A New Mixed-Integer Linear Programming Model for Omnichannel Inventory Optimization with an Empirical Application

Özet

The rapid acceleration of digitalization, the widespread adoption of mobile technology, and shifting consumer purchasing patterns have elevated the strategic importance of omnichannel frameworks within retail supply chain networks. The convergence of physical stores, online platforms, mobile applications, and marketplaces has resulted in heightened customer expectations for smooth interactions across all touchpoints. Inventory allocation, order fulfillment, inter-channel transfers, and delivery decisions are then expressed as a multidimensional planning problem. In the cosmetics industry, effective decision-making relies on a carefully planned inventory placement strategy that incorporates multiple distribution channels, directing customer demand to the most suitable channel and controlling goods movement quantities between channels to achieve efficient operations and the highest possible profitability. This study proposes a novel mixed-integer linear programming (MILP) model designed to maximize the total profit within an omnichannel distribution network. The model combines sales revenue, inventory costs, transportation and transfer expenses, and initial procurement costs to guide product distribution across channels. By integrating these elements, it provides a unified framework for determining shipments between channels and setting initial stock levels. The model was evaluated on a genuine cosmetics dataset via FICO Xpress, with its solution quality, computational efficiency, and scenario-based sensitivity assessed across diverse product, channel, and time configurations. The study’s findings indicate that the proposed model not only bridges the existing gap in multi-channel decision-making research but also provides a practical decision-support tool for managing multi-channel logistics networks. Furthermore, the model has strong potential to be enhanced through heuristic or metaheuristic extensions, enabling even higher performance for large-scale problem instances.