A Statistical Learning Approach to Personalization in Revenue Management

We develop a general framework for modeling decision problems in which actions can be personalized by taking into account the information available to the decision maker. We demonstrate the application of our method to customized pricing and personalized assortment optimization. We show that learning under our model takes place reliably by establishing finite- sample convergence guarantees for model parameters which hold regardless of the number of customer types, which can be potentially uncountable. The parameter convergence guarantees are then extended to performance guarantees in decision problems. In particular, we provide high-probability bound on the gap between the expected revenue of the best action taken under the estimated parameters and the revenue generated by a decision-maker with full knowledge of the demand distribution. We conduct simulated experiments to demonstrate the performance of our method. We further test our method on real transaction data for airline seating reservations and show that effective customized pricing can increase revenue by at least 7% over the best single-price.