Creating a slotting model based on a previous run Maxdiff project with LCA

Question

Hi all,

I have a project which is a repeat from a couple of years ago when a colleague did the project. When first ran she did a latent class segmentation with 5 classes, based on a maxdiff with 24 items, with 15 rotations (4 items in each).

Now we would like to build a slotting model based on the HB scores and LCA segments, so we can cut the number of items and rotations and use predictive analysis to slot in the new respondents in the excisting segmentation model.

Does anyone have any experience with this?

Answer 1

This is commonly called developing a "Typing Tool" for MaxDiff, where you develop a short MaxDiff questionnaire that only has (typically) about 3 to 5 "golden" MaxDiff questions (only using a subset of the items from the original questionnaire).  The answers to these "golden" MaxDiff questions are used to assign respondents with a high degree of accuracy into the original segmentation scheme that had been developed from the full questionnaire.  This assignment can be done on-the-fly at the moment of taking the questionnaire, or afterwards.

Our "Sawtooth Analytics" consulting group has done many projects like this over the last 5 years.  There are other consulting agencies who also regularly do this.  If you are interested in following up with our Sawtooth Analytics group about this potential project, please contact Keith Chrzan at keith@sawtoothsoftware.com.