Trends, seasonality and anomalies in time series

The Time series step runs the site’s Time Series Explorer on a dataset with a date column. Open the example Website traffic: trend and anomalies to follow along: it contains ten weeks of daily visits (the tool’s own example data).

Set up the step

Open the Workflow desk and select Weekday pattern and anomalies in its list. The settings:

• Date column: date (empty = the first date column).
• Value columns: visits. You can tick up to eight series.
• Resample to: Automatic keeps the data’s own spacing; choose Week or Month to aggregate, with Combine values per period (average, sum, last …).
• Smoothing and Smoothing window: a 7-period moving average suits daily data with a weekly rhythm.
• Seasonality: Day of the week for daily data, Month of the year for monthly data.
• Anomaly detection: z-score of the residuals with Anomaly threshold 3 flags values more than three standard deviations away from trend + season.

Press Run (or Ctrl+Enter).

Read the findings

For the example data the findings say:

• visits: +26.02 % from the first to the last value (2026-01-05 to 2026-03-15).
• visits: the linear trend (160.84 per year) explains little of the variation (R² 0.463) — the trend exists, but day-to-day swings are larger.
• visits: 2 anomalies, the first on 2026-01-28.
• visits: a noticeable weekday pattern (strength 0.471) — weekends are clearly lower.

Below come the tool’s charts (data with smoothing and anomalies, the seasonal profile, year-over-year when there is enough history) and their tables.

Follow the data over time

When you replace the dataset with a newer export and run the step again, RavingDesk keeps the previous result. Changes since the previous run lists every number that moved — the change in %, the trend per year, the number of anomalies and the last value — and the report shows the same list.

Monthly data and year-over-year

The example Seasonality and year-over-year has three years of monthly units for two products. Its time-series step uses Seasonality: Month of the year and a 3-month window. Product B shows a clear linear trend of 46.54 per year (R² 0.997): almost all of its change is steady growth, while Product A is dominated by the season.

Tips

• Long tables (one row per date and series) work too: set Series column (long format).
• Anomalies are statistical outliers, not errors. Check them against what happened (a campaign, an outage) and write it into Your note for the report.