Smoothing and moving average
Smoothing and moving average: simple forecast methods for stable series without strong seasonality. A moving average takes the mean of the latest periods, exponential smoothing weights fresh data more than old. They suit goods with even demand.
Formula
Moving average = mean of the last N periods; level = α × latest sales + (1 − α) × previous level
How to read
Suppose 90, 100 and 110 units were sold in three weeks: the moving average is 100 units. They do not see seasonality: on seasonal data they lose to a method with a seasonal index. Choose the window by series stability, not at random.
Where it is discussed
Topics
Calculations
- Intermittent demand: Croston and SBA for rare sizes and accessories
- Sales forecast: method, horizon, error
- Holt-Winters: a forecast with trend and seasonality, error on held-out periods