within attributes

All sixteen options are then printed onto laminated A4 cards for respondents to evaluate.
The questionnaire must be very explicit and easy to interpret in its outline of what the respondent should
do with the cards. An ideal script could be:
“I am now going to pass you some cards. All cards have four aspects of a holiday offer in Spain printed on
them which relate to:
• Price of the holiday
• The time of year the holiday is offered

www.job.oticon.dk

18
Download free eBooks at bookboon.com

Click
Click on
on the
the ad
ad to
to read
read more
more


An Outline of Conjoint Analysis

The Power of Conjoint Analysis

• Accommodation type

• Amenities available”
“What I’d like you to do is read the text then create two piles of cards; the first pile being a range of holiday
offers you consider are attractive and the second pile being a range of holidays that you find unattractive.
It doesn’t matter how many cards are in each pile.”
The respondent has now created his/her pile of holiday offers – one good/attractive pile, one poor/
unattractive piles.
The interviewer now hands the respondent the good/attractive pile and asks the respondent to sort this
pile into the order of preference with the overall most preferred at the top and the least preferred at the
bottom. The respondent now ranks from most preferred to least preferred. The interviewer now records
the card numbers on the grid on the questionnaire with the most preferred in position 1, the second most
preferred in position 2 and so on. The interviewer must take care and record the card number sequence
in the exact same order that the respondent has sorted the cards. At this stage the grid should have the
following hypothetical appearance, assuming eight cards have appeared in the good/attractive pile:
Overall most
preferred rank = 1

2

3

4

5

6

7

8

Card number

14

8

3

1

7

16

2

9

Order

9

10

11

12

13

14

15

16

Order

Fig. 13 Ranking of most preferred options – holiday example

The interviewer continues:
“Now please do the same for the second pile, (poor/unattractive options), with the most preferred at the top
and the least preferred at the bottom”
Again the interviewer must ensure the recorded sequence coincides with the order in which the
respondent has sorted the cards. The grid on the questionnaire should now have the following hypothetical
appearance:

19
Download free eBooks at bookboon.com


An Outline of Conjoint Analysis

The Power of Conjoint Analysis

Overall most
preferred rank = 1

2

3

4

5

6

7

8

Card number

14

8

3

1

7

16

2

9

Order

9

10

11

12

13

14

15

16

Card number

5

13

5

10

4

11

12

15

Order

Fig. 14 Ranking of all options – holiday example

At the quality control stage, pre data entry, the data analyst should ensure there are:
• No card numbers are duplicated
• All cards have been rated
• There are no omissions in the grid
This particular process is not just limited to PAPI (paper assisted) questionnaires. They can also be
applied to online questionnaires. The analytical procedure remains the same; it’s just the method of
presenting the options to online respondents that differs. One tried and tested method we have used
recently is to present each group of options in turn, on a screen, and ask the respondent to give a score
of between 1 and 100, the higher the score the more desirable the group of options is. The reason for
requesting a score of between 1 and 100 is the likelihood of a respondent giving the same score twice
is relatively remote compared with a score of between 1 and 20. In the event of an identical score being
awarded to a previous range of options, the programme will prompt with a message requesting that a
different score be awarded.
To use our holiday example, we now have scores for each of the 16 options evaluated:
Option

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

Score

92

90

45

72

36

10

85

12

1

60

70

23

18

100

40

38

Fig. 15 Online ranking of all options – holiday example

The rating of these options shows that there is a high preference for some of these options, (options 1,
2, 4, 7, 11, 14), an indifferent level of preference, (options 3, 10, 15 and low levels, (options 5, 6, 8, 9,
12, 13, 16).
What we can now do is sort these scores from highest to lowest to enable us to have a ranking of the 16
options presented to the respondent as is demonstrated in the table below:

20
Download free eBooks at bookboon.com


An Outline of Conjoint Analysis

The Power of Conjoint Analysis

Score

100

92

90

85

72

70

60

45

40

38

36

23

18

12

10

1

Order

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

Option number

14

1

2

7

4

11

10

3

15

16

5

12

13

8

6

9

Fig. 16 Online ranking of all options sorted from most preferred to least preferred

