Identifying the Root Causes of Stockout Events in e-commerce Using Machine Learning Techniques

Abstract

The year 2020 marked an unprecedented worldwide growth in e-commerce driven mainly by the COVID- 19 pandemic. The lockdown restrictions created significant spikes in the demand for several products causing severe disruptions throughout the supply chains. The pandemic created significant challenges for companies to maintain efficiency in the supply chain and product availability on the digital shelves. Stockout events rose considerably in the online platforms, and companies across industries needed to find ways to address the problem. The focus of this project was to identify the main reasons that lead to stockouts for the sponsoring company to a major online retailer and to develop a model to predict the stockouts. Using supervised machine learning models, we developed a model that predicts the missing order quantities for every specific order. The analysis shows that the variables associated to the demand such as order quantity have a higher impact than variables associated with the supply, such as inventory on hand. Additionally, the product categories and brands associated with each category play an important role in the stockout prediction. With the continued growth of e-commerce and customers changing their shopping preferences, our predictive model will help the sponsoring company analyze the orders and make informative decisions to predict the stockouts and improve the inventory allocation.