CONJOINT ANALYSIS

[DATE]
[COMPANY NAME]
[Company address]
CONTEN
T
 Introduction
 Types of Conjoint Analysis
 Key Terms for Conjoint Analysis
 Application For Conjoint Analysis
 Example of conjoint analysis using excel
 Example of conjoint analysis using SPSS
 Merits and Demerits of Conjoint Analysis
 Litigations
 References
 Introduc
Market Research is
tion
Frequently concerned with finding out which
characteristics of a product or service is most important to consumers.
Now a days for any person there is very much options available to choose not
only in electric appliance almost in every field even in health lines they prefer
multiple different types of drugs with their own importance so at this time
what to prefer is more concerned and for that a manufacturer(company) also
are concerned that they give best preferred product to the customer and at
that time the use of conjoint analysis came to into existence this method help
us which attributes does customer prefer the most and which type of product
a company should made this analysis help to find that
Definition
(i) “Conjoint analysis is a set of techniques ideally suited to studying
customers’ choice processes and determining trade-offs.
(from applied conjoint analysis by V.R.Rao)(Book)
(ii) “Conjoint analysis is a survey-based statistical technique used in market
research(mainly) that helps determine how people value different attributes
(feature, function, benefits) that make up an individual product or
service."(Wikipedia)
A controlled set of potential products or services is shown to survey
respondents and by analysing how they make preferences between these
products, the implicit valuation of the individual elements making up the
product or service can be determined. These implicit valuations(utilities or
part-worths) can be used to create market models that estimate market
share , revenue and even profitability of new designs .

The objective of conjoint analysis is to determine what combination of a

limited number of attributes is most influential on respondent choice or
decision making.
In simple language we can say that the conjoint analysis will help us to find the
importance of each attribute as per the consumer’s choice.

Conjoint analysis has its origins in psychology and was developed by Robert
Luce and John Tukey in 1964.
Conjoint analysis originated in mathematical psychology and was developed
by marketing professor Paul E. Green at the Wharton School of the University
of Pennsylvania. Other prominent conjoint analysis pioneers include
professor V. "Seenu" Srinivasan of Stanford University who developed a linear
programming (LINMAP) procedure for rank ordered data as well as a self-
explicated approach, and Jordan Louviere (University of Iowa) who invented
and developed choice-based approaches to conjoint analysis and related
techniques such as best–worst scaling.
Types of
Conjoint
Choice-Based Conjoint (CBC) Analysis: This is one of the most common
Analysis
forms of conjoint analysis and is used to identify how a respondent
values combinations of features.
Adaptive Conjoint Analysis (ACA): This form of analysis customizes each
respondent's survey experience based on their answers to early
questions. It’s often leveraged in studies where several features or
attributes are being evaluated to streamline the process and extract the
most valuable insights from each respondent.
Best/Worst Conjoint: Similar to choice-based conjoint, this method
shows respondents a set of concepts. In each set, respondents are asked
to pick the most/least (or best/worst) concepts. This approach is used
when a product or service has features that cause both positive and
negative reactions.
Full-profile conjoint analysis: This method displays many concepts and asks
respondents to rate each one based on the likelihood of purchase. This
method is outdated and was primarily used prior to the introduction of survey
tools that offer choice-based conjoint. Asking to rate lots of concepts at once
is error-prone, quickly causes fatigue, and yields low-quality data.

Rating or Ranking Conjoint: Ranking and rating conjoint was the method used
for full-profile conjoint analysis. As software has progressed, it is now possible
to conduct rating or ranking conjoint similar to a choice-based conjoint.
Respondents are shown a set of concepts and asked to rank or rate each
concept. They could rank by entering a value for each concept, which sums to
100 for each set, or they could enter a number based on a scale. This method
is also sometimes referred to as "Continuous Sum Conjoint".

Menu-Based Conjoint: Menu-based conjoint is a new conjoint method. This

method gives respondents the ability to pick multiple levels from a menu. For
example, a car manufacturer could ask respondents to choose a base model
and price, just like choice-based conjoint. But then they could also ask to
check a box for each additional feature desired such as "Alloy Wheels for
$1,500", "Sunroof for $1,000", or "Parking Assist for $1,500".

