To provide their patients with the care they need as quickly as possible, pharmacies are supplied by wholesaler-distributors who provide them with a half-day delivery guarantee for most product references. For this purpose, they have set up an efficient and complex supply chain. To further improve the efficiency of their delivery services, some of them want to use machine learning tools to predict future orders and anticipate their inventory needs. This paper investigates different machine learning models for the prediction of sales on molecules of a French wholesaler-distributor. This paper focuses on four molecules and compares the results of the models predictions on these molecules