← Files Power BI Report CopilotARCHIVED FILE

references/performance-evidence.md

899 Bytes · Nov 20, 2026 · 06:46 UTC

↓ Download file

See the change to this file →

# Multi-layer performance troubleshooting
Observe source execution, connector/Power Query, semantic-model storage, DAX, report visual/tablix processing, renderer and capacity separately. Collect minimal evidence appropriate to the failing layer: source plan, Power Query Diagnostics, Performance Analyzer, DAX Studio/VertiPaq, refresh duration, capacity signals and export timing.
Classify the slow path before changing it. Big models can be slow from poor relationships/cardinality, high-cost DAX iterators, overbroad visuals, chatty DirectQuery, paginated dataset retrieval or renderer complexity. Measure before and after and preserve correctness benchmarks. A hypothetical optimization is not a measured result.
Suggested result fields: baseline, environment, query shape, concurrency, suspected layer, proposed experiment, expected direction, actual measurement, regression check and rollback.

SHA-256: 317b0504bfc1baa5f28ca6b51a08903318ede5d246f2442d5008f79f870eee6d