Observed and unobserved preference heterogeneity in brand-choice models

This paper extends the scanner-based choice literature by explicitly incorporating individual-level brand-preference data. We illustrate our model using a unique data set that combines survey and scanner data collected from the same individuals. The addition of individual-specific brand-preference information significantly improves fit and prediction. Furthermore, this "observed" heterogeneity better explains choice than does "unobserved" heterogeneity in the standard scanner model's parameters. More importantly, we find that the standard model underestimates the importance of consumers' brand preferences and overestimates both brand loyalties and price sensitivities. Brand loyalty is overestimated because models without preference information confound state dependence, heterogeneity, and preference effects. Price sensitivities are inflated because the "average" preference-based consumer is implicitly assumed to be more willing to switch from his preferred brand than is the "real" preference-based consumer. Further, standard models overestimate the heterogeneity in price and loyalty sensitivities and misidentify both price- and loyalty-sensitive consumers. The managerial implications of our findings and the applicability of our methodology when survey data are collected infrequently and for only a subsample of consumers are pursued. We demonstrate that even under these circumstances better populationwide pricing and promotion decisions are identified and more accurate targeting results.