A Fabric Pipeline Checklist Before You Trust Portfolio Dashboards
May 1, 2026By Justin K. O. Bridgeford, Founder & Principal, Strategy & Systems
Validate SQL and Fabric pipelines before Power BI goes live—ownership, row counts, and failure alerts for portfolio datasets.
Why dashboards fail before Power BI does
Power BI gets blamed when the upstream export was wrong on Tuesday. For property and insurance portfolios, the fix is rarely a prettier chart—it is a Fabric or SQL pipeline with validation rules, named owners, and refresh windows that respect close calendars.
This checklist is what we run before promoting executive dashboards tied to large exposure datasets.
1. Label system of record per field
Every column in your analyst export should map to a system of record: SQL warehouse table, Fabric lakehouse path, Salesforce object, or manual override with approver name.
If two teams calculate occupancy differently, the pipeline should not merge them silently—branch definitions or document the executive metric choice.
- Owner per metric (analyst, underwriting, operations)
- Written definition for portfolio vs account vs unit grain
- Change log when definitions shift after model refresh
2. Validate row counts and keys before load
Large portfolios surface problems at scale—duplicate keys, null property codes, and orphaned units. Build SQL checks that fail the job when thresholds breach, not dashboards that show red after leadership already met.
- Row count within expected band vs prior run
- Primary key uniqueness on property + unit + account
- Null rate caps on fields that feed DAX measures
3. Fabric jobs with observable failures
Automated pipelines should notify operations when they fail—email, Teams, or a log table analysts already review. Silent retries that drop rows are worse than a delayed refresh.
- Job naming standard per dataset and environment
- Retry policy documented with max attempts
- Failure digest to a named on-call rotation
4. Handoff package for Power BI developers
Analysts promote models faster when they receive a data dictionary, sample primary keys, and refresh SLA—not only a lakehouse path.
Include test cases for two DAX measures leadership will ask about in the first meeting.
- Metric dictionary with SQL source queries
- Refresh schedule aligned to portfolio close
- Known exceptions list (properties under renovation, etc.)
When you need outside help
If your team cannot answer who owns rent roll vs pipeline revenue before building Fabric jobs, start with a Digital Operations Audit. We map datasets, document gaps, and sequence fixes before a full Power BI rollout.
