Demand Forecasting with Machine Learning
From time series to reliable forecasts: methods, metrics and practice
Audience
Demand planners, S&OP/IBP owners, controllers with planning responsibilities, data analysts in supply chain functions. Basic numeracy is sufficient; programming skills are helpful but not required.
Starting point
Many companies forecast with Excel trends or the standard methods of their ERP system – and wonder about systematic plan deviations. Modern ML methods can be deployed today without a data science team and deliver measurably better results.
Contents
- Overview: which forecasting methods exist – from exponential smoothing to gradient boosting (XGBoost/LightGBM) – and when is which appropriate?
- Why point forecasts are not enough: quantifying and using forecast uncertainty
- Feature engineering made clear: seasonality, calendar effects, price promotions, external drivers
- Measuring forecast quality: MAPE, RMSE, bias – and which metric misleads you when
- Forecast accuracy vs. business value: when is a better forecast actually worth it?
- Hands-on: a complete forecast pipeline on real sample data – prepare data, train the model, evaluate results, export forecasts
- Anchoring forecasting in the S&OP/IBP process: from technical forecast to accepted planning baseline
Your takeaway
You will be able to evaluate ML forecasting methods on solid ground, set up a pilot environment yourself, and talk to vendors or your own IT as an equal.
Your trainer
Dr. Jan Fränkle
20+ years of machine learning and AI, founder of ITM-predictive, lecturer and consultant. Teaches complex methods in a way that lets you apply them the very next day.
Interested in this seminar?
Dates, pricing and in-house variants are best discussed directly – book a short call or send us an email.
All seminarsRequest a training
Briefly describe what you have in mind – we will get back to you within one business day with a proposal.