Nicolas Vandeput on machine learning, inventory simulation and planner roles

A long-form discussion with Nicolas Vandeput covers Python adoption, master-data limitations, outlier treatment, forecast explainability, the changing role of planners, inventory simulation, safety-stock assumptions and the practical limits of external signals.

The episode supports a governance model in which machine learning generates a controlled baseline while planners focus on data quality, exceptional information and decision validation. Lightweight models still require IT ownership, monitoring and defined interfaces with planning systems.