What it does

Lets report consumers pull the data behind a visual into Excel or CSV. In the service you choose between summarised data (what the visual shows) and underlying data (the unaggregated rows from the model). Dashboard tiles export to CSV only.

Key facts

• Summarised export gives only the columns and measures the visual uses, filtered as the visual is filtered. The first row of the file records the applied filters.
• Underlying data removes aggregation and adds other columns from the semantic model, so it is far larger and far more revealing.
• Documented service limits: Excel with live connections up to 500,000 rows, Excel .xlsx up to 150,000 rows, CSV up to 30,000 rows.
• The September 2026 feature summary flags increased Excel export row limits as coming soon. Check the figures above against the page before quoting numbers to a client.
• Exports respect row-level security and the sensitivity label. A live-connection file inherits the semantic model's label; a plain .xlsx inherits the report's.
• If a visual holds only measures from a measures-only table, underlying export returns one value per measure.
• Desktop exports CSV only. A dashboard tile exports to CSV, and if the tile comes from a labelled report you're warned and pointed at the report page instead.
• Admins and designers can restrict export per report and tenant-wide.

When to use / skip

Export is the pressure valve for "I just need it in Excel". Leave it on for analysts, and turn it off for reports with data users shouldn't be walking out with. If the real need is a recurring Excel view, give them Analyze in Excel against the semantic model rather than a monthly export.

Configuration decisions

• Tenant export settings and any per-report override.
• Whether underlying data export is allowed, which is the riskier of the two.
• Sensitivity labels that follow the exported file.
• Whether Defender for Cloud Apps policies block downloads to unmanaged devices.

Gotchas

• Users hit the row cap silently on large visuals and assume the file is complete.
• Underlying data exposes columns that aren't on the visual. RLS is respected, but column-level surprises are common.
• A displayed applied-filters row isn't a guarantee every listed filter shaped the data. Microsoft calls out specific cases.
• Exported files leave your governance boundary. Labels help only if the recipient's tooling honours them.

Consultant notes

• Put export on the security review checklist for any report that touches personal or commercial data. It's the most common leak path and the least discussed.
• When a user says the export is wrong, check the applied-filters row and the summarised-versus-underlying choice before you open the model.
• Offer Analyze in Excel early. It removes most of the export demand and keeps the model in charge of the numbers.

---

Revisit when the increased Excel row limits are published.