Quantitative Models for Managing Multi-Node Store Replenishment Logistics in the Fast-Food Supply Chain

Abstract

Fast-food restaurants represent a significant sector of the retail industry, with consumers spending over $310 billion in a single year, that has shown consistent yearly average growth of 6% over the last 10 years. However, there is a surprising scarcity of studies focused on the replenishment logistics of this sector. Considering how replenishment logistics are critical to any fast-food company success and how costly the logistical challenges in this area can be, it makes this a particularly interesting area of study. The work discussed in this dissertation addresses this gap by developing quantitative models and decision tools focused on the replenishment logistics in the fast-food industry.

We introduce the store replenishment problem (SRP), which is concerned with minimizing the logistics costs associated with replenishing stores in a network over a fixed time horizon. The objective function of the SRP was defined in collaboration with a supply chain group at a well-known fast-food chain to include four critical cost components associated with the replenishment logistics: transportation, labor, fleet size, and route-time overage costs. We formulate the SRP as a mixed-integer program using a set-partitioning approach. The formulation uses a set of pre-generated potential routes as input and then concurrently determines fleet size, delivery routes, and chooses which routes are going to be completed by a single-driver or a team of two. Using real-world data, we show that the proposed heuristic outperforms the current industry baseline and that the multi-component objective function obtains superior solutions to those obtained with one-dimensional objectives, such as minimizing only the travel distance or the fleet size.

The second model presented focuses on a generalization of the SRP, where building a flexible delivery schedule is included as a decision variable (SRP-FS). This flexible schedule determines both the timing and the quantity of the deliveries at each store while observing the storage capacity of the stores. The objective of the SRP-FS is also to minimize replenishment costs and uses the same multi-component cost structure as the SRP. We developed a two-step simulated annealing metaheuristic, that incorporates an adjusted Savings Algorithm to solve the vehicle routing component. A series of self-generated test problems and real-world data from our industry collaborator are used to evaluate the performance of the heuristic. The results show that the proposed metaheuristic is capable of finding good solutions in reasonable times and that significant cost reductions can be obtained by introducing a flexible delivery schedule.

The last model discussed in this dissertation is concerned with the impact of the store network composition on the SRP. Due to the nature of the fast-food industry, store locations are not uniformly spread through regions and can result in areas with high density of stores areas with relatively isolated stores in remote locations. This mixed composition in the store network can present a logistical challenge for decision-makers when planning store replenishment routes. We propose a quantitative model that exploits the clustered nature of the store network into the solution approach. Using a clustering heuristic, we are able to simplify the decision space of the problem and formulate the SRP as a bin-packing problem to assign clusters to routes. Our computational results show that the proposed heuristic outperforms the original SRP method in almost every test instance, particularly in instances based on real-world data from our industry collaborator.