Understand statistics and unusual patterns

Use statistics to find questions, not declare answers

Minimum, maximum, average, and ranked values help you find the moments worth examining in a large dataset. They are starting points: a maximum may be a genuine demand peak, a normal operating event, or an artefact of incomplete data.

The detailed chart window calculates a short summary from the values it currently renders and can narrow a refreshed result to its highest or lowest values. This guide explains what those numbers represent and how to move from an extreme value to a defensible investigation.

Open the chart window

In Chat, locate Vega’s response containing the chart or table and select Open, identified by the external-link icon in the response actions. This opens the detailed chart window. The Statistics control and Min, Max, and Avg summary described below are in its right-hand panel. See Open and adjust the detailed chart window for the complete opening flow.

Read Min, Max, and Avg

Before interpreting the summary, confirm the period and interval, metrics and components, and active filters.

The Statistics panel shows:

• Min — the smallest numeric value in the currently rendered datasets, with its time label.
• Max — the largest numeric value in the currently rendered datasets, with its time label.
• Avg — the arithmetic mean of all numeric points across the rendered datasets.

Missing or non-numeric points are omitted from this calculation. When more than one metric or component series is displayed, the summary scans all of them together. In that case, identify which series contains the extreme by switching to Table or simplifying the scope.

The average is an average of displayed points, not automatically a time-weighted average and not an energy total. Its meaning therefore depends on interval, metric, filters, and number of series.

Rank high or low values

Use the statistic selector to choose Top N, Bottom N, or None. When using Top or Bottom, enter a positive value for N, then select Refresh.

For example, to find demand peaks:

1. Select Peak Power and Load.
2. Select one component and a complete period.
3. Choose the interval appropriate to the decision.
4. Set Statistics to Top N and set N to 5.
5. Select Refresh.
6. Open Table and record the timestamps and values.

The ranked result is based on the selected interval. The top five daily values and the top five 15-minute values answer different questions.

Investigate an unusual point

1. Record the component, metric, unit, timestamp, interval, and active filters.
2. Narrow the period around the point.
3. Change to a finer interval and refresh.
4. Inspect the table rows immediately before and after it.
5. Check the component’s Measurements calendar for partial or missing coverage.
6. Review its Health findings for the same day.
7. Search the Event Log by site, severity, and relevant terms.
8. Ask Vega to distinguish observed evidence from possible causes.

A single high value can be a brief spike, the start of a sustained change, a restarted process, a clock-boundary effect, or a data-quality problem. The surrounding rows show whether it persisted.

Check whether the comparison is fair

Before calling a value unusual, compare like with like:

• complete periods of the same duration,
• the same component boundary and measurement direction,
• the same unit and interval,
• equivalent operating hours or day types, and
• similar data coverage.

If filters are active, apply the same filters to the baseline. If the selected period includes very few points, say so explicitly.

Ask Vega for a defensible explanation

Use a prompt that requires evidence:

For the selected component, explain the maximum shown on 18 August 2026. Show the surrounding hourly values, relevant health findings and events, data coverage, and at least two possible operational explanations. Label each explanation as supported or hypothetical.

Then follow Verify Vega’s answers before acting. A statistical extreme is a starting point for investigation, not an automatic anomaly diagnosis.