Review measurements, asset health, and events

Trace an insight back to its evidence

A chart can show that something changed, but it cannot establish on its own whether the change was operational, expected, or caused by incomplete data. Vega provides several connected views for answering that next question: Inventory identifies the component, the measurement calendar shows coverage, Health presents detector findings, and the Event Log preserves the related event details.

This guide follows that evidence trail in order, starting with the exact component and timestamp from the result you are investigating.

Start with the exact component and timestamp from the visualization. If the scope is still uncertain, use Select metrics, meters, and sites.

Find the component in Inventory

1. Open Inventory from the left navigation.
2. Search for the site.
3. Review its Health badge.
4. Open the site and select Components.
5. Locate the component by name and ID in the hierarchy.
6. Select its name to open the component details.

The site table and component tree can show OK, Info, Warning, Major, Critical, Issues, No status, or Unknown, depending on the available analysis. No status is not the same as OK; it means no usable status summary is present.

Inspect measurement coverage

1. In the component details, select Measurements.
2. Keep the view on Electrical for electricity measurements.
3. Select the year with the previous and next controls.
4. Read Last synced and Current sync window.
5. Use the legend to identify Completed, Pending, Failed, Excluded, and Outside window days.
6. Look for dashed Partial coverage (<100%) and orange No coverage (0%) outlines.
7. Select a day to inspect its timestamped values in the day-detail table.
8. Select Refresh view after a synchronization job completes.

The calendar reports availability and coverage. It does not prove that a present value is operationally correct. Check units, directions, repeated values, and an external source when accuracy is critical.

If synchronization is running, Vega shows a status notice and stream labels. The lower section lists recent synchronization jobs, scanned days, completion state, and any error text.

Use synchronization actions cautiously

Trigger Sync asks Vega to synchronize missing data for the whole site. Use it only when you are authorized to start connector work.

Resync All is destructive: the confirmation states that it deletes measurement history for the whole site and may remove data that a later synchronization cannot restore. Do not use it as a routine troubleshooting step. Escalate persistent gaps to the data owner or support team.

Understand Sentinels and detector findings

Sentinels are Vega’s automated checks of synchronized measurements and related analytical inputs. Each check is a detector. When a detector’s conditions are met, it creates a finding with a title, severity, summary, and supporting facts. Those findings feed the statuses shown in Health and the corresponding entries in the Event Log; these views are therefore two presentations of the same underlying analysis rather than independent diagnoses.

The available detectors depend on the site’s assets, data, enabled product features, and account configuration. They can identify:

• missing or incomplete actual and historical measurements,
• completion of an initial synchronization or a gap between synchronization windows,
• missing expected-production inputs,
• unusually low radiation that makes weather-sensitive comparisons unreliable,
• PV production close to zero despite available radiation,
• production that falls sharply against a recent baseline or differs materially from expected production, and
• a panel or microinverter performing materially below a sufficiently complete group of comparable peers.

A finding means that the detector’s configured rule matched; it does not by itself establish the physical cause. Some checks require a minimum absolute difference as well as a percentage threshold, enough complete data, or a sufficiently large and complete peer group. Low radiation, missing inputs, initial synchronization, and stronger overlapping findings can also prevent or suppress other detectors. Consequently, do not interpret the absence of one expected finding as proof that the asset was operating normally.

Each detector has a configured severity—Info, Warning, Major, or Critical—and some accounts use custom thresholds, severity levels, or enabled states. Use the values and thresholds shown in the finding instead of assuming that another account or an earlier event used the same rule. If your permissions expose Sentinels in account settings, that page shows the rules configured for the active account; changing them is an administrative action and changes future analysis behavior.

When investigating a finding:

1. Note the detector title, analysis date, severity, and analyzed-record count.
2. Read its supporting facts, including measured values, baselines, differences, thresholds, coverage, or suppressed checks when present.
3. Confirm the same date and component in Measurements.
4. Check the component hierarchy and configuration in Inventory.
5. Locate the matching entry in the Event Log using the detector or component name.
6. Select Ask AI only after establishing that evidence, and ask Vega to separate detector facts from possible causes.

An Info finding can still explain why another comparison did not run, while a Critical finding still requires confirmation against measurements and operational context. Severity communicates the configured importance of the matched rule, not certainty about its cause.

Review site health

1. Return to the site and select Health.
2. Read the site status summary.
3. Follow any linked component name in the summary.
4. Select Ask AI when you need Vega to explain the site status while keeping the Health page visible.

The site summary is a shortcut to affected components. Use the component-level history for dates and individual detector findings.

Review component health history

1. Open the component and select Health.
2. Use the left and right controls to move through 14-day history windows.
3. Select the day that contains the chart observation.
4. Review Analyzed At, Highest Severity, and Analyzed Records.
5. Read each detector card and its supporting facts.
6. Select Ask AI on a finding for a contextual investigation.

No health status for this day means the day has no result. No alerts detected for this day means an analysis result exists but contains no detector findings. Keep that distinction when documenting evidence.

Search the Event Log

1. Open Event Log from the left navigation.
2. Use Search for a detector name, component name, ID, or text from the finding.
3. Filter Severity by Info, Warning, Major, or Critical if needed.
4. Filter Site to one or more locations.
5. Expand a row with its chevron.
6. Review the event Body and Payload.
7. Use the paginator to inspect earlier pages.
8. Select Clear filters when finished.

The Event Log stores filters in the browser, so a later visit can still be narrowed by an earlier search. Clear them before concluding that no event exists.

Build an evidence trail

For each investigation, record:

• the chart or table value and timestamp,
• the component name and ID,
• measurement coverage for that day,
• health severity and detector facts,
• matching event-log entries, and
• operational context supplied by the site team.

Then ask Vega to explain only what those sources support. Continue with Work with notifications and contextual investigations when the issue originated from the notification bell.