Semantic Model Audit — Client...

Findings (8)

Missing RLS on 3 models with sensitive data

CRITICAL

Security

Sales Revenue, Marketing Attribution, and Supply Chain Ops models expose all data to all workspace members without Row-Level Security.

→ Implement RLS roles mapped to Azure AD groups. Priority: Sales Revenue (contains revenue targets and compensation data).

Marketing Attribution model exceeds 400MB

HIGH

Performance

At 412MB, this Import model is approaching the 1GB Pro license limit and causing 45-minute refresh times.

→ Partition the model by date or switch historical data to DirectQuery/Composite mode. Consider incremental refresh.

42 unused measures across all models

HIGH

Optimization

15% of all measures are never referenced in any report or dashboard, adding unnecessary complexity and refresh overhead.

→ Archive unused measures to a separate 'deprecated' display folder, then remove after 30-day verification period.

Circular dependency detected in Finance model

HIGH

Data Modeling

Table relationships between FactSales → DimProduct → FactInventory → DimProduct create a circular reference affecting calculation accuracy.

→ Redesign the star schema to eliminate the circular path. Use bridge tables if many-to-many relationships are needed.

8 duplicate measure definitions

MEDIUM

Governance

Measures like 'Total Revenue' and 'Revenue Total' exist with identical DAX in multiple models, creating maintenance burden.

→ Consolidate into a shared semantic model using composite models or DirectQuery chaining.

No descriptions on 85% of tables and measures

MEDIUM

Governance

Only Finance Consolidated has proper documentation. Other models lack descriptions, making self-service discovery difficult.

→ Add descriptions to all tables and key measures. Use Tabular Editor for bulk updates.

No incremental refresh configured

MEDIUM

Performance

All 4 Import models do full refresh daily, even though only recent data changes. This wastes capacity and extends refresh windows.

→ Configure incremental refresh with 3-year archive, 7-day incremental window on all Import models.

DirectQuery performance below threshold

LOW

Performance

Finance Consolidated DirectQuery averages 4.2s per visual query, above the 2s recommended threshold.

→ Add aggregation tables for common queries. Review the SQL source query performance.

Recommendations

1Implement RLS on Sales Revenue, Marketing Attribution, and Supply Chain models immediately

2Partition Marketing Attribution model and configure incremental refresh

3Audit and remove 42 unused measures (15% of total)

4Resolve circular dependency in Finance Consolidated model

5Add descriptions to all undocumented tables and measures using Tabular Editor

6Consolidate 8 duplicate measures into shared definitions

7Configure incremental refresh on all Import-mode models

8Add aggregation tables to Finance DirectQuery model to improve visual query performance

name	workspace	tables	measures	rls_enabled	storage_mode
Sales Revenue Model	Sales Analytics	14	52	False	Import
HR Metrics Dashboard	HR Analytics	8	28	True	Import
Finance Consolidated	Finance	22	95	True	DirectQuery
Marketing Attribution	Marketing	11	38	False	Import
Supply Chain Ops	Operations	18	67	False	Composite

Findings (8)

Recommendations

Models Audited