Partial-profile versus full-profile conjoint analysis

The pairwise ranking task referred to above can be simplified even more by
defining each hypothetical alternative included in a choice set on just two
attributes (or criteria) at a time. This type of conjoint analysis is known as
‘partial-profile’ conjoint analysis.
In contrast, full-profile conjoint analysis is based on choice sets where the
alternatives are defined on all attributes together at a time (e.g. six or more).
Examples of partial- and full-profile choice sets respectively appear in Figure 1
and 2 below (for a conjoint analysis into smartphones).

Pairwise ranking full-profile choice sets is more cognitively difficult (and

slower) than pairwise ranking partial-profile choice sets. Therefore, all else
being equal, conjoint analysis results obtained using full-profiles are less likely
to be accurate, albeit such choice sets are arguably more realistic

Figure 1: Example of a partial-profile choice set

Figure 2: Example of a full-profile choice set
Key Terms For Conjoint Analysis
Attributes (Features): The product features being evaluated by the analysis.
Examples of characteristics for Laptops: Brand, Size, Color, and Battery Life.
Levels: The specifications of each attribute. Examples of standards for Laptops
include Brands: Samsung, Dell, Apple, and Asus.
Task: The number of times the respondent must make a choice. The example
shows the first of the five functions as indicated by “Step 1 of 5.”
Concept or Profile: The hypothetical product or offering. This is a set of
attributes with different levels that are displayed at each task count. There are
usually at least two to choose from.
Relative importance: “attribute importance,” which depicts which of the
various attributes of a product/service is more or less important when making
a purchasing decision. Example of Laptop Relative Importance: Brand 35%,
Price 30%, Size 15%, Battery Life 15%, and Color 5%.
Part-Worths/Utility values: Part-Worths, or utility values, is how much weight
an attribute level carries with a respondent. The individual factors that lead to
a product’s overall value to consumers are part-worths. Example part-worths
for Laptops Brands: Samsung – 0.11, Dell 0.10, Apple 0.17, and Asus -0.16.
Profiles: Discover the ultimate product with the highest utility value. At a
glance, QuestionPro lets you compare all the possible combinations of product
profiles ranked by utility value to build the product or service that the market
wants.
Market share simulation: One of the most unique and fascinating aspects of
conjoint analysis is the conjoint simulator. This gives you the ability to
“predict” the consumer’s choice for new products and concepts that may not
exist. Measure the gain or loss in market share based on changes to existing
products in the given market.

Brand Premium: How much more will help a customer pay for a Samsung
versus an LG television? Assigning price as an attribute and tying that to a
brand attribute returns a model for a $ per utility distribution. This is
leveraged to compute the actual dollar amount relative to any characteristic.
When the analysis is done relative to the brand, you get to put a price on your
brand.
Price elasticity and demand curve: Price elasticity relates to the aggregate
demand for a product and the demand curve’s shape. We calculate it by
plotting the demand (frequency count/total response) at different price levels.
Softwar
e to be
used
Application
s of
Conjoint analysis in marketing

Classic questions that marketers are looking for solutions to involve which
features to offer in a product or service, how to price that product, and who to
target with each of multiple offerings.

Conjoint analysis can help answer all these market research questions. It
allows you to test thousands of potential configurations and prices for your
product using a single market research survey.

Conjoint analysis for pricing research

Figuring out how much to charge for a product/service that delivers value to
customers while maximizing revenue or profit is the age-old business
dilemma.

Directly asking people their willingness to pay in pricing research is

problematic due to its difficulty to answer, yielding subjective, unreliable
responses.

How much are you willing to pay to double the storage on your next phone or
laptop, for example? It's almost an impossible question to answer without a
realistic context involving competitive offerings with differing storage and
prices.

Given the choices gathered across a sample of respondents, we can tease out
the value of a product’s different features and conduct choice simulations to
estimate price sensitivity, willingness to pay, and overall demand for different
product configurations.
Conjoint analysis in product development

Product managers want to understand their customers’ needs as well as the

competitive landscape to help them bring to market the right product at the
right price.

