Price Customization and Within Chain Data do not mix!
- Rakesh Niraj
- S. Siddarth
European Journal of Marketing
We propose that purchase histories available from store loyalty cards may not yield accurate estimates of consumer preferences and marketing mix responsiveness and that purchase data from across-store purchase histories is essential. We find support for this thesis by creating both across-store and store-specific purchase histories for a large sample of households in a scanner panel and showing that a standard latent class model of consumers’ purchase incidence, brand choice and purchase quantity decisions estimated on the former dataset outperforms the equivalent model estimated on the latter dataset. Recognizing that retailers may potentially have access to purchase data on thousands of consumers in their loyalty card programs, we also examine how parameter recoverability is affected by the number of households in the sample. We find that relatively small samples of the across-store data recover parameter estimates fairly well but that even large samples drawn from the store-specific data may not. We use the parameters derived from the store-specific dataset to develop customer-specific prices for both retailer and manufacturer targeting. Surprisingly, the profits from these targeting strategies are lower than those realized by delivering the optimal common price-cut to all customers, i.e., without any customization, but derived from the model estimated on across-store purchase histories. These results suggest that it may be suboptimal for retailers and manufacturers to use store-specific information from loyalty cards to develop targeted price promotions.