Prévision et gestion de stocks basés sur les données de panier dans la supply chain omnicanal

Omni-channel retailing has become the strategy for a competitive edge for most of the retailers relying on a network of brick-and-mortar stores. The COVID-19 pandemic accelerated the trend with new consumer preferences along with a technology-driven and data-driven redesign of the retail industry. In an omnichannel environment, the customer has several options to inspect, to order, and to receive an ordered product via a delivery or a pick up of the product. With this ultimate customer journey, any omnichannel retailer is particularly challenged in terms of inventory and replenishment decisions, transportation planning, and forecasting, which increase tenfold when sales are mainly in baskets and not only individual items. Accordingly, the thesis builds on a basket data-driven approach to answer some of these challenges when it comes to forecasting and inventory planning. First, we propose a novel omnichannel forecasting approach using basket data that provides an improvement in forecasting accuracy and inventory performance. Second, an anticipatory shipping strategy is designed and tested that builds on basket data prediction, which provides an improvement in the delivery lead time to customers and reduce costs. Third, we develop and compare the performance of several hierarchical forecasting approaches considering basket data. All these contributions, build methodologically on advanced forecasting and machine learning techniques, graph theory, and inventory simulation. In addition, they involve data from a large European retailer in the cosmetics industry and thus provide valuable insights for practice.