A/B tests are limited regarding how many product variations may be tested.
With conjoint analysis, thousands of product configurations may be tested
using a single conjoint survey. We can tease out the importance of features
and predict the likely acceptance of just about any possible product
configuration and price.

And the end result? An optimal product package ready for successful launch.

Conjoint analysis for branding, package design, and product claims

Because conjoint analysis presents products in a realistic way, much like

buyers would see them in the real world, the researcher can effectively test a
multitude of package design and the accompanying claims/branding.

Due to the complex and aesthetic nature of package design, interactions

(synergies) between design, colors, and messages can and do occur.

Choice-Based Conjoint (CBC) can detect and model many kinds of interaction
effects with precision, making it an excellent solution for optimizing branding,
package design and product claims.

At our recent conference, researchers from Procter & Gamble showed how
they use our tools for packaging and product claims.

Conjoint analysis for needs-based market segmentation

Conjoint analysis is widely used for conducting market simulations, predicting

demand, and estimating price sensitivity for various product configurations.
But, we can also use conjoint data to group respondents based on their
differing preferences. Our conjoint software delivers easy-to-use algorithms
for grouping respondents that minimize the differences in preference within
the groups while maximizing the differences in preference between the
groups.

The needs-based segments can be cross-tabulated against other survey data

to profile attractive segments and find the greatest opportunities for target
market success. Specifically, you can design unique products to appeal to key
market segments to enhance your competitiveness in the marketplace.

Conjoint analysis in healthcare

Health economists use conjoint analysis and other kinds of choice experiments
to understand the tradeoffs patients and healthcare providers make in
choosing medical treatments, which often feature decisions about the quality
and length of life.

This is well illustrated by the Patient Preference Initiative, a framework

developed through a public/private partnership between medical technology
companies and the Food and Drug Administration (FDA) to collect patient
preference data, that recommends conjoint analysis and MaxDiff as good
solutions for creating quantitative assessments of the relative desirability or
acceptability to patients of different medical treatment approaches.
Conjoint Analysis in Transportation Economics

Many of the innovations we rely on today in our conjoint and choice studies
started off in the transportation literature, including the Nobel-prize winning
work of Dan McFadden, who invented the statistical model on which modern
conjoint analysis and choice experiments rely.

Conjoint analysis helps transportation researchers understand how travelers

and urban commuters value and tradeoff various aspects of their travel
experience.

To this day the transportation research literature remains one of the best
places to find ongoing academic work on choice-based conjoint methods.

Conjoint analysis in environmental economics

Like many traditional marketing projects, environmental impact studies often

seek to quantify both the demand and revenue of certain projects with the
costs and potential impact of the affected area.

Conjoint analysis can be an extremely helpful tool in these situations to model

both sides of a give-and-take relationship between investment and
environmental conservation.
The state of North Carolina surveyed a massive 33,000 hunters to get their
input on deer hunting season changes to help raise the natural population of
bucks.

The direct feedback from constituents trading off realistic scenarios (you can't
have the longest season with the most tags at the cheapest price) helped
guide lawmakers in an extremely practical way.

Example
Using
 Excel
In this example, we assume the reader has a basic understanding of multiple
regression analysis.

A traditional conjoint analysis is really just a multiple regression problem. The

respondent’s ratings for the product concepts form the dependent variable.
The characteristics of the product (the attribute levels) are the independent
(predictor) variables. The estimated betas associated with the independent
variables are the utilities (preference scores) for the levels. The R-Square for
the regression characterizes the internal consistency of the respondent.
Consider a conjoint analysis problem with three attributes, each with levels as
follows: Brand
ABC

Color Red Blue

Price
$50
$100
$150

For simplicity, let’s consider a full-factorial experimental design. A full-

factorial design includes all possible combinations of the attributes. There are
(3)·(2)·(3) = 18 possible product concepts (commonly called cards) that can be
created from these three attributes. Further assume that respondents rate
each of the 18 product concepts on a score from 0 to 10, where 10 represents
the highest degree of preference.
Assume the data for one respondent are as follows (as if in an Excel
spreadsheet):

A B C D E
1 Card# Brand Color Price Preferen
ce
2 1 1 1 1 5
3 2 1 1 2 5
4 3 1 1 3 0
5 4 1 2 1 8
6 5 1 2 2 5
7 6 1 2 3 2
8 7 2 1 1 7
9 8 2 1 2 5
10 9 2 1 3 3
11 10 2 2 1 9
12 11 2 2 2 6
13 12 2 2 3 5
14 13 3 1 1 10
15 14 3 1 2 7
16 15 3 1 3 5
17 16 3 2 1 9
18 17 3 2 2 7
19 18 3 2 3 6

The first card is made up of level 1 of each of the attributes, or (Brand A, Red,
$50). The respondent rated that card a “5” on the preference scale.

