Machine Learning Applications in Retail Price Optimization: Balancing Profitability with Customer Engagement

Abstract

There is a growing academic interest in the development of machine learning (ML) models applied to price optimization. Historically, retail prices were set with exogenous reference to competitors, a base price was selected for the promoted product, and a simple markup on cost was applied to all items in a product category. The emergence of e-commerce provided retailers with both opportunity and challenge, as they faced competitive pressures to adjust prices across thousands of products on an hourly basis. Price optimization is now being pursued using methods ranging from simple heuristics to sophisticated, machine-learned price models based on historical sales data. ML models can mine pricing-relevant relationships and patterns from transactional data generated by the ongoing management of price changes. Demand models can then simulate the impact of price adjustments on sales volume, revenue, and profit (or margin). Price models can recommend a price to be applied to the future and/or a scheduled price adjustment to capitalize on the forecasted demand change. Pre-emptive pricing models can be used to profitably pre-empt a new competitor or the next move from an existing competitor.