Pricing and Monetization Strategies Readings & Articles:
Pricing Readings Lecture 2:
Do it yourself conjoint:
Conjoint Analysis is most commonly used quantitative market research method to quantify
consumer preferences for products and services. Applying conjoint analysis has 6 steps:
1. Select attributes
2. Select levels for these attributes
3. Create product profiles
4. Collect data
5. Estimate partworths
6. Derive insights and make predictions
Step 1: Select attributes
Any conjoint analysis will only be useful to the extent that it will enable better decision
making, in terms of what product configuration and/or price to offer. Thus, the attributes
need to have the potential to sway consumers’ choices and can actually be changed or
controlled by the firm.
Do’s and Don’ts:
- Do not use too many attributes: keep number below 6. Increasing the number of
attributes increase the burden on consumers in 2 ways: (i) it creates a need for
longer surveys, (ii) it makes each question harder to answer.
- Focus on the attributes upon which managerial decisions need to be made:
collecting additional info on must-have attributes will not lead to better decisions.
Could use an introduction that informs them of the values they should assume to
avoid assumptions.
- Do not use subjective attributes and do not use ambiguous level descriptors: style,
aesthetic appeal, good, modern, should all be avoided or levels that are defined as
ranges. The more room left for interpretation by using subjective/ambiguous
attributes, the more likely their true preferences are not measured by responses.
- Do not use infeasible combinations: do not use vegetarian yes/no and then choose
between fish, chicken, beef, because that is infeasible.
Step 2: Select levels for the attributes included
The levels should be selected to allow effective decision making on the part of the firm and
be meaningful to consumers (and respondents to the survey).
Do’s and Don’ts:
- Keep the number of levels similar across attributes.
- Do not use highly unrealistic levels: no point of including levels of attributes that
respondents will not encounter in practice or that they find incredulous.
- Do not use “confounded” levels: if an attribute level is defined as meaning one of
several things (e.g., round or square) or as meaning the combination of several things
(e.g., flat and black), then it is impossible to determine which of these components
drive preferences.
, - Compare apples to apples: make sure that the different levels of the same attribute
only differ on 1 dimension. If you use 10€ a month and 80€ a year, you will not be
able to conclusively determine whether the preference comes from framing or
different amounts.
- Use mutually exclusive levels: define binary attributes.
Step 3: Create product profiles
The standard format of Rating-Based Conjoint Analysis, in which consumers are asked to
provide independent ratings of various profiles, on a numeric scale. Which is convenient to
administer, and it yields data that are easy to analyse using standard software.
Alternative format which became more popular is Choice-Based Conjoint Analysis, in which
consumers are asked to make a series of choices between 2 or more profiles. While more
realistic, this format gives rise to data that may only be analysed using advanced statistical
packages.
It is highly recommended to use existing libraries or software packages for optimising
profiles. Designing the survey by yourself, and specifically which profiles to include, without
consulting such resources may lower the statistical power of your questionnaire.
SAS (statistical software package) can be used to create optimal sets of profiles
corresponding to any number of attributes and levels. The function %mkturns will help you
decide how many profiles should be included in your questionnaire and the function %mktex
creates the set of profiles. For example, if there are 5 attributes with 3, 3, 4, 4, and 3 levels
respectively, %mkturns (3 3 4 4 3) will generate a few reasonable number of profiles to
include in the questionnaire; and %mkturns (3 3 4 4 3, n=24) will create an optimal set of 24
profiles.
Step 4: Collect data
Data may be collected using traditional paper and pencil questionnaires. Alternatively, you
may conduct your questionnaire online. Today, it is very easy and affordable to run a
conjoint analysis survey online (Qualtrics, Surveymonkey).
Step 5: Estimate Partworths
Estimate consumers’ preferences (partworths) based on the data. We follow the standard
approach of using a linear regression to link a consumer’s ratings of the various profiles in
the questionnaire to their preferences for various attribute levels.
If data has been collected on a small number of consumers, it is easy to run 1 regression for
each consumer (using Excel).
If data have been collected from a larger sample of consumers, it may be tedious to estimate
each consumers’ preferences separately using Excel.
One option is to use other statistical packages that allow running a large number of
regressions with minimal effort (R).
Another option is to group consumers together and estimate one regression per group
(cluster analysis).
,For example, it is possible estimate one large regression for all consumers in the sample, of a
few regressions by splitting the sample into groups and running a separate regression for
each of them (e.g., age). So, if there are 50 consumers in a group and each consumer rated
15 profiles described by 10 attribute levels, this regression will have 50 x 15= 750
observations (rows) and 10 columns. The estimated partworths will be our estimate of the
average preference in the group.
Step 6: Derive insights and make predictions
Several insights and predictions may be derived once partworths have been estimated. For
example, WTP for a unit change may be estimated for any attribute tested in the survey.
Market shares may also be predicted for any set of alternatives described along the
attributes measured in the survey.
The companion spreadsheet allows making such predictions from partworth data you
supply.
Here we went with setting the baseline profile to a snack box with: Apple juice, Toblerone,
Peanut butter, Chips, $5
The ideal profile for this consumer is the profile that would have each attribute at the
consumer’s most preferred level.
, It is not recommended to use this approach to predict market shares among a group of
consumers. In that case it is preferable to run 1 regression for each consumer in the group
and use the companion spreadsheet “Market Outcome Predictor”.
Understanding Conjoint In 15 Minutes:
Conjoint analysis is used to determine what features a new product should have and how it
should be priced, and it is far less expensive and more flexible than concept testing.
New golf ball, there are 3 important product features:
1. Average Driving Distance
2. Average Ball life
3. Price
Range of feasible alternatives for each of these features, for instance:
Obviously, the market’s “ideal” ball would be:
And the “ideal” ball from a cost of manufacturing perspective would be:
Assuming that it costs less to produce a ball that travels a shorter distance and has a shorter
life