We now have a ranking which resembled that derived from the PAPI engagement method. Both databases
can now be integrated to increase the sample size and thereby add value to the analysis.
We now have a full dataset of preference rankings for all 16 options to enable us to build a rating based
model based on evaluating the interrelationships between each of the 16 service levels using linear
regression analysis. A full understanding of conjoint requires years of experience in mathematical
statistics, computer programming and spreadsheets. Full understanding however is not required to
conduct and interpret conjoint analysis. It is only essential to understand the two key outputs of conjoint
analysis:

21
Download free eBooks at bookboon.com

Click
Click on
on the
the ad
ad to
to read
read more
more


An Outline of Conjoint Analysis

The Power of Conjoint Analysis

• Utility values – these are a measure of quantitative values attached to each level within each
attribute. Utility values, in essence, are a ‘common currency’ translating attractiveness and

desirability of relationships between levels into a positive score or, conversely, the lack of
‘fit’ between other relationships between level sets into a quantitative negative value. Utility
values are obtained from an ordinary linear regression analysis using the rank data as the
dependent variable and the profile designs/cards as the independent variables.
• These utility values, analogous to the coefficients of correlation derived from the regression
analyses, are called part-worths and can be used to evaluate the relative importance of the
attributes to which they belong.
The results show which combinations of levels generate the highest levels of attractiveness or
desirability and which particular attributes most influence preference and the relative importance
of each attribute.
Utility values are initially calculated at respondent levels. The regression is repeated for each respondent’s
data. The data analysis, once completed, can be averaged over all the respondents to show the average
utility value for each level within each attribute. Please note the importance and utility scores are purely
for illustrative purposes – this also applies to the case studies in the next chapter.
With these data, we can now map out the importance scores for each attribute and the utility scores for
each level.
So let us populate our holiday model with some hypothetical importance and utility scores:
Utility scores
ATTRIBUTE

UTILITY SCORES

Price

£200

£500

£1,000

£1,500

Importance 38.2%

2.17

4.10

3.48

1.18

February

Easter

May

August

Importance 26.4%

-3.04

0.64

2.13

1.8

Accommodation

Tent

Apartment

Hotel

Villa

Importance 23.8%

-2.46

1.85

3.73

1.64

En-suite Bathroom

Near Beach

Near Golf Course

Near Bullring

5.23

4.61

-0.87

-2.82

Time of Year

Amenities
Importance 11.6%

Fig. 17 Importance and utility scores for holiday model

22
Download free eBooks at bookboon.com


An Outline of Conjoint Analysis

The Power of Conjoint Analysis

Firstly, let us have a look at the relative importance score:
Attribute

%

Price

38.2

Time of Year

26.4

Accommodation

23.8

Amenities

11.6

Total

100.0
Fig. 18 Importance scores sorted from most to least preferred

Statistically, in terms of the sample size, price is the most important attribute. There is no statistically
significant difference between Time of Year and Accommodation Type, so both are equally important,
though not as important as price. Amenities is the least important attribute. So one can conclude from this
that Price, Time of Year and Accommodation Type are the three key drivers in the decision making process.
SPSS has also computed the individual utility scores for each level within each attribute. What we need
to establish is which combinations of levels derive the highest utility scores to enable the executives of
the holiday company to formulate their offer(s) to the marketplace.
Remember, earlier in this chapter, we stated that the four levels within these four attributes yields 256
permutations – (4x4x4x4). We are now in a position to simulate the highest possible levels of utilities
based on all the 256 combinations or permutations of the four levels within the four attributes.

23
Download free eBooks at bookboon.com

An Outline of Conjoint Analysis

The Power of Conjoint Analysis

We would strongly advise an analyst (or end user of these data) to set up a spreadsheet and map out all
256 permutations. This sounds a daunting task but there is a logic behind ensuring all 256 are mutually
exclusively covered without any duplications. Here is some guidance:
1. In row 1 of a spreasheet, type in your headings – PRICE, TIME, ACCOMMODATION,
AMENITY, UTILITYp, UTILITYt, UTILITYacc, UTILITYam
2. Enter the lowest price level £200 into Cell A2 and its utility score 2.17 into Cell E2
3. Copy these Cells down to A65 and E65 respectively
4. In Cell A66 enter £500 with its utility score 4.10 into Cell E66
5. Copy these cells down to A129 and E129 respectively
6. Repeat these processes for £1,000 (utility score 3.48) and for £1,500 (utility score 1.18) by
copying and pasting or dragging these values down 64 Cells
7. Enter ‘February’ in cell B2 and its utility score -3.04 into cell F2
8. Drag or copy these down 15 cells in turn into cells B17 and F17 respectively
9. Enter ‘Easter’ into cell B18 and its utility score 0.64 into cell F18
10.Drag or copy these down 15 cells into cells B38 and F38 respectively
11.Type ‘May’ into cell B34 and its utility score 2.13 into cell F34
12.Drag or copy these down 15 cells into cells B49 and F49 respectively
13.Type ‘August’ into cell B50 and its utility score 1.8 into cell F50
14.Drag or copy these down 15 Cells into cells B65 and F65 respectively

