Predicting Human Discretion to Adjust Algorithmic Prescription: A Large-Scale Field Experiment in Warehouse Operations

Conventional optimization algorithms that prescribe order packing instructions (which items to pack in which sequence in which box or bin) focus on bin volume utilization (efficiency) yet tend to overlook human behavioral deviations. We observed the order fulfillment operations at Cainiao, the logistics division of Alibaba which is the largest e-commerce platform in China. We found that workers deviated from the algorithmic prescriptions for more than 5.8% of packages, typically by switching to a larger bin than recommended. This switch increases packing time as well as material and environmental costs. We identify such behavioral deviations and propose a new algorithm that predicts discretionary behavior (e.g., switching to a larger bin) using machine learning techniques to pro-actively adjust the algorithmic prescription. We conducted a large-scale randomized field experiment with the Alibaba Group involving 757 workers and 782,360 packages from August 27, 2018 to September 9, 2018. During the experiment period, we randomly assign orders from Alibaba to either receive our ?human-centric bin packing algorithm? (treatment group) or Alibaba's original algorithm without behavioral adjustment (control group). Our field experiment results show that, by anticipating and incorporating human behavior, our new algorithm reduces the deviation probability of workers from 30.1% to 24.5% and improves their average packing time of targeted packages (i.e., packages for which workers are more likely to deviate) by 4.5%. This idea of incorporating human deviation to improve optimization algorithms could also be generalized to other processes in logistics and operations.