JMP 13 Preview: New MaxDiff platform for consumer research
Aug 31, 2016 10:31 AM
MaxDiff (maximum difference scaling) is a new platform in JMP 13 that will be helpful to anyone who does consumer research. It enables a specialized type of choice model where respondents are asked to evaluate items (product attributes, …) in sets of three to five, choosing the most preferred and least preferred in each set.
This relatively recent methodology developed by Jordan Louviere et al. is very effective in ranking existing products and avoids the ambiguity respondents encounter when trying to do a forced ranking of a large number of items. The most and least important things are clear, but the relative rankings in the middle tend to be unclear.
MaxDiff can help rank items that don't easily break down into features.
This platform was added because JMP users who were doing consumer research said it was high on their priority list. They really wanted to stay in JMP to do this rather than use R or another software package.
Regular readers of the JMP Blog may recall the Potato chip smackdown: Winners and losers post a few years ago by Melinda Theilbar, JMP research statistician developer. That was actually a MaxDiff analysis using JMP 12, and it took a LOT of data work. So Melinda was motivated to make life easier for customers who need to do this kind of choice modeling.
In fact, she enjoys using MaxDiff regularly for a project outside of her JMP work. She co-founded Research Triangle Analysts, a nonprofit forum for data enthusiasts to promote useful techniques, meaningful analysis and effective communication of findings. Melinda surveys this group each year, asking members to rate talks, topics and speakers. MaxDiff lets her easily see what are the most and least important of these among members of this group.
While there are users who already appreciate what MaxDiff choice models yield, Melinda believes others will get access to a great new method they may not have had in their analytical tool set.
“It’s hard to get data on things you most need to measure. This method provides a way to get a handle on data that you really need, but is often difficult to effectively gather,” Melinda says.
We are very appreciative of the input we received during our Early Adopter program on this new platform, in particular from Walt Paczkowski, founder of Data Analytics Corp.:
"JMP continued to push the envelope in consumer research with the addition of two new MaxDiff modeling platforms in its newest release, JMP 13. One allows you to create a MaxDiff design matrix while the other gives you the ability to estimate MaxDiff utilities both at the aggregate and disaggregate levels. As should be expected from all the JMP platforms, these new additions are fully integrated with all the power and versatility of JMP making JMP the only software that allows you to design, estimate and analyze a range of choice models. Anyone who works in the consumer choice area will find these new additions to be a great advance for their work."
For Melinda, what's most exciting about MaxDiff is its ease of use and the opportunity to introduce our users to this technique: "Many could really benefit from it but may not have invested in the tools or the time needed to try it."
Now, it’s been implemented in an intuitive form you would expect in JMP. Melinda will blog about how to use Max Diff later this fall. And as we noted in another post about association analysis (which Melinda also worked on), she will be presenting at Discovery Summit next month on new features in choice modeling in JMP and JMP Pro.
For more about what's coming in JMP 13, visit the preview site.