Data-Driven Optimization for Commodity Procurement under Price Uncertainty
We study a practice-motivated multi-period stochastic commodity procurement problem under price uncertainty with forward and spot purchase options. Existing approaches are based on parametric price models, which inevitably involve price model misspecification and generalization error. We propose a non-parametric, data-driven approach (DDA) that is consistent with the optimal procurement policy structure but without requiring the a-priori specification and estimation of stochastic price processes. In addition to historical prices, DDA is able to leverage real-time feature data, such as economic indicators, in solving the problem. This paper provides a framework for prescriptive analytics in dynamic commodity procurement, with optimal purchase policies learned directly from data as functions of features, via mixed integer linear programming (MILP) under cost minimization objectives. Hence, DDA focuses on optimal decisions rather than optimal predictions. Furthermore, we combine optimization with regularization from machine learning (ML) to extract decision-relevant data from noise. Based on numerical experiments and empirical data, we show that there is a significant value of feature data for commodity procurement when procurement policy parameters are learned as functions of features. However, overfitting deteriorates the performance of data-driven solutions, which asks for ML extensions that improve out-of-sample generalization. In a practical application, compared to an internal best practice benchmark, DDA would have generated savings of on average 9.1 million Euros p.a. (4.33%) for ten years of backtesting and potential savings of 7.7 million Euros over 18 months after handing over the project results and DDA tools. A practical benefit of DDA is that it yields simple but optimally structured decision rules that are easy-to-interpret and easy-to-operationalize. Furthermore, DDA is generalizable and applicable to many other procurement settings.
Stefan Minner is a Full Professor for Logistics and Supply Chain Management at the School of Management, Technical University of Munich, Germany. Before, he held positions at the Universities of Paderborn, Mannheim, and Vienna. He studied Business Administration with specialization in Operations Research at the University of Bielefeld, received his doctoral degree from the Otto-von-Guericke University in Magdeburg, and was a postdoctoral researcher at the University of Calgary. His primary research interests are logistics network design under uncertainty, transportation optimization including smart mobility concepts, and inventory management. Recent projects include logistics for the automotive industry, retail operations, and last-mile city logistics. In his consulting work, Stefan Minner has cooperated with several national and international companies from various industries. Stefan Minner serves on several editorial boards of logistics and operations research journals. Currently, Stefan Minner is the Editor-in-Chief of the International Journal of Production Economics. His research was published in many peer reviewed journals, including Manufacturing & Service Operations Management, Operations Research, Production and Operations Management, Transportation Research Part B, European Journal of Operational Research and OR Spectrum. In 2014, he was listed among the top 25 most productive researchers in Business Administration in Germany by Handelsblatt. He is a fellow of the International Society for Inventory Research (ISIR), vice-chairman of the scientific advisory board of the German Logistics Association (BVL), a member of the Research Committee of the European Logistics Association (ELA), and the speaker of the research training group “Advanced Optimization in a Networked Economy (AdONE)” at Technical University of Munich.
Professor Stefan Minner
4:30 - 5:30 pm