OpenForecast launched its open course, Demand Forecasting Principles, running in November with eight live sessions over four weeks, starting on 3 November, delivered by Nikos Kourentzes and Ivan Svetunkov.

We help companies get their stock levels right — distributors, manufacturers, spare parts operations, retailers. We start by understanding how your process actually runs and measuring how accurate the forecasts really are. What the measurement finds is usually not what people expect: - Demand hidden by stockouts. Sales history records a zero when you had nothing to sell, and the model learns that nobody wanted it. - Error measures that reward the wrong thing. A forecast can score well on accuracy and still be the wrong input for the decision it feeds. - Safety stock set by convention. Service targets chosen by habit rather than calculation, costing money at both ends. - Models that cannot represent your demand. A promotion, a price change, or a large amount of slow movers break tools built for steady weekly sales, and no amount of tuning fixes a model that was developed for a different situation. - A process built around the software, not the decision. Forecasts produced with arbitrary periodicity, not aligning with the ordering decisions, at a level of aggregation nobody uses, for a horizon that does not match the lead time. - Overrides nobody tracks. When the system is not trusted, planners adjust by hand — and most of those adjustments make the forecast worse. If any of that sounds familiar, a conversation is the cheapest way to find out how much it is costing you.