Integrated Warehouse Location and Inventory Decisions in a Multi-location Newsvendor Problem

TL;DR

This paper analyzes a multi-location newsvendor supply chain with centralized and decentralized inventory decisions, considering complex transportation costs, and proposes solutions demonstrating the benefits of centralization in large networks.

Contribution

It introduces a comprehensive modeling framework for warehouse location and inventory decisions with complex transportation costs, providing closed-form solutions and a novel Q-search algorithm.

Findings

Centralized models outperform decentralized ones in profit and service level for large networks.

Closed-form solutions are derived for certain transportation cost dependencies.

The Q-search algorithm effectively finds high-quality solutions for complex models.

Abstract

In this paper, we investigate a supply chain network with a supplier and multiple retailers. The supplier can either take orders from retailers directly, or choose to build a warehouse somewhere in the network to centralize the ordering from retailers. Meanwhile, we take three modes of transportation cost into account, including distance-dependent, quantity-dependent, and distance-quantity-dependent costs, to formulate six models. For the three decentralized models, we provide closed-form solutions to compute the optimal order quantity of each retailer. For the centralized models, we develop closed-form solutions for the first two models as the transportation cost only depends on either distance or order quantity; but when it depends on both, the model becomes a non-linear programming problem. We develop a solution algorithm named Q-search to find a high-quality solution that includes the order quantity and the warehouse location. The experiment results show that the centralized model outperforms the decentralized model in large networks in terms of profit and service level.