How Volumetric Works

Volumetric tracking watches the row count of a table or file over time. Once tracking is enabled on a container, Qualytics records its row count on a regular sweep, and Volumetric checks compare each new measurement against configured thresholds. Deviations from the thresholds are surfaced as anomalies.

Enabling volumetric tracking

Volumetric tracking is a per-container toggle in the container's Observability Settings. If a container has no volumetric tracking enabled, its card shows an empty heatmap with an Enable action to turn tracking on.

• Volume Tracking is the base toggle. It has to be on for any volumetric measurement to happen.
• Create AI Volumetric checks is a separate toggle. When on, Qualytics automatically authors volumetric checks once enough measurements have accumulated (the check pill carries an AI badge to indicate it was inferred).
• Turning Volume Tracking off stops new measurements from being recorded. Existing authored checks are preserved but stop receiving new data.

You can also author a Volumetric check manually from the Quality Checks flow. Manually authored checks do not carry the AI badge.

The measurement sweep

Qualytics runs a scheduler-driven sweep every hour at minute 30 (UTC). Each container gets one Volume measurement per UTC day: the first successful attempt of the day sticks, and later hourly ticks skip containers that already have a measurement for that UTC day. Large datastores can take a few hourly ticks to fully sweep before every container has recorded its measurement; a container that missed an earlier tick is picked up automatically on the next one.

Containers that are marked Inaccessible are excluded from the sweep. Three consecutive failed measurements will flip a container's status to Inaccessible. Only a subsequent successful measurement restores the container to Available: aborted or timed-out attempts do not clear the status, even if the underlying source has recovered.

Check properties

A single container can have more than one Volumetric check, each with its own window size:

• Measurement Period (window size, in days). The window over which the row count is measured. A 1 Day check compares each day's count against the thresholds; a 7 Days check compares weekly totals. Multiple checks with different window sizes can coexist on the same container.
• Measure. The measured value. For Volumetric checks this is always Row Count.
• Comparison. How the measurement is evaluated: Absolute Change, Absolute Value, or Percentage Change. See Comparisons for the details of each.
• Min / Max thresholds. The boundaries of the expected range. Values outside the range depend on the comparison type but always fall into one of two outcomes: raise an anomaly, or mark the measurement out-of-threshold on the chart.

What the chart shows

• Bars. Each bar is a measurement for the selected window. The bar's height represents the row count.
• Threshold lines. The chart overlays the Min / Max lines (or the equivalent for change-based comparisons).
• Salmon-tinted points. Measurements that fell outside the threshold range but did not raise an anomaly are colored salmon on the chart line. This is distinct from the heatmap coloring below.

The heatmap

Below the chart, the heatmap presents one square per day for the selected timeframe:

• Cyan. A measurement was recorded and no anomaly was raised.
• Orange. At least one anomaly was recorded for that day.
• Grey. No measurement exists for that day (tracking was off, the container was inaccessible, or the day predates the first measurement).

Hovering a square surfaces the date, the anomaly count, the last row count with a trend arrow versus the previous period, and the last modification time. For a day whose measurement has not landed yet (the current UTC day before the sweep reaches this container, or a day that missed its earlier tick), the row-count field in the tooltip renders as — until the measurement is recorded. That — means "not yet measured", not "zero rows".

Timeframe and grouping

The chart supports four timeframes and a Group By control that adapts to the selected timeframe:

The Report Date control lets you anchor the chart to a specific end date; the timeframe extends backward from that date.

Anomaly outcomes

When a measurement falls outside the configured thresholds, the outcome is one of:

• Anomaly recorded. The container's active anomaly count increases; the heatmap turns orange for that day; the standard anomaly workflows apply.
• Out-of-threshold, no anomaly. The measurement is flagged visually on the chart line (salmon color) but does not create an anomaly record. This can happen when the check is configured with tolerance rules that dampen isolated deviations.