Have an idea?

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

Volumetric CVA?

I have a question on a design like CVA. We have a governmental service without any competition and only one service is offered at the time.
So, it only makes sense to have tasks with a single offer. Now we like to estimate the effect of a different offer on purchase intent and the volume bought.

For purchase intent we can use the CVA/HB analysis, while for the volume I was thinking rather of the OLS.

Each offer is made up by three attributes.

Disadvantage is that we need many tasks to get sufficient data, leading to fatigueness. And the pairwise comparison is forced, so it seems we are study with the old-fashioned single concept CVA.

Unless anyone has other suggestions?
asked Sep 14, 2017 by Andy

1 Answer

0 votes
CVA with single-concept design will work, as long as the volumes that the respondent will specify are in the range 0-999.  The software will not support values larger than 999.

You can use either OLS or HB-OLS estimation.  Just make sure in the Analysis Manager to click the gear button (the settings button), then select the Scale link, then specify that Recode= "None".

With the dependent variable in its natural scale (0-999), meaning Recode="None", the sum of the raw part-worth utilities for the product concept will be the least-squares fit to the volume the respondent specified.

In other words, don't use zero-centered diffs rescaling to make the predictions of volume.  Build your own simulator in Excel and compute the total predicted volume per respondent as simply the sum of the raw part-worths.

Or, if using our Advanced Simulation Module simulator, I think there is a way to tell it to report "total utility" rather than "shares of preference" and in that case our simulator will simply add the raw part-worths and report the total utility (which for you means total volume) for any product concept you specify in the simulator.
answered Sep 14, 2017 by Bryan Orme Platinum Sawtooth Software, Inc. (132,290 points)
Thank you, Bryan! That helps.