Anonymized engagement summary: Power BI weld quality and project-controls reporting, rebuilt on Microsoft Fabric, for an industrial construction contractor (2026)
Delivered — dated delivery record and client correspondence on file (EPC-reported; documents private)First-pass editionAn industrial construction contractor (client not named) · 2026 · Project ID EPC-2026-ANON-CONSTR-PBIFABRIC
Last updated by Errin O'Connor, Founder & Chief AI Architect, EPC Group
Opening answer. An industrial construction contractor runs its weld quality (QA/QC) and project-controls reporting in Power BI. The reports draw on two outside systems, a weld-tracking system and a project-controls system, through an Azure Synapse data warehouse. Between January and August 2026 EPC Group:
- rebuilt and reconciled the Power BI reports on a warehouse connected end to end
- carried new fields through the weld data path
- moved both integrations from Azure Synapse to a Microsoft Fabric warehouse
- set up access and release management for the reporting team
Disclosure. This is an EPC-reported anonymized engagement summary. The client is not named, and neither are the third-party systems involved. The supporting records — the executed statement of work, the dated monthly delivery record and the client correspondence that corroborates it — are retained privately and are not published in the Evidence Center, so this account is outside the Evidence Center's documented-engagement counts. Every engagement fact below is taken from those documents.
Verification class
Delivered — dated delivery record and client correspondence on file (EPC-reported; documents private). The statement of work was executed in November 2025. The delivery record lists, month by month, what each team member delivered from December 2025 to August 2026. Where engineering task logs are missing (January, July and August 2026), the record is rebuilt from dated client correspondence. No client acceptance form or business-outcome measurement is claimed.
Buyer problem and scope
| Item | What the documents state |
|---|---|
| Buyer | An industrial construction contractