Pagina |1
Conjoint Analysis
Lecture 1
Introduction
Conjoint analysis is needed to have a better understanding of the following:
- how to successfully design new products
- how to reposition products that are not working out fine
- how to design optimal pricing strategies
- how to develop a successful product line / portfolio
With the latter, one can think of a completely new product line or an extra addition to an
existing product line.
Every year, many new products and services are introduced, but only a small fraction
succeeds. This is a major issue for companies since a product launch is very expensive; a lot
of investments have to be made. Associated costs are advertising costs, R&D costs, costs to
convince retailers to carry your product(s) and the possible costs to your image. R&D is
needed for companies in many different industries to keep up with what is going on in the
market and to stay ahead of competitors.
Conjoint analysis is needed since product decisions are becoming increasingly complex and
there is a lot of money at stake. A product is characterised by many, many, many attributes
but they are not all of equal importance to consumers; it differs per person. Nowadays,
consumers are often confronted with many different features / attributes that they have to
decide on; there are many different products. Trade-offs have to be made to decide which
product to choose. Keep in mind that with e.g. high technological products, the attributes
considered important do not all have to be highly technical as well. For example, the colour of
the product can also be found of importance.
An attribute is characterised by several levels: e.g. attribute is colour; the levels are black,
white, gold, etc.
When using conjoint analysis, we use a decompositional view since the utility of consumers is
dependent of the attribute levels of a product.
Example
Product: smartwatch
Product’s attributes: brand; bracelet colour
Attributes’ levels: Apple, Samsung; black, white, red
Product: a white Apple smartwatch
, Pagina |2
The appeal of the smartwatch = utility of Apple + utility of white
So, the utility of the stimulus is the summation of the utility of different attribute levels.
Choice-based conjoint analysis can be used to determine the best combination of product
attribute levels to form a successful product.
Latent-class segmentation can be used to segment consumers. This is needed because not all
consumers want the same; they have different preferences for e.g. brand, price.
So, collected market data can be used to perform conjoint analysis and market segmentation.
These results can be used to make a simulation (predict the success of different product
options) and predict market shares. When combining the latter with data on accompanying
costs, one can find the most profitable options.
Conjoint analysis – an introduction
Conjoint means to join together. It is a quantitative
research technique that asks respondents to either rank,
rate or choose among multiple products / services. This
can be done, respectively, by ranking-based, rating-based
and choice-based conjoint.
Products are represented as bundles of attributes; levels
of each attribute define the product. A stimulus is a
summation of different attributes, while an attribute is a summation of levels.
All forms of conjoint analysis evaluate different stimuli (=products).
Ranking-based conjoint
Choose the most-preferred product, then the second most-preferred product, … until the least-
preferred product. It is about an overall evaluation of all possible products and is followed by
a ranking on preference. This is the least interesting method.
Rating-based conjoint
With each task a customer has to rate one different stimulus. This can be done by giving the
stimulus a score on a scale of 0-10 or by using Likert scales.
Choice-based conjoint
With each task, a customer is shown multiple different products at once and the customer has
to choose the most-preferred product only. So each task means one choice; for each tasks
there is a set of different products. This set can contain 2 to 4 different products.
Managerial goals of conjoint analysis
- To unveil the relative importance of the different attributes of products/services, i.e.
determine how individuals make trade-offs between attributes of a product.
o Examples: between price and brand name, between colour and performance,
between price and performance, …
- To assess the preference for different levels of a given attribute, i.e. determine how
individuals value alternative options for an attribute
o Examples: black vs. red vs. white smartwatch?
Preference = utility
Part-worths = preference for levels of attributes
,Pagina |3
, Pagina |4
The main outcomes of conjoint analysis are:
- Importance (utilities) – which products attributes’ levels influence the purchase
decisions
- Preference shares – how do customers choose between different products in specific
market situations; do we observe different customer segments
- Market simulation – how should product attributes be changed to compete with the
products available on the market
Segmenting is possible to the heterogeneity in the data.
When planning on doing a conjoint analysis, a researcher should specify the attributes that
one uses to decide on when buying a product. The choice of attributes and levels of attributes
is crucial, as well as how well they are defined and presented to the respondents / customers.
It namely influences the outcome and quality of the research.
There is a difference in what consumers find of importance – e.g. resolution & digital zoom
vs. easy & elegance.
Conjoint analysis is a preferred research method because in direct surveys, respondents might
say they consider all attributes important. This is not informative for the researcher. Conjoint
enforces trade-offs between attributes. This is because all attributes are evaluated at once;
there are no dominant alternatives. Respondents evaluate “complete” products with both
strong and weak attributes. Furthermore, conjoint reduces problem of socially desirable
answers and it adds realism. The latter is because in real-life consumers evaluate products, not
isolated attributes (do they consciously know which attributes matter?). Lastly, conjoint
analysis is straightforward.
The disadvantages of conjoint analysis are that typically, only small number of attributes can
be included. Respondents need to process multiple attributes simultaneously and too much of
them can lead to overload – takes a lot of effort to process all. Respondents can only handle a
max of 6 to 7 attributes. When doing a proper pre-analysis, one can limit the number of
attributes, leading to more homogeneity and less correlation. Furthermore, the research
method assumes that the products one has to choose among, are done with high involvement.
It is assumed that respondents “know” the attribute levels, which is usually only possible after
search effort. To limit this disadvantage one can choose to explain the attributes in a
questionnaire.
Rating-based conjoint
With this form of conjoint, respondents are asked to evaluate hypothetical products on e.g. a
10-point scale. The number of hypothetical products depends on the amount of attributes one
wants to research as well as the levels of these attributes.
e.g. three attributes that have two levels each 2 × 2 × 2 = 8 possible products / stimuli.
The linear regression model is used to estimate the preferences for certain stimuli. The model
is as follows (the number of attributes can be different):
𝑅𝑎𝑡𝑖𝑛𝑔𝑖𝑗 =∝ +𝛽𝐴𝑡𝑡𝑟𝑖𝑏𝑢𝑡𝑒1𝑖𝑗 + 𝛾𝐴𝑡𝑡𝑟𝑖𝑏𝑢𝑡𝑒2𝑖𝑗 + 𝛿𝐴𝑡𝑡𝑟𝑖𝑏𝑢𝑡𝑒3𝑖𝑗 + 𝜀𝑖𝑗
Where: i = ith customer
j = jth stimuli
the attributes can either take the value 1 or 0 (dummy variables)
The 𝛽, 𝛾 𝑎𝑛𝑑 𝛿 in the regression reflect the part-worths.
Conjoint Analysis
Lecture 1
Introduction
Conjoint analysis is needed to have a better understanding of the following:
- how to successfully design new products
- how to reposition products that are not working out fine
- how to design optimal pricing strategies
- how to develop a successful product line / portfolio
With the latter, one can think of a completely new product line or an extra addition to an
existing product line.
Every year, many new products and services are introduced, but only a small fraction
succeeds. This is a major issue for companies since a product launch is very expensive; a lot
of investments have to be made. Associated costs are advertising costs, R&D costs, costs to
convince retailers to carry your product(s) and the possible costs to your image. R&D is
needed for companies in many different industries to keep up with what is going on in the
market and to stay ahead of competitors.
Conjoint analysis is needed since product decisions are becoming increasingly complex and
there is a lot of money at stake. A product is characterised by many, many, many attributes
but they are not all of equal importance to consumers; it differs per person. Nowadays,
consumers are often confronted with many different features / attributes that they have to
decide on; there are many different products. Trade-offs have to be made to decide which
product to choose. Keep in mind that with e.g. high technological products, the attributes
considered important do not all have to be highly technical as well. For example, the colour of
the product can also be found of importance.
An attribute is characterised by several levels: e.g. attribute is colour; the levels are black,
white, gold, etc.
When using conjoint analysis, we use a decompositional view since the utility of consumers is
dependent of the attribute levels of a product.
Example
Product: smartwatch
Product’s attributes: brand; bracelet colour
Attributes’ levels: Apple, Samsung; black, white, red
Product: a white Apple smartwatch
, Pagina |2
The appeal of the smartwatch = utility of Apple + utility of white
So, the utility of the stimulus is the summation of the utility of different attribute levels.
Choice-based conjoint analysis can be used to determine the best combination of product
attribute levels to form a successful product.
Latent-class segmentation can be used to segment consumers. This is needed because not all
consumers want the same; they have different preferences for e.g. brand, price.
So, collected market data can be used to perform conjoint analysis and market segmentation.
These results can be used to make a simulation (predict the success of different product
options) and predict market shares. When combining the latter with data on accompanying
costs, one can find the most profitable options.
