Have an idea?

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

Estimating without the Dual Response Question

Hi - I am running a partial profile study with a dual response option. Is there any way that I can estimate utilities without including the dual response question? Do I have to manipulate the CHO? If so, how?

The motivation behind this being that we want to see the difference between preference of products to their adoption.
asked Nov 6, 2013 by Hem Chander

1 Answer

+2 votes
This is a tricky question because of all the issues involved.

By "dual response option" you are referring to the option to use the "dual-response None".  With dual-response None, respondents are asked if they would really purchase the concept they just chose within the CBC task.

First, you should be aware that the proper scaling of the None parameter is problematic in partial-profile CBC.  Respondents often often have difficulty saying whether they would buy or not a product concept if that concept is only described using a subset of the attributes.   That said, the counter argument is that the None parameter just represents a relative propensity or not for respondents to be wanting to "buy" the product in question, so we aren't really dealing in absolutes anyway, even if we were using full-profile CBC.

I'm assuming you are asking this question because you haven't yet fielded this study and are wondering if you need the "None" parameter or not.  Specifically, you state that you need to "see the difference between preference of products".  You can do that in market simulator analysis without the use of the None.

But, then you use the three words "to their adoption".  If you are trying to use results of CBC to estimate what percent of respondents will actually adopt the product (buy the product), then this is not something that CBC can do well.   Respondents are notoriously bad whether using the "None" category in CBC or whether using traditional 5-point purchase intent scales in standard concept testing, at telling researchers if they are really going to purchase a product or not.   So, using CBC to forecast actual adoption rates is tricky and involves some history and outside knowledge to bring to the equation.  If you've run many CBCs before in your product category and have been able to compare the predicted adoption rates to actual later sales (by including the dual-response None in your CBC experiments), then you'll probably get a good feel for how declared "None" rates in CBC exercises actually translate to real purchases in the market place.   

There are issues of awareness, distribution, advertising, length of time on the market, etc. that are needed to predict actual adoption rates for products.  CBC alone doesn't include those issues.
answered Nov 6, 2013 by Bryan Orme Platinum Sawtooth Software, Inc. (131,990 points)