Time Series Decomposer
Overview
The Time Series Decomposer separates time series data into three fundamental components: trend, seasonal, and residual. This classical decomposition technique helps you understand underlying patterns in your data by isolating long-term growth (trend), repeating cycles (seasonal), and random fluctuations (residual). The tool supports both additive and multiplicative decomposition methods, making it suitable for various types of time series data from sales forecasting to web analytics.
Tips
- Use additive decomposition when seasonal variations stay constant over time, multiplicative when they grow proportionally with the trend
- Set the seasonal period to match your data’s cycle length (7 for daily data with weekly patterns, 12 for monthly data with yearly patterns)
- Check the residual component - it should look like random noise with no patterns; visible patterns suggest your model is missing something
- Adjust the trend window size to control smoothness - larger windows create smoother trends but may miss shorter-term changes
- Examine residual statistics (mean near zero, low skewness) to validate the quality of your decomposition
- Use the deseasonalized data (original minus seasonal component) for clearer trend analysis and reporting
- Upload at least 2 complete seasonal cycles for reliable seasonal pattern estimation