Conjoint analysis – an introduction
Conjoint means to join together. It is a quantitative
research technique that asks respondents to either rank,
rate or choose among multiple products / services. This
can be done, respectively, by ranking-based, rating-based
and choice-based conjoint.
Products are represented as bundles of attributes; levels
of each attribute define the product. A stimulus is a
summation of different attributes, while an attribute is a summation of levels.
All forms of conjoint analysis evaluate different stimuli (=products).
Ranking-based conjoint
Choose the most-preferred product, then the second most-preferred product, … until the least-
preferred product. It is about an overall evaluation of all possible products and is followed by
a ranking on preference. This is the least interesting method.
Rating-based conjoint
With each task a customer has to rate one different stimulus. This can be done by giving the
stimulus a score on a scale of 0-10 or by using Likert scales.
Choice-based conjoint
With each task, a customer is shown multiple different products at once and the customer has
to choose the most-preferred product only. So each task means one choice; for each tasks
there is a set of different products. This set can contain 2 to 4 different products.
Managerial goals of conjoint analysis
- To unveil the relative importance of the different attributes of products/services, i.e.
determine how individuals make trade-offs between attributes of a product.
o Examples: between price and brand name, between colour and performance,
between price and performance, …
- To assess the preference for different levels of a given attribute, i.e. determine how
individuals value alternative options for an attribute
o Examples: black vs. red vs. white smartwatch?
Preference = utility
Part-worths = preference for levels of attributes
,Pagina |3
, Pagina |4
The main outcomes of conjoint analysis are:
- Importance (utilities) – which products attributes’ levels influence the purchase
decisions
- Preference shares – how do customers choose between different products in specific
market situations; do we observe different customer segments
- Market simulation – how should product attributes be changed to compete with the
products available on the market
Segmenting is possible to the heterogeneity in the data.
When planning on doing a conjoint analysis, a researcher should specify the attributes that
one uses to decide on when buying a product. The choice of attributes and levels of attributes
is crucial, as well as how well they are defined and presented to the respondents / customers.
It namely influences the outcome and quality of the research.
There is a difference in what consumers find of importance – e.g. resolution & digital zoom
vs. easy & elegance.
Conjoint analysis is a preferred research method because in direct surveys, respondents might
say they consider all attributes important. This is not informative for the researcher. Conjoint
enforces trade-offs between attributes. This is because all attributes are evaluated at once;
there are no dominant alternatives. Respondents evaluate “complete” products with both
strong and weak attributes. Furthermore, conjoint reduces problem of socially desirable
answers and it adds realism. The latter is because in real-life consumers evaluate products, not
isolated attributes (do they consciously know which attributes matter?). Lastly, conjoint
analysis is straightforward.
The disadvantages of conjoint analysis are that typically, only small number of attributes can
be included. Respondents need to process multiple attributes simultaneously and too much of
them can lead to overload – takes a lot of effort to process all. Respondents can only handle a
max of 6 to 7 attributes. When doing a proper pre-analysis, one can limit the number of
attributes, leading to more homogeneity and less correlation. Furthermore, the research
method assumes that the products one has to choose among, are done with high involvement.
It is assumed that respondents “know” the attribute levels, which is usually only possible after
search effort. To limit this disadvantage one can choose to explain the attributes in a
questionnaire.
Rating-based conjoint
With this form of conjoint, respondents are asked to evaluate hypothetical products on e.g. a
10-point scale. The number of hypothetical products depends on the amount of attributes one
wants to research as well as the levels of these attributes.
e.g. three attributes that have two levels each 2 × 2 × 2 = 8 possible products / stimuli.
The linear regression model is used to estimate the preferences for certain stimuli. The model
is as follows (the number of attributes can be different):
𝑅𝑎𝑡𝑖𝑛𝑔𝑖𝑗 =∝ +𝛽𝐴𝑡𝑡𝑟𝑖𝑏𝑢𝑡𝑒1𝑖𝑗 + 𝛾𝐴𝑡𝑡𝑟𝑖𝑏𝑢𝑡𝑒2𝑖𝑗 + 𝛿𝐴𝑡𝑡𝑟𝑖𝑏𝑢𝑡𝑒3𝑖𝑗 + 𝜀𝑖𝑗
Where: i = ith customer
j = jth stimuli
the attributes can either take the value 1 or 0 (dummy variables)
The 𝛽, 𝛾 𝑎𝑛𝑑 𝛿 in the regression reflect the part-worths.
