14Case Study 5: Approximate Bayes and Personalized Pricing

Abstract

Approximate Bayes methods have become increasingly popular in recent years given the need for scalability and speed in business applications. This case study demonstrates the use of such methods in the context of personalized pricing.

With the wide availability of granular customer data there has been a push toward the scaling of decisions to the individual level so as to provide personalized marketing offerings. The core idea in this endeavor is to estimate heterogeneous effects using cross‐sectional data and then use these estimates to customize marketing mix elements. Typically, the challenge lies in the high‐dimensionality of the covariate set as well as the number of individuals which can be addressed using ML tools. In this case study, we examine the issue of personalization of pricing decisions using a approximate Bayes implementation of binary Logit model with L1 penalization.

14.1 HETEROGENEITY AND HETEROGENEOUS TREATMENT EFFECTS

The building block of any personalization scheme is heterogeneity. Typically, we think of heterogeneity in terms of the differences in how customers react to a particular marketing treatment. For example, different customers react differently to the same price or to the same advertising message. Effective personalization requires not only that customers exhibit such heterogeneity in responsiveness but also that the firm be able to offer differentiated treatments targeting ...

Get Bayesian Statistics and Marketing, 2nd Edition now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.