Have an idea?

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

Is the HB analysis for a ACBC accurate if you only include screener questions?

I would like to use the ACBC screener format questions only for a survey I am building. Originally, I was planning on working with the raw data for analysis outside of Sawtooth, but I decided to test a few responses in Sawtooth's HB analysis anyway to see what would happen. I was surprised to see that average utilities and importances still appeared in the analysis manager. Can those results be trusted or does the HB for ACBC only work accurately when BYO, screener, and choice tasks are all included?
asked Oct 4, 2018 by carohut (245 points)

1 Answer

+1 vote
 
Best answer
The ACBC/HB program will work well (under the right conditions below) if only using the Screener section...or if only using the Choice Tournament section.

The Screener section alone is equivalent to a binary choice experiment, where the question "asked" for each product profile is, "Is this concept a possibility for you or not".  In other words "choose/not choose" for each concept.  It's like a card-sort conjoint where respondents sort an array of conjoint cards into just two buckets.

If skipping the BYO section, then ACBC uses a balanced and near-orthogonal design, similar to any other traditional conjoint experiment.  That's a great starting point.

If you use a *healthy number of concepts (product profiles), then the Screener section alone should do pretty well...to the degree that respondents don't say that nearly all the concept are "not a possibility" or that all the concepts are "a possibility".

Imagine that all your respondents say that all cards are "not a possibility".  All you will learn is that the None parameter (the "not a possibility" parameter) is extremely high and nothing else matters (there is no difference among all the attributes and levels in your study).

If all respondents say that all cards are a possibility, then all you learn is that the None parameter is extremely low and nothing else matters (there is no difference among all the attributes and levels in your study).

*Healthy number of cards asked for a card-sort conjoint or binary choice conjoint:

[(Total Number of Levels)-(Total Number of Attributes)+1]*2
answered Oct 4, 2018 by Bryan Orme Platinum Sawtooth Software, Inc. (201,165 points)
selected Oct 4, 2018 by carohut
Oops, I originally typed "-1" in the last equation, when I really meant to type "+1"
...