BI semantic layer · rule powerbi-table-reads-unknown
Power BI tables reading unmodelled tables
A Metacenta review checks this under the rule Power BI tables read tables the warehouse layer builds. Everything below applies whether or not you ever commission one.
What this rule checks
For each loaded Power BI table, we read the warehouse table names its partition source navigates to. The rule flags a table when one of those names matches nothing the transformation layer builds or declares as a source. We compare names only, not schemas.
Why it matters
Its numbers sit outside the tested, documented pipeline. A change upstream cannot warn anyone that it will break the report.
How to fix it
Point Power BI tables at modelled tables. Point the table at the modelled table that serves it. If another pipeline builds that table on purpose, tell us, and the finding comes out of your report.
When it is fine to leave
A reference table maintained by hand and loaded directly, such as a sales-target sheet. Better still, load it through the pipeline as a seed, so it is tested too.
What we need to check it
The .tmdl files from a PBIP project, plus a warehouse-side artefact: manifest.json, a SQL project or a lineage export. Without one, the rule declines. Calculated tables, native queries and sources we cannot reduce to a table name are not judged. Findings are medium confidence.