The problem we study is inspired by an industrial collaboration between a third-party logistics, DHL Supply Chain, and a large French restaurant chain. As part of this partnership, DHL Supply Chain coordinates the actors of a domestic logistics network composed of suppliers, warehouses and restaurants. Over a certain time horizon, the restaurants issue requests of generic products (frozen products, beverages, etc.) that are manufactured by the suppliers. The mission of DHL Supply Chain is to ensure the supply of the restaurants. For that purpose, the company determines the shipping origin of each product ordered and designs a loading plan that characterizes the routes followed by the goods. DHL Supply Chain wants to develop innovative solutions to improve its competitiveness and optimize the profitability of its logistics operations. In this thesis, we present the Logistics Service Network Design Problem (LSNDP) which form! alizes the problematic of planning transportation operations in a supply chain. Our work aims to provide methodological solutions for solving industrial instances of the LSNDP. However, these industrial instances are too complex to be solved by generic operations research methods. We thus propose several algorithms that overcome the scaling of the parameters. In particular, we develop a graph reduction heuristic, as well as a dynamic Benders strategy that adapts to the increasing number of products. Through various computational studies, we evaluate the scalability of each algorithm with respect to the considered parameter. Finally, we combine these methods for the resolution of a real case.