Time Series Decomposition: Split Trend, Seasonal and Residual Components
What is Time Series Decomposition?
Time series decomposition separates a series into three components: trend (long-term direction), seasonal (repeating pattern with a fixed period, e.g. weekly, monthly or yearly) and residual (irregular remainder). It is the first step of most forecasting workflows: once the components are separated you can model the trend and seasonality separately and detect anomalies in the residual. The tool supports both additive decomposition (components sum to the series) and multiplicative decomposition (components multiply to the series), and lets you choose the seasonal period (for example 12 for monthly data with yearly seasonality, 7 for daily data with weekly seasonality).
Additive or Multiplicative?
Choose additive when the seasonal amplitude stays roughly constant over time, and multiplicative when the amplitude grows with the trend level (common in sales, traffic and defect counts). As a quick check, plot the series: if the ups and downs look like a fixed-width band, use additive; if the band widens as the series rises, use multiplicative. After decomposition, look at the residual component: a stable residual centred near zero with no obvious pattern indicates the model captures the structure well; a residual that still shows a wave suggests the wrong period or model.
How to Read the Results
The tool outputs four panels: observed, trend, seasonal and residual. The seasonal component repeats exactly every period and is centred near zero (additive) or one (multiplicative); the trend line shows the underlying direction after removing seasonality. Use the seasonal strength to decide whether seasonality is worth modelling: if the seasonal swing is large relative to the residual, include it in the forecast. Remaining patterns in the residual - a trend or a wave - mean the decomposition can be improved by adjusting the period or switching between additive and multiplicative.
How to Use It
After sign-in, paste your time series (one value per row, with optional time labels), choose the seasonal period and model type, then review the four panels. You can export the chart and the component values for further analysis in forecasting tools. The decomposition result also feeds directly into Holt-Winters and ARIMA tools for the next forecasting step.