Choose Language Hide Translation Bar
Community Manager Community Manager
Learn Conjoint (Discrete Choice) Experimental Design and Analysis

Conjoint designs are planned experiments that determine the features people want in a product and what they are willing to pay for them. Market researchers show respondents a controlled set of potential products or services and then analyze how respondents make preferences.


On April 12, Leo Wright demonstrated how to use JMP for conjoint design and analysis using a case study on a new laptop product with various hard disk size, processor speed, battery life and cost attributes. He showed how to create the design, run the model, analyze the model, use the JMP Profiler for trade-off analysis and incorporate respondent demographics to identify possible product choices by gender and job type.


In this study, he found that the company could probably charge either $1000 or $1200 for an attractive product and that, for this product, females are more price conscious than males.



Use Term Estimates from test survey to populate Prior Means.


Points people found interesting were:

1. Prior Mean allows you to exclude the feature combinations, or dominating alternatives, that are clearly the most attractive (in this case, large disk, fast processor, longest battery life and lowest cost). One can guess at prior means or assign them using the parameter estimates received from a test survey answered by a small set of respondents.

2. During interactive what-if analysis, the JMP Profiler Utility Score is helpful for determining the ideal feature set and what the respondents believe is an acceptable price for an enhanced set of features.



Postive Utility score shows that customers are willing to spend $1,200 for this feature set.


Want to try this on your own? The Discrete Choice Design case study and instructions are available in JMP in the DOE Guide, Chapter 8, p. 161. (Help>Books>JMP DOE Guide). The laptop data files are accessible from JMP 8 menu Help>Sample Data Files.


Interested in other information about Design of Experiments? Consider signing up for one of our live DOE Webcasts.





Article Labels

    There are no labels assigned to this post.


Marcel wrote:

I would be interested in the issue of excluding feature combinations with prior mean as is mentioned in the blog entry. There is nothing about it in the handbooks. I would really appreciate any help on that subject.


Martin Demel wrote:

Hi Marcel,

In this community entry you'll find an example where the Prior mean Setting creates such a low probability value that you almost exclude a certain Feature combination:

May this helps you further,