Does overrepresentation frequency of brands affect results?


Likely yes; however, these effects would be small in the aggregate analysis, and can usually be ignored. Significant overrepresentation of brands in a conjoint (i.e. 80% of all products are from one brand) may bias respondents simply due to the frequent exposure, but the effect in the results would generally be so small that they can be ignored.

Conducting market research experiments requires reliable data, so it’s desirable to gather the most realistic information and data. Nevertheless, some points have to be taken into account:

  • The choice task is meant to realistically mimic, not recreate the real-world purchase decision.
  • In general, conjoint does not attempt to account of real-world factors such as shelf availability, promotion frequency, advertising effectiveness, and other factors. Nevertheless, it can still consistently deliver actionable and verifiable results. Furthermore, it is possible to apply availability adjustments to conjoint simulations in an attempt to match simulated preference shares to observed volume shares.
  • We assume that a “well informed” customer reads each available brand independent from its shelf exposure.