Custom user metrics report

Hi,

What you are seeing is expected and comes from how the auto-generated Usage Metrics report is calculated upstream versus what is exposed in the Usage Metrics dataset.

The standard report is effectively working on a sessionized, report-level semantic layer, while the dataset you connect to is event-level telemetry (page renders, interactions, navigation events, etc.). The built-in ViewsCount measure therefore tends to overcount compared to the standard report because it includes multiple events within a single viewing session.

In practice, the closest approximation is to reconstruct a session grain and then aggregate at report level:

Report Views =
COUNTROWS (
    SUMMARIZE (
        UsageMetrics,
        UsageMetrics[ReportId],
        UsageMetrics[UserId],
        UsageMetrics[SessionId]
    )
)

If SessionId is unreliable or not populated, you can approximate sessionization using time bucketing:

Report Views (Time Bucket Approx) =
COUNTROWS (
    SUMMARIZE (
        UsageMetrics,
        UsageMetrics[ReportId],
        UsageMetrics[UserId],
        FORMAT ( UsageMetrics[Timestamp], "yyyymmddhh" )
    )
)

Technically, you will never get a 1:1 match with the auto-generated report because Microsoft applies additional filtering and normalization before the metrics are surfaced. The recommended approach is to define a consistent session-based view definition and use that across custom governance or adoption reports.

If you can share which Usage Metrics tables you are using, I can help refine the session grain further.

Hope this helps.