Have an idea?

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

Comparing utility of brand vs. total utility

Hi all,

Methodological question: assuming you have run a CBC with HB analysis to get point estimates per respondent per attribute level, and your attributes look like the following:

Brand: Brand A, Brand  B
Cost: <four price points>
Reward: <four reward points levels>

Can you, justifiably, do the following per respondent?
- Note which part-worth is higher for which brand (i.e. which brand is preferred)
- Take the non-preferred brand, and determine what combination (sum) of part-worths for the other levels result in the total utility exceeding their part-worth for their preferred brand, and then conclude that that is what is required to cause them to switch brands?

Is this methodologically justifiable or is it breaking the rules/assumptions of CBC?

asked Jul 10, 2012 by anonymous
retagged Apr 21, 2014 by Walter Williams

1 Answer

0 votes
Best answer
To get results you want, you should compare full products of preferred and unpreferred  brand.
I suppose you should compare with real market offer of preferred brand. So if we state that preferred brand is B1+C1+R1, you should choose utilities C2 and R2 utilities for brand B2 to make it's product utility u(B2) + u(C2) + u(R2) > u(B1) + u(C1) + u(R1). There will be a lot of such combinations probably, so you should use some optimization rule e.g minimal cost for company. As you don't have many levels of your attributes you could conduct full search, but you don't have to. Especially if your attributes are ordered.
answered Jul 10, 2012 by lkomenda Bronze (2,830 points)