Abstract: In the age of Machine Learning, many data scientists have started using more complicated approaches for forecasting, such as GAM, XGBoost, k-NN, Artificial Neural Networks etc. Due to the increased interest in these methods, the simpler and robust approaches might look less attractive than the more innovative ones and as a result tend to be neglected. However, the good old exponential smoothing still works well in many situations and is still widely used in practice, especially in demand planning. In this talk, Ivan will give a short overview of the history of exponential smoothing, show how it has evolved over time, explain why it has been so popular and discuss how it can be used in the modern demand forecasting in a variety of situations.

This will be a modified version of my talk in the CMAF FFT webinar in December.