Managing inventories for perishable e-groceries: The value of probabilistic information

E-grocery retailing allows customers to order products online for delivery within a chosen future time slot. To remain competitive, retailers aim to meet high customer expectations regarding product availability by strategically setting very high service level targets. However, maintaining excess inventory incurs holding costs and leads to spoilage of perishable products, with associated environmental impacts. Retailers face multiple sources of uncertainty, including stochastic customer demand, stochastic spoilage, and potential supply shortages. This renders the determination of optimal replenishment quantities both challenging and crucial for long-term business growth. Fortunately, comprehensive new data sets routinely collected by retailers enable a data-driven approach to controlling inventory levels. This approach includes predictive and prescriptive analytics to (1) estimate suitable underlying probability distributions to represent the inherent uncertainty in the inventory process and to (2) integrate those forecasts into a comprehensive multi-period optimisation framework. In this paper, we propose a stochastic lookahead policy to solve the corresponding optimisation problem, thus supporting the retailers' inventory management decisions by minimising expected costs while maintaining a specified service level target. By explicitly deriving the value of probabilistic information, we provide guidance for retailers on which sources of uncertainty warrant investments in data collection and processing.