Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

Segmentation using MaxDiff

I'm considering using MaxDiff utilities as inputs for a segmentation.  However, the normal course of affairs for segmentation, after the segments have been finalized, is to develop a subset of questions (i.e., "golden" questions) for classifying new respondents. I'm curious though, if the segmentation was originally based on 30 items, but then that number was later cut down to 10 items (i.e, "golden questions"), wouldn't that affect the utilities? Which would in turn affect the algorithm and the probabilities of belonging to each segment?
asked Dec 17, 2012 by anonymous

1 Answer

0 votes
As I understand your question, Yes, would affect the utilities if the survey was run again with only the reduced set.

That said, after "ranking" set of attribute (level) utilities I suspect the top N would include the reduced set so perhaps all is not lost for entry into your segmentation model (which you haven't noted, also how you derived the reduced set?)
answered Dec 27, 2012 by zstatman (200 points)