Optimization of item availability prompts in the context of non-deterministic inventory data

Abstract

A system receives a request for a set of items at a warehouse from a user device, and determines a set of candidate items responsive to the request. The system applies a trained item availability model to each candidate item to determine a prediction of a likelihood that the candidate item is available for pickup at the warehouse. A subset of candidate items that have a prediction below a threshold is classified as low availability. The computer system also determines a cap of low availability items to present to a user based on a user utility curve. The user utility curve is modeled based on user utility associated with amounts of low availability items presented. The low availability items are filtered to an amount within the determined cap. The filtered low availability items are sent to the user device for presentation in a user interface.

CPC Classifications