Have an idea?

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

Calibration Weighting


I'm sure this is very straight forward but I've managed to confuse myself. I am looking to apply weighting to individual respondents based on their utility scores (From HB post Conjoint) to calibrate the overall results in line with the market.

At the moment I am getting the following results:
Product A: 30%
Product B: 40%
Product C: 10%
Product D: 20%

But the real life market is 22%, 45%, 15% and 18%.

How do I calculate the weight to apply to each respondent to reflect this?

asked Feb 26, 2018 by DaveT (245 points)

1 Answer

0 votes
Weighting respondents is only appropriate if you believe the main reason for the deviance between simulated shares of preference from conjoint analysis and market shares is that you did not sample the respondents properly.  However, this is usually not the main reason for differences between simulated shares of preference & market shares.  The more common differences are: lack of even distribution of brands across the market, lack of even awareness, missing attributes.

Adjusting simulated shares of preference to match market shares is a dangerous practice, and should only be taken after full consideration of the ramifications.

Our newest market simulator within Lighthouse Studio (or the standalone desktop version) provides a way to adjust the predicted shares of preference to match the actual market shares (but not using the weighting of respondent), should you want to pursue this route.  For documentation on that, please refer to: http://www.sawtoothsoftware.com/help/lighthouse-studio/manual/index.html?hid_externaleffects.html

On the other hand, if you really want to develop respondent-level weights such that simulated shares of preference match market shares, then a recommended way to go is to build your simulator in Excel using the "share of preference-logit" simulation rule.  Our Lighthouse Studio software also can export a beginning Excel simulator for you.  Just make sure to choose "share of preference" as the simulation method.  Then, modify your Excel simulator to allow individual-level respondent weights, such that the simulated share of preference is a weighted share of preference.  Then, you can use Excel's Solver tool to modify the weights for each respondent such that the simulated shares of preference match the target market shares.

Different approaches to adjust market simulators to match target market shares, including strengths and weaknesses are described in our 2006 white paper at: http://www.sawtoothsoftware.com/download/techpap/externaleffects.pdf

If you don't want to use a Solver tool, but would rather do the calculations to find the weighting for each respondent to lead to the target simulated shares of preference, this is described starting on page 12 of the article cited directly above.
answered Feb 26, 2018 by Bryan Orme Platinum Sawtooth Software, Inc. (198,315 points)
edited Feb 26, 2018 by Bryan Orme