Causal inference improving warehouse productivity: zoned storage and killer items

Abstract

E-commerce companies need help with customer service experience: faster and more frequent deliveries. Then, order fulfillment becomes critical to establish a competitive advantage. The main objective of this project is to determine whether a new key product assortment, called a class-based scattered storage policy, improves order-picking operations in one of the main warehouses of the sponsor company. This e-commerce firm operates in an emerging market. As mentioned earlier, this project addresses the problem by running an A/B quasi-experiment in the warehouse, showing findings directly from a real context for the first time. For this purpose, the warehouse was split into control and treatment sections during peak season when speed is most required. The effect of the proposed storage policy is studied by comparing picking productivity through a two-sample t-test. The samples are chosen using the Coarsened Exact Matching algorithm to have similar data to analyze in observable characteristics. The result of this work indicates that the class-based scattered storage policy does not lead to an improvement in picking productivity. It can be attributed to real-context features that are presented and discussed. Additionally, strong recommendations are given to include the findings in future research.