Human-machine interaction in retail demand forecasting

Speaker: Prof. Dr. Kai Hoberg

Abstract: While AI is generally very good at generating demand forecasts in retail settings, planners in retail companies can typically override these forecasts. We aim to understand when humans feel that they can add value to the AI generated forecast and when they do add value. In our research we analyze a real-world data set that contains 30 million forecasts at the SKU-store-day level plus additional variables, e.g. related to products, weather, or holidays. Our results show under which conditions planners’ interventions should be enabled. We propose managerial implications for the best use of human knowledge to improve the process.