Publication Date
Authored by
Geoffrey Allen, Shoichi Ishida
Advisor(s): Eva Ponce
Topic(s) Covered:
  • Network Design

The foodservice distribution industry in the United States is projected to experience a 40% growth from 2022 to 2025. As the industry expands, an increasing number of companies are adopting an omnichannel supply chain strategy. In this context, our sponsor company, a large U.S. foodservice distribution company, aims to optimize the end-to-end supply chain network, including omnichannel options, in order to minimize the overall supply chain cost. This objective is typically achieved through a technique known as supply chain network design, which has been extensively researched. However, while there is abundant literature discussing omnichannel supply chain strategies and network design, only 4% of the reviewed supply chain research specifically addresses B2B applications. Given the substantial differences between B2C and B2B businesses (e.g., lead-time requirements, network and cost structure, number of facilities and customers), a new customized model is required. Additionally, many previous studies have primarily focused on cost factors directly proportional to the volume of flows within the network, such as product cost, transportation cost, and warehouse handling cost. However, to achieve the most optimal solution, we must consider the impact of inventory positioning. Changes in inventory positioning can significantly influence the ability of the model to leverage inventory pooling, ultimately affecting safety stock costs. Since this pooling effect generally exhibits nonlinear behavior, many previous studies have either overlooked it or proposed separate models for network optimization and inventory positioning. To address these issues, we propose a mixed-integer non-linear programming (MINLP) model that simultaneously optimizes supply chain network flows and inventory positioning. We have solved this model by reformulating the square rooted term representing safety stock cost as a quadratic constraint, which can be solved by commercial solvers. Additionally, we have developed a tailored algorithm using outer approximation (OA) to expedite the solving process for our inherently complex model, known as NP-Hard. Depending on the specific products, our results demonstrate potential cost reductions of 3-9% in transportation, 2-8% in warehouse handling, and up to 50% in inventory costs. Furthermore, we have achieved more than nine times greater efficiency with our tailored algorithm compared to Gurobi's default solve. Lastly, our model exhibits the possibility of future expansion into a multi-item model, which would have a significantly greater impact on the company.