After collecting the respondent data, the next step is to code the data in an
appropriate manner for estimating utilities using multiple regression. We use
a procedure called dummy coding for the independent variables (the product
characteristics). In its simplest form, dummy coding uses a “1” to reflect the
presence of a feature, and a “0” to represent its absence. For example, we can
code the Brand attribute as three separate columns.

Brand A Brand B Brand C

If Brand is “A”, then dummy 1 0 0
codes =
If Brand is “B”, then dummy 0 1 0
codes =
If Brand is “C”, then dummy 0 0 1
codes =
Applying dummy-coding for all attributes results in an array of columns as
follows:

A B C D E F G H I J
1 Card # A B C Red Blue $50 $100 $150 Preferen
ce
2 1 1 0 0 1 0 1 0 0 5
3 2 1 0 0 1 0 0 1 0 5
4 3 1 0 0 1 0 0 0 1 0
5 4 1 0 0 0 1 1 0 0 8
6 5 1 0 0 0 1 0 1 0 5
7 6 1 0 0 0 1 0 0 1 2
8 7 0 1 0 1 0 1 0 0 7
9 8 0 1 0 1 0 0 1 0 5
10 9 0 1 0 1 0 0 0 1 3
11 10 0 1 0 0 1 1 0 0 9
12 11 0 1 0 0 1 0 1 0 6
13 12 0 1 0 0 1 0 0 1 5
14 13 0 0 1 1 0 1 0 0 10
15 14 0 0 1 1 0 0 1 0 7
16 15 0 0 1 1 0 0 0 1 5
17 16 0 0 1 0 1 1 0 0 9
18 17 0 0 1 0 1 0 1 0 7
19 18 0 0 1 0 1 0 0 1 6

Again, we see that card 1 is defined as (Brand A, Red, $50), but we have
expanded the layout to reflect dummy coding.

To this point, the coding has been very straightforward. But, there is one
complication that must be resolved. In multiple regression analysis, no
independent variable may be perfectly predictable based on the state of any
other independent variable or combination of independent variables. If so,
the regression procedure could not separate the effects of the confounded
variables. We have that problem with the data above, since, for example, we
can perfectly predict the state of brand A based on the states for brands B and
C. This situation is termed linear dependency.
To resolve this linear dependency, we omit one column from each attribute. It
really doesn’t matter which column (level) we drop, and for this example we
have excluded the first level for each attribute, to produce the modified data
table below:

A B C D E F G
1 Card # B C Blue $100 $150 Preferen
ce
2 1 0 0 0 0 0 5
3 2 0 0 0 1 0 5
4 3 0 0 0 0 1 0
5 4 0 0 1 0 0 8
6 5 0 0 1 1 0 5
7 6 0 0 1 0 1 2
8 7 1 0 0 0 0 7
9 8 1 0 0 1 0 5
10 9 1 0 0 0 1 3
11 10 1 0 1 0 0 9
12 11 1 0 1 1 0 6
13 12 1 0 1 0 1 5
14 13 0 1 0 0 0 10
15 14 0 1 0 1 0 7
16 15 0 1 0 0 1 5
17 16 0 1 1 0 0 9
18 17 0 1 1 1 0 7
19 18 0 1 1 0 1 6

Even though it appears that one level from each attribute is missing from the
data, they are really implicitly included as reference levels for each attribute.
The explicitly coded levels are estimated as contrasts with respect to the
omitted levels, which are defined as “0.”

