STL stands for Seasonal-Trend decomposition using Loess. It is a technique introduced by Robert Cleveland, William Cleveland, Jean McRae, and Irma Terpenning in a 1990 paper in the Journal of Official Statistics. The method takes a single time series and separates it into three additive parts: a slowly changing trend, a repeating seasonal pattern, and a remainder that captures everything left over, including noise and unusual events.
What makes STL useful is its flexibility. It uses loess, a form of local smoothing, so the trend and seasonal shapes can bend and evolve over time rather than being forced into a fixed form. It can handle any kind of seasonality, lets the analyst control how quickly the seasonal pattern is allowed to change, and has a robust mode that prevents a few outliers from distorting the whole decomposition. The result is a clear, visual breakdown that makes it easy to see the underlying direction of a series with the seasonal wiggles stripped away.
Decomposition like this is a standard first step in time-series analysis. It feeds directly into forecasting and is built into modern tools, including the structure used by Facebook’s Prophet.
Why business readers should care: STL turns a noisy, seasonal sales or traffic chart into a clean trend line, making it far easier to tell genuine growth from a holiday bump.