Learning Stochastic Binary Feedback on a Sampled Hierarchical Belief Model: Optimal Pricing of Contracts in the Truckload Trucking Market

Abstract

We apply optimal learning to the truckload trucking market to understand industry phenomena and optimize revenue through contract pricing. In the trucking market, brokers arrange contracts between two firms, the trucking service provider (the carrier) and the customer (the shipper). The shipper and the carrier have competing and disguised pricing objectives which makes their responses to contract offers stochastic and binary. The broker’s challenge is to offer contract prices that are accepted by both firms. In the learning problem, we optimize and learn from the perspective of the broker and seek to learn the true pricing objectives of both firms. In the mathematical model, we first develop a family of logistic functions that describe a firm’s probabilistic response to a particular contract price. We design a sequential decision making model to learn the true responses of each firm for a single trucking route between two locations, primarily using the knowledge gradient policy for sampled beliefs. The single route model is then expanded, adapted, and used in conjunction with a hierarchical model to efficiently learn the responses of each firm over a multiple route trucking network. We demonstrate that the use of these models significantly increases broker revenue. Overall, this research presents a powerful framework for learning and optimal pricing in the truckload trucking market.