View Monitor Results¶
Use the Monitors dashboard to track anomalies detected across your monitored tables. The dashboard shows anomaly counts by type (Freshness, Volume, Schema, Metric) and lets you drill into trend charts to investigate issues.
Identify tables with anomalies¶
Select a Table Group from the dropdown to view its monitored tables. The anomaly summary at the top shows total anomaly counts broken down by monitor type.
- Use the Anomaly type filter to narrow the list to tables with specific anomaly types (Freshness, Volume, Schema, or Metrics).
- Check Show changes to see latest update times, row count changes, and schema change details alongside the anomaly counts.
- Tables with anomalies are highlighted for quick identification.
Note
Monitors in Training mode (not yet enough data for predictions) or Pending status (configured but not yet run) are indicated in the anomaly summary. No anomaly detection occurs during training.
Investigate an anomaly¶
Click any anomaly count or the :material-insights: (view trends icon) on a table to open the trend charts dialog. The dialog shows time series charts for each monitor type, making it easy to spot patterns and determine whether an anomaly is a one-time spike or part of a trend.
When investigating anomalies:
- Review the trend chart — Look at the historical context. Anomalies appear as red markers; the gray prediction band shows the expected range.
- Check for correlated anomalies — Did multiple monitor types flag issues at the same time? Schema changes can often explain Volume or Freshness anomalies (e.g., a dropped column causing a pipeline failure).
- Cross-reference with operations — Compare the anomaly timing with known events: deployments, data migrations, pipeline changes, or maintenance windows.
- Adjust if needed — If the anomaly was expected, consider adjusting the monitor's sensitivity or threshold mode. See Configure Monitors.
Interpret results by monitor type¶
For an overview of what each monitor type does and how the prediction system works, see Monitor Tables.
Freshness anomalies¶
- Late — The table has not been updated, and the elapsed time exceeds the expected staleness threshold. Usually indicates a pipeline that stopped running or a delayed upstream process.
- Earlier than expected — The table was updated, but sooner than the learned pattern predicts. May indicate an ad-hoc manual update or a schedule change in an upstream pipeline.
- Later than expected — The table was updated, but later than the learned pattern predicts. May indicate a delayed upstream process or a gradual drift in pipeline timing.
Volume anomalies¶
The row count is outside the predicted or configured range. Common causes:
- A bulk data load or delete operation.
- A pipeline failure that prevented expected data from arriving.
- A data retention policy that removed more rows than expected.
Schema anomalies¶
Any detected schema change (column additions, deletions, or modifications) is always flagged as an anomaly. Click the schema indicators in the dashboard or trend chart to see the full change log with affected columns and data types.
Metric anomalies¶
The custom metric value fell outside the predicted or configured range. Investigation depends on the nature of the metric being tracked.