A0922
Title: Challenges in achieving explainability and control with supply chain forecasts
Authors: Rishab Guha - Amazon (United States) [presenting]
Andrea Tambalotti - Amazon (United States)
Phillip Jang - Amazon (United States)
Abstract: AI and deep learning methods have revolutionized many forecasting applications but have not achieved widespread adoption in industry for "macro" forecasting (e.g., forecasting aggregate revenue). This paper identifies three critical capabilities that traditional macroeconometrics methods achieve but current AI approaches lack: (1) multivariate consistency at scale, (2) explainable and controllable long-run assumptions, and (3) flexible incorporation of forward-looking external inputs. A Bayesian vector autoregression state-space framework is described, which builds on models used in macroeconometric forecasting, and is used in production at a major e-commerce retailer, where the forecasts influence billions of dollars in spending decisions. By detailing how traditional time series methods solve these challenges today, concrete opportunities for researchers are identified to develop hybrid approaches that combine the accuracy advantages of modern AI with the explainability and control benefits of traditional methods.
