An incentive-based delivery scheme and its effect evaluated via explainable machine learning

Abstract

The current delivery mode acquiesces E-shopping consumers to provide only a single address for delivery, despite their potential to have multiple addresses available. Inspired by this, we propose a new delivery mode with an economic incentive scheme to encourage consumers to provide more addresses and empower the delivery operator to determine the final delivery address following a certain optimization criteria. To examine the incentive's effect, we conducted a survey. The survey reveals a substantial, near-linear impact on promoting multiple address provision through the incentive, resulting in a 32% increase in consumers providing additional addresses. We develop an eXtreme Gradient Boosting model, which outperformed Logistic Regression and Support Vector Machine, to explore the relationship between address provision decision and E-shopping behavior. Augmented by Shapley Additive Explanations, the model can interpret how both the incentive and E-shopping behavior influence address provision. In addition to the incentive, factors such as the number of available addresses and the average price of the parcel also significantly influence the decision-making process for providing delivery addresses. The insights extracted from this study can provide a foundation for policymakers to establish more practical delivery management policies.