Data-driven stochastic markdown optimization for fashion retail

Abstract

An effective price markdown strategy is important for any retailer to profitably liquidate seasonal fashion products with finite life-time. While conventional markdown approaches are largely rule-based or parametric, we propose an approach with two novel components: i) A data-driven price elasticity model that estimates future sales as function of offered discounts and other product and merchandizing attributes. ii) A dynamic programming based optimizer that recommends an optimal discount policy to be followed in the entire planning horizon, that maximizes the expected revenue and also allows for a pre-specified markdown budget. The proposed approach was piloted with a leading fashion retailer and yielded encouraging results.