Microsoft ExcelTM (we have used Excel from Office 2000 in this example) offers
a simple multiple regression tool, under Tools + Data Analysis + Regression
(you must have installed the Analysis Toolpak add-in). Using the tool, specify
the preference score (column G) as the dependent variable (the Input Y
Range) and the five dummy-coded attribute columns (columns B through F) as
independent variables (the Input X range).
You should also make sure a constant is estimated (this usually happens by
default). The mathematical expression of the model is as follows:
Y = b1(Brand B) + b2(Brand C) + b3(Blue) + b4($100) + b5($150) + constant + e

where:
Y = respondent’s preference for the product concept,
b1 through b5 are beta weights (utilities) for the features, e is an error term,
and
the reference levels are equal to “0.”
The solution minimizes the sum of squares of the errors over all observations.
A portion of the output from Excel is as follows:

SUMMARY OUTPUT

Regression Statistics

Multiple R 0.9489
02
R Square 0.9004
15
Adjusted R
Square 0.8589
21
Standard 0.9428
Error 09
Observations 18

Coefficients Standard Error t StatP-value Lower 95% Upper 95% Lower

95.0% Upper 95.0%
Intercept 5.833333 0.54433 10.716 1.69E- 4.64733 7.01932 4.64733 7.0193
1 52 07 8 9 8 29
X 1.666667 0.54433 3.0618 0.00986 0.48067 2.85266 0.48067 2.8526
Variable 1 62 5 1 2 1 62
1
X 3.166667 0.54433 5.8175 8.24E- 1.98067 4.35266 1.98067 4.3526
 Variable 1 38 05 1 2 1 62
2
X 1.111111 0.44444 2.5 0.02791 0.14275 2.07947 0.14275 2.0794
Variable 4 5 2 72
3
X -2.16667 0.54433 - 0.00182 - - -3.35266 -
Variable 1 3.9804 5 3.35266 0.98067 0.9806
4 2 7
X -4.5 0.54433 - 2.68E- -5.686 -3.314 -5.686 -3.314
Variable 1 8.2670 06
5 3

Using that output (after rounding to two decimals places of precision), the
utilities (Coefficients) are:

Brand
A 0.00
B 1.67
C 3.17
Color
Red 0.00
Blue 1.11
Price
$50 0.00
$100 -2.17
$150 -4.50

The constant is 5.83, and the fit for this respondent (R-Square) is 0.90. The fit
ranges from a low of 0 to a high of 1.0. The standard errors of the coefficients
(betas) reflect how precisely we are able to estimate the betas with this
design. Lower standard errors are better. The remaining statistics presented
in Excel’s output are beyond the scope of this paper, and are generally not of
much use when considering individual-level conjoint analysis problems.
Notes:

One can easily generalize how many parameters (independent

variables plus the constant) are involved in any conjoint analysis
problem as #Levels - #Attributes + 1. Most traditional conjoint
analysis problems solve a separate regression equation for each
respondent. Therefore, to estimate utilities, the respondent must
have evaluated at least as many cards as parameters to be
estimated. When the respondent answers the minimum number of
conjoint cards to enable estimation, this is called a saturated design.
While such a design is easiest on the respondent, it leaves no room
for respondent error. It also always yields an R-square of 100, and
therefore no ability to assess respondent consistency.

Most good conjoint designs in practice include more observations

than parameters to be estimated (usually 1.5 to 3 times more). The
design above has three times as many cards (observations) as
parameters to be estimated. These designs usually lead to more
stable estimates of respondent utilities than saturated designs.

Also note that in practice (except with the smallest problems), asking
respondents to evaluate all possible combinations of the attribute
levels is usually not practical. Design catalogs and computer
programs are available to find efficient fractional factorial designs.
Fractional factorial designs show just an efficient subset of the
possible combinations, and still provide enough degrees of freedom
to estimate utilities.

