Seasonal Decomposition Tool Calculator

Upload values, choose season length, and run decomposition. View trend, seasonal indices, and residual diagnostics. Download reports fast and validate patterns with confidence easily.

Calculator Inputs

Paste your time series. Optionally include labels or dates in the first column.
Use multiplicative for proportional seasonality.
Examples: 12 (monthly), 7 (daily weekly cycle), 4 (quarterly).
Interpolation helps moving averages when data is sparse.
Additive: indices sum to zero. Multiplicative: average equals one.
Use fewer decimals for reporting, more for auditing.
Charts appear in the results section.
Accepted formats: one value per line, or label,value. Use NA for missing points.

Example Data Table

This sample shows an upward trend with a yearly cycle.

Label Value
2024-01120
2024-02128
2024-03135
2024-04142
2024-05150
2024-06160
2024-07172
2024-08168
2024-09155
2024-10145
2024-11138
2024-12148
2025-01125
2025-02132
2025-03140
2025-04148
2025-05156
2025-06166
2025-07180
2025-08175
2025-09162
2025-10152
2025-11145
2025-12155

Formula Used

This tool applies classical seasonal decomposition using a centered moving average trend and seasonal indices by position.

Additive model
Observedt = Trendt + Seasonalt + Residualt
Seasonal indices are normalized so they sum to zero across one period.
Multiplicative model
Observedt = Trendt × Seasonalt × Residualt
Seasonal indices are normalized so their average equals one.
Trend estimation
Trendt = centered moving average over the chosen period
For even periods, the tool uses a two-step centering commonly used in time-series textbooks.
Seasonal index for a position p
Sp = average of detrended values at position p
Detrended is Observed − Trend (additive) or Observed ÷ Trend (multiplicative).

How to Use This Calculator

  1. Paste your observations in the data box, one row per time step.
  2. If you have dates or labels, use the label,value format.
  3. Set the season length (period), such as 12 for monthly data.
  4. Pick additive for constant seasonal swings, multiplicative for proportional swings.
  5. Click Decompose Series to see results under the header.
  6. Use the download buttons to export your tables for reports.

Data structure and input quality

Seasonal decomposition assumes observations arrive at regular intervals, such as sales, weekly defects, or monthly energy usage. For stability, provide at least two full seasons; with a 12-step period, 24 to 36 points improves index averaging. Keep units consistent, remove duplicated timestamps, and review one-off spikes as special events. Missing points can be written as NA and handled by skipping or interpolating internal gaps.

Season length and moving-average coverage

The period controls seasonal grouping and the trend smoother. Trend is estimated with a centered moving average of length m. When m is even, a second centering step aligns the trend to integer positions, so edge values can be unavailable. Use n ≥ 2m for a workable output in practice, and n ≥ 3m for steadier indices. Example: n=36 with m=12 yields a trend for most months and clearer seasonality.

Model choice: additive versus multiplicative

Choose additive when seasonal swings have constant magnitude, for example ±15 units regardless of baseline. Choose multiplicative when swings scale with the level, such as +8% in peak months and −6% in off months. Multiplicative decomposition requires positive observations because it uses ratios. Indices are normalized so additive indices sum to zero per period, while multiplicative indices average to one, easing cross-series comparison for clearer interpretation.

Diagnostics that support reliable reporting

Residual diagnostics summarize what remains after removing trend and seasonality. Additive residuals should have a mean near 0 and a stable spread; multiplicative residuals should cluster around 1. Seasonal strength approaches 1 when residual variance is small relative to detrended variance. Fit metrics quantify accuracy on fitted points: MAE is typical absolute error, RMSE penalizes large misses, and MAPE expresses percentage error when observed values are nonzero.

Using components for forecasting workflows

Use the deseasonalized series to model underlying movement with regression, smoothing, or autoregressive methods, then reapply seasonal indices to reconstruct forecasts. Compare fitted and observed values to spot regime changes, and monitor residuals for unusual cycles or breaks. If residual mean drifts or error metrics rise, revisit the period and model, or add external drivers. Export the component table to share key assumptions and reproducible calculations with stakeholders.

FAQs

How do I choose the season length (period)?

Match the period to the repeating cycle in your sampling. Monthly data often uses 12, daily data with weekly seasonality uses 7, and quarterly series uses 4. If unsure, test nearby values and compare error metrics.

Why is the trend sometimes shown as a dash at the edges?

Centered moving averages need points on both sides of a time step. Near the beginning and end, the full window is unavailable, so the trend is intentionally left blank to avoid biased estimates.

Should I skip missing values or interpolate them?

Skip is safer when gaps are large or irregular. Interpolation is helpful for small internal gaps because it restores continuity for the moving average. Avoid interpolation when missingness is systematic or when values jump abruptly.

When is a multiplicative model appropriate?

Use multiplicative when seasonal effects scale with the level, such as percentage swings. All observations must be positive because the method uses ratios. If your data includes zeros or negatives, choose additive.

What does seasonal strength indicate?

Seasonal strength summarizes how much variation remains seasonal after accounting for trend. Values closer to 1 suggest a strong, consistent seasonal pattern, while values near 0 suggest weak seasonality or high residual noise.

What do the CSV and PDF downloads include?

Exports use the component table displayed in results: label, observed, trend, seasonal, fitted, residual, and deseasonalized values. The PDF is a formatted landscape report, and the CSV is a raw file for analysis tools.

Related Calculators

Moving Average CalculatorTime Series Forecast ToolPartial Autocorrelation ToolStationarity Test ToolADF Test CalculatorKPSS Test CalculatorHolt Winters ToolSeasonal Index CalculatorDifferencing CalculatorAnomaly Detection Tool

Important Note: All the Calculators listed in this site are for educational purpose only and we do not guarentee the accuracy of results. Please do consult with other sources as well.