Python · Power BI · dataprof · Data quality
Metric Evidence
A Power BI report that puts each metric next to dataprof evidence about its data.
Revenue changed: which segments contributed, and which data limitations affect the reading? A local Power BI project shows dataprof profiles per month and region, the snapshot’s coverage, and a check that the rows Power BI loaded are the rows dataprof profiled. Synthetic data, no Power BI Service, no paid licence.
Updated from public sources ·
Evidence beside the number, not instead of it
Three situations are planted in synthetic orders: a real drop in one region, category labels going missing in the same region and month, and an extract that stops mid-month. Partial months are compared on the same days, so the report shows −8.5% where a naive full-month comparison would claim −39.8%. Data observations describe the data; they are never presented as the reason revenue moved.
Checked against an independent reference
Each evidence measure picks exactly one scope, so overlapping profiles are never added together. A Python reference recomputes the results in integer cents, a read-only script reconciles the live DAX against it, and a mutation check changes one value at a time to prove the snapshot match breaks. The global quality score is deliberately unused.