An Explore-Then-Commit Strategy for Revenue Management via Sequential Estimation

Abstract

We introduce and analyze a sequential price experimentation approach within the framework of Explore-Then-Commit strategies, which effectively balances the trade-off between learning about demand and maximizing revenue. We assume that the demand function follows a parametric family with unknown parameters and derive a closed-form stopping rule for learning the demand function. Our method allows decision-makers to adaptively determine when to stop gathering information and set prices, eliminating the need to specify a fixed sample size in advance. We demonstrate that our proposed strategy achieves asymptotically optimal min-max regret. This means that, even in the worst-case scenario where an adversary selects demand parameters to maximize the retailer's regret based on sample size, our approach successfully mitigates this risk. However, this level of robustness does not hold for strategies with a predetermined sample size. Furthermore, we extend our policy to a parametric choice model and provide numerical evidence showing that our approach outperforms fixed sample size policies.

Keywords: Revenue management, sequential estimation, data-driven decision making, pricing, asymptotic analysis

JEL Classification: C44