Have an idea?

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

Latent Class vs. clustering by purchased product


I am conducting a CBC and also want to cluster the different target groups. Does it make more sense to...

1. run latent class?
2. analyze the data with HB and the SMRT market simulator, define as many clusters as there are competing products and assign respondents to the product/cluster with the highest purchase probability?

Option #2 seems better to interpret, however it is not "true" clustering but rather describing which respondent would buy which product...

Which option is more advisable?
Do I have to make any configurations when exporting the data to SPSS for cluster description?

asked Jul 26, 2015 by Ulf

1 Answer

0 votes

I think option 2 is a perfectly reasonable way to segment, if it makes sense with your analysis plan.  It may not derive "latent" segments using fancy statistics, but not all situations require that kind of segmentation.  Once you classify respondents via simulation you can profile them in SPSS as you would any other segments.
answered Jul 26, 2015 by Keith Chrzan Gold Sawtooth Software, Inc. (48,200 points)
Thanks for your answer! Do yo happen to know where in SMRT I can see which respondent would buy which product? I only get the shares per product when I run my simulations, but for the clustering part I would need the disaggregate values for each respondent as well...?
Yes, when you select a scenario, press the "edit" key.  Then select "Individual Results to File" option in the lower left corner.