How to build up MaxDiff simulator

Question

Hi

My client wants to apply MaxDiff method to examine features importance for its product. But there is a question raised by him. Is there any MaxDiff simulator generated or produced after analysis? I reviewed the website and couldn't find MaxDiff simulator. It would be grateful if someone could guide me how to produce MaxDiff simulator.

Client's objective is:
1) view importance at total level and different subgroups;

2) for different subgroups, the top 10 most important features in sequential order (from most important to least important)

3) for different subgroups, I need to find the importance of optimal feature combinations which meets specific value (e.g. the importance of feature combinations are sum to 80% (this 80% could be varied))

4) in optimal feature combinations calculation, some features cannot be revealed at the same time (e.g. we have two battery life features, one is 3-hr battery life and the other one is 5-hr battery life. In feature combinations calculation, these two features cannot display simultaneously)

Answer 1

We have an online tool at www.maxdiffanalyzer.com that is probably what you are looking for.  Once you have estimated your item scores, you can save them to a CSV and upload them to a project at that site.  You can also upload a CSV of segmentation variables in order to split out the scores, simulations, and optimization runs by those variables.

So, #1 is taken care of by the segmenting

#2 would require sorting, which currently isn't implemented on the fly.  But, once I turned on a segmentation variable to look at my item scores, I can download that summary to a CSV and sort it easily in Excel, for example

For #3, you might be interested in the TURF optimization capabilities.  TURF stands for total unduplicated reach and frequency, and is a neat analysis approach for MaxDiff items.  Say, for instance, I can only stock 5 ice cream flavors.  I might run a MaxDiff on 20 flavors, and one option would be to just choose the top 5 flavors on average.  But, if the same people who like flavor 1 also like flavor 2, TURF would allow me to stock the 5 flavors that "reaches" the greatest number of people.  Since MaxDiff scores are continuous, rather than discrete, there are a few options for informing TURF how someone is counted as "reached' or not.  But, there is not a specific option to make things sum to 80, though I think that would be pretty hard to pick the "right" number to target (too arbitrary)

For #4, TURF supports both prohibitions (items cannot appear together in a bundle) as well as forcing specific items to a bundle (if I have to have vanilla, what are the best other 4 flavors to include).

The MaxDiff Analyzer is $250 per year for an account, and you can sign up for a free 30-day trial at www.maxdiffanalyzer.com