Modèles de réapprovisionnement de stocks sous incertitudes et perturbations dans le contexte d'e-détaillant
The aim of this PhD research is to study the impact of the inventory inaccuracy issue on replenishment policies under the wholesaling / e-retailing context. The inventory inaccuracy is defined as a discrepancy between the quantity shown by the Informational System and the quantity actually Physically available for sales. Such discrepancy has as root many sources of errors such as execution, transaction, misplacement, supply unreliability or shrinkage errors. We provide empirical evidence on the wide presence of inventory inaccuracies through one case study of wholesaling organizations. In e-retailing context customers’ demands are remotely satisfied based on the inventory level shown in the information system. In fact, when the information system is subject to inaccuracies, satisfying a demand based on inaccurate information could have therefore the non-satisfaction of a commitment when shipping the products, particularly when the physical stock is not aligned with the information system. The main interest of the research community was about studying the inaccuracy under the retail context. After motivating our research empirically and by the literature review, we develop three inventory frameworks subject to inventory inaccuracies under the wholesaling / e-retailing context with different configurations of the stochastic errors describing the inaccuracy issue. The error configuration could be additive for some sources of inaccuracies such as transaction as well as multiplicative (also known as stochastically proportional) for other sources such as shrinkage. Our first inventory framework aims to extend single-period (Newsvendor) model to the e-retailing/wholesaling context where the study of the inaccuracy issue is intuitively and mathematically more challenging. In Addition, to providing managerial insights, we compare the behavior of the optimal ordering strategies under the additive and the multiplicative error settings. In fact, being aware that the assumption, about the error setting (additive or multiplicative likely modeled) is sometimes not easy to validate, we proof that the inventory manager has to prefer the multiplicative error configuration in case of doubt about the error modeling. Our second inventory framework is an extension of the former to the case of two selling periods. By solving exactly the two-period problem, we show that the behavior of the optimal ordering quantities is not monotonic, and conclude that a myopic policy could not be optimal. Thanks to a numerical study, we derive some interesting managerial insights about the management of the error risk between the two selling periods. Thanks to the theoretical results of our second inventory framework, we propose a third framework dealing with the inventory inaccuracy issue under a multi-period setting by assuming cost minimizing as a target but also by considering two Cycle Service Level constraints to achieve, the former for the demand satisfaction and the latter for the sales commitment satisfaction. We provide a comprehensive numerical study by comparing several policies and inspection strategies permitting to align the inventory levels.
KEYWORDS: inventory management; disruptions; multiplicative / additive error modeling; errors; RFID; e-retailer