Empirical Estimation of Behavioural Biases in Spare Parts Inventory Management

In increasingly automated supply chains, human planners are often assisted by (advanced) algorithms to make inventory management decisions. We investigate the role of human planners in spare parts inventory management. Spare parts demand is characterized by very low and intermittent demands. This means that in general, the decision is to only have a limited number of items on stock. The resulting stockout probabilities are usually very low. We worked together with a large Original Equipment Manufacturer (OEM) in the semiconductor industry to answer these questions. We find that planners deviate from the optimal solution provided by the algorithm. Empirically, using structural estimation, we find a behavioral explanation for these adjustments. We identify that stockout costs are significantly overestimated by planners. We show that using the estimated parameters in the original algorithm will yield a solution which is directionally in line with, but more efficient than, the human planner adjustments.

Joan Stip is a part-time PhD candidate at the department of Industrial Engineering at Eindhoven University of Technology. He earned his MSc and PD Eng degree at the same school. His research interests are on service logistics, inventory management, human algorithm interaction, and simulation and optimization methodologies. Next to his research, he is working at ASML as a supply chain engineer in the customer supply chain management department.