Optimal Feature-Based Market Segmentation and Pricing

We study semipersonalized pricing policies in which a seller uses customer features to segment the market and offer segment-specific prices. Although such policies are common, determining the optimal segmentation and corresponding prices is computationally challenging, leading practitioners to rely on heuristic methods that first segment the market and then optimize the prices for each segment. We address this issue by studying the joint optimization of feature-based market segmentation and pricing strategies under the natural assumption that the seller has trained a regression model which maps customer features to valuations. When the noise in this regression model is independent across features and satisfies a monotone hazard-rate (MHR) condition, we show that the optimal market segmentation and pricing problem can be computed efficiently. We then analyze structural properties of the optimal feature-based market segmentation and pricing, quantify its performance relative to feature-based personalized pricing strategies, and give theory to guide practitioners in choosing the number of segments to use. Finally, we conduct a case study using home mortgage data and show that our approach can achieve nearly all of the available revenue using only a few well-chosen segments, significantly outperforming heuristic methods.

Supplemental Material: All supplemental materials, including the code, data, and files required to reproduce the results, are available at https://doi.org/10.1287/opre.2022.0490.