AI-driven Policy Optimization For Transparent And Operationally Effective Pricing

Abstract

Interpretability and Operational Effectiveness are two practical limitations of current AI based price optimization. The rationale behind price changes is unclear and the projected gains from offline studies often disappear in live tests. We present an automated, dynamic prescriptive policy optimizer that combines 'Blackbox' machine learning and mathematical programming models to identify near-optimal and transparent pricing recommendations that can perform effectively in practice.