Track 12 · Time Series and Forecasting · Lesson 4 of 11
Simple exponential smoothing
Update and forecast with simple exponential smoothing, interpret the smoothing parameter α, and choose α honestly on held-out data.
35 minAdvanced7 graded exercisesMembers
A store manager wants next week's sales forecast every Sunday. The last week alone is too noisy; the average of two years ignores that things change. Simple exponential smoothing sits between the two: an average that gives recent weeks more weight, updated with one line of arithmetic.
Two simple forecasts bracket the idea:
- The naive forecast uses only the last value: . It reacts instantly, including to noise.
- The mean forecast uses the average of all past values. It is stable, but treats a value from two years ago as just as relevant as last week.
Simple exponential smoothing (SES) keeps a running level and nudges it towards each new observation by a fraction .
Included with a Formula Quest membership
It has 7 graded exercises with feedback, hints and worked solutions. After it you will be able to:
- Update the SES level by hand and produce its flat forecast
- Explain the exponentially decaying weights α(1 − α)^k
- Interpret large and small α as short and long memory
- Choose α by holdout error and compare with a simple benchmark
- Recognise when SES is the wrong method (trend, seasonality)