Personalized Pricing and the Value of Past Purchase Histories: An Empirical Perspective

This paper quantifies the value of purchase history data for behavior-based personalized pricing in grocery retail. We estimate a structural demand model with latent consumer types across 24 product categories within a large US supermarket. Using the transaction data from 17,756 consumers enrolled in a loyalty program, we conduct counterfactual experiments in which the supermarket updates its beliefs about consumer preferences using Bayes's rule and sets personalized prices accordingly. Personalized pricing based on purchase history increases profits by approximately 4% across all categories and has asymmetric effects on consumers. While most consumers enjoy a small discount over uniform prices, a small fraction of price-insensitive consumers face substantially higher prices under personalization. These findings have important implications for the design of optimal loyalty programs: the supermarket benefits from charging a small membership fee and only the most price-insensitive consumers opt out of the program.