Rafinezhad Mostafa: "Uncovering Retailers’ Inventory Policies"

Managing inventory effectively is a critical challenge for retailers, as both excessive and insufficient stock levels result in financial and operational inefficiencies. Accordingly, operations management research has extensively investigated optimal inventory policies under different assumptions. Many of these policies have been embedded into ERP systems, which are now widely used in practice. However, despite the central role of inventory policies in order fulfillment, little research has been conducted that aims to understand which of these policies are applied by retailers, which could have important implications for demand planning of suppliers. This study aims to bridge this gap by systematically analyzing real-world inventory policies using a large empirical order dataset.

We utilize data from a leading global wholesaler in the travel retail market, which supplies hundreds of retailer customers across diverse segments, including airports, ships, and border shops. The dataset comprises order histories for tens of thousands of SKUs across multiple product categories, spanning two years. The study’s primary objectives are: (1) to characterize the inventory policies of the retailers based on observed order patterns and (2) to identify deviations from standard policies to understand the factors influencing non-standard ordering behaviors.

To achieve these objectives, the study employs a machine learning-based approach, integrating Recurrent Neural Networks. By this approach, the model processes raw sequential order data directly, enabling it to capture demand fluctuations and replenishment patterns over time. Preliminary findings indicate that real-world inventory policies frequently deviate from theoretically optimal models due to various operational and behavioral constraints. Retailers often override automated replenishment systems in response to unforeseen demand fluctuations, supply chain disruptions, and promotional activities.

This study contributes to the field of operations management by presenting a novel data-driven approach to characterize inventory policies and ordering behaviors. The findings have practical implications for both retailers seeking to refine their inventory decision-making processes and suppliers aiming to optimize stock planning through more accurate order forecasting.

Mostafa Rafienezhad started his PhD at Kühne Logistics University in November 2023. His research focuses on investigating new approaches and methodologies to advance the use of data and analytics for demand planning. Prior to joining KLU, Mostafa studied Logistics & Supply Chain Management at the University of Luxembourg, part of Massachusetts Institute of Technology's SCALE Network.