When One Forecast Is Not Enough: Machine Learning for Smarter Demand Planning  

What if the biggest forecasting mistake is assuming that one model can predict everything? This talk explores a real-world demand forecasting and inventory optimization framework where statistical methods, machine learning, deep learning, game theory, and chaos-inspired approaches compete across thousands of product time series. Going beyond forecast accuracy, the session shows how rolling evaluation, multi-metric model selection, and inventory optimization can turn predictions into better business decisions – reducing stockouts and excess inventory while improving product availability and inventory efficiency.