The standard errors for the Color attribute are lower than for Brand
and Price (recall that lower standard errors imply greater precision of
the beta estimate). Because Color only has two levels (as compared
to three each for Brand and Price), each color level has more
representation within the design. Therefore, more information is
provided for each color level than is provided for any level of the
three-level attributes.
Data Analysis using SPSS Example
Here we have taken a example of a mobile company who wants to
manufacture a new mobile with multiple different attributes like we
have considered 4 attributes in the following mobile as:-
Storage:- (64,128)GB
Camera:- (20,64)MP
Processor:- (754,788,810)Snapdragon
Price:- (16000,18000,19000)
Now as we see that from this attributes which one customer would
like to prefer that’s Important for mobile company and which
attribute would have more importance for customer to find out that
is company`s moto but now as we can see there are many
combinations that company could give to the customers like in the
above case we have total combinations:-
Storage(2)*Camera(2)*Processor(3)*Price(3)=36 so there are total 36
combinations of mobile which company could give but company cant
make all types of mobile so which one would be the best for that
company can prepare a 36 type mobile and then give it to the
customers for ratings but that’s very time consuming and for
customers its boring so we can get same information form less type
of combinations in SPSS there is a feature called orthogonal designs
so it would see all the combinations and would recommend less
preference for the customer
Now Here as we see by using Orthogonal designs we have got 9
combinations of mobile from 36 combinations of mobile that
customer would like to prefer so just now circulate the following
template and ask all the customers just to fill the template by giving
the following rank 1 to 9 which would they prefer first and last rank
wise and after completion of this process enter the data in new spss
data set as shown in below figure
As seen above we have taken eight customers review they have
given the rank as they want as we know we have nine combination
we have named it has P1 to P9 now after getting these we have
conjoint function in SPSS which is inbuilt function now just we have
to write the syntax and it will give us the output
So here In above we could see we have run the syntax we would get
the output as
As seen above the summary is very big but here we have shown the
important part that is utlity so as seen above in camera the 64 MP
camera is given more preference and as in storage 128 GB is
preferred more and processor 710 and price as 16000Rs so by seeing
above thing company should go for these type of model if its not
feasible for them then the relative importance is also given in which
it is saying that customer prefer more to camera , Processor , price
and last storage so if company cant give 128 gb then it can give some
model in 64gb storage .
 Merits
and
Demerit
s of
Advantages[edit]
estimates psychological tradeoffs that consumers make when

Conjoint
evaluating several attributes together
can measure preferences at the individual level

Analysis
uncovers real or hidden drivers which may not be apparent to
respondents themselves
mimics realistic choice or shopping task
able to use physical objects
if appropriately designed, can model interactions between attributes
may be used to develop needs-based segmentation, when applying
models that recognize respondent heterogeneity of tastes
Disadvantages[edit]
designing conjoint studies can be complex
when facing too many product features and product profiles,
respondents often resort to simplification strategies
difficult to use for product positioning research because there is no
procedure for converting perceptions about actual features to
perceptions about a reduced set of underlying features
respondents are unable to articulate attitudes toward new
categories, or may feel forced to think about issues they would
otherwise not give much thought to
poorly designed studies may over-value emotionally-laden product
features and undervalue concrete features
does not take into account the quantity of products purchased per
respondent, but weighting respondents by their self-reported
purchase volume or extensions such as volumetric conjoint analysis
may remedy this
 Litigations
Federal courts in the United States have allowed expert witnesses to
use conjoint analysis to support their opinions on the damages that
an infringer of a patent should pay to compensate the patent holder
for violating its rights.[4] Nonetheless, legal scholars have noted that
the Federal Circuit's jurisprudence on the use of conjoint analysis in
patent-damages calculations remains in a formative stage. [5]
One example of this is how Apple used a conjoint analysis to prove
the damages suffered by Samsung's copyright infringement, and
increase their compensation in the case.

Courtyard by Marriott is a brand of hotels owned by Marriott

International. This mid-priced range of hotels designed for business
travelers, also accommodate families. Marriott used conjoint analysis
to design a hotel services.
Referen
ces
 https://en.wikipedia.org/wiki/Conjoint_analysis
 https://sawtoothsoftware.com/conjoint-analysis
 https://www.1000minds.com/conjoint-analysis/what-is-
conjoint-analysis
 https://www.qualtrics.com/au/experience-management/
research/types-of-conjoint/
 https://www.surveyking.com/help/conjoint-analysis-explained
 https://www.questionpro.com/blog/what-is-conjoint-analysis/
 https://online.hbs.edu/blog/post/what-is-conjoint-analysis
 Book:- Conjoint Analysis By V.R.RAO