Turning a challenge into a learning curve.
Just another day at the office for a high performer.
Accenture Boot Camp – your toughest test yet
Choose Accenture for a career where the variety of opportunities and challenges allows you to make a

difference every day. A place where you can develop your potential and grow professionally, working
alongside talented colleagues. The only place where you can learn from our unrivalled experience, while
helping our global clients achieve high performance. If this is your idea of a typical working day, then
Accenture is the place to be.
It all starts at Boot Camp. It’s 48 hours
that will stimulate your mind and
enhance your career prospects. You’ll
spend time with other students, top
Accenture Consultants and special
guests. An inspirational two days

packed with intellectual challenges
and activities designed to let you
discover what it really means to be a
high performer in business. We can’t
tell you everything about Boot Camp,
but expect a fast-paced, exhilarating

and intense learning experience.
It could be your toughest test yet,
which is exactly what will make it
your biggest opportunity.
Find out more and apply online.

Visit accenture.com/bootcamp

24
Download free eBooks at bookboon.com

Click

Click on
on the
the ad
ad to
to read
read more
more


An Outline of Conjoint Analysis

The Power of Conjoint Analysis

15.Now copy and paste B2:B65 and F2:F65three times commencing with cells B66 and F66
respectively till you have completely filled all 256 rows
16.Type ‘Tent’ into Cell C2 and its utility score -2.46 into Cell G2
17.Drag or copy these down 3 Cells to C5 and G5 respectively
18.Type ‘Apartment’ into cell C6 and its utility score 1.85 into cell G6
19.Drag or copy these down 3 cells into cells C9 and G9 respectively
20.Type ‘Hotel’ into cell C10 and its utility score 3.73 into cell G10
21.Drag or copy these down 3 Cells into cells C13 and G13 respectively
22.Type ‘Villa’ into Cell c14 and its utility score 1.64 into cell G14
23.Drag or copy these down 3 Cells into cells C17 and G17
24.Now copy and paste C2:C17 and G2:G17 into cells C18 and G18 respectively till you have
completely filled all 256 rows
25.Type ‘Bathroom’ into cell D2 and its utility score 5.23 into cell H2
26.Type ‘Beach’ into cell D3 and its utility score 4.61 into cell H3
27.Type ‘Golf ’ into cell D4 and its utility score -0.87 into cell H4
28.Type ‘Bullring’ into cell D5 and its utility score -2.82 into cell H5
29.Now copy and paste D2:D5 and H2:G5 into cells D6 and H6 respectively till you have

completely filled all 256 rows
30. E
 nter a new column heading in Cell I1 and name it ‘Total Utility’. Now simply add across
the utility scores in Cells E2:H2 and drag down this entry till you have reached Cell I257
31.Now sort the spreadsheet from highest to lowest Total Utility
Of the 256 aggregated utility scores, now let us have a look at what the top ten combined utilities and
the bottom ten are telling us:
Option

Description

Total Utility

1

Hotel with en-suite bathroom in May @ £500

15.19

2

Hotel with en-suite bathroom in August @ £500

14.86

3

Hotel with en-suite bathroom in May @ £1,000

14.57

4

Hotel near beach in May @ £500

14.57

5

Hotel with en-suite bathroom in August @ £1,000

14.24

6

Hotel near beach in August @ £500

14.24

7

Hotel near beach in May @ £1,000

13.95

8

Hotel with en-suite bathroom at Easter @ £500

13.70

9

Hotel near beach in August @ £1,000

13.62

10

Apartment with en-suite bathroom in May @ £500

13.31

Fig. 19 Top ten aggregated utility scores for holiday model

25
Download free eBooks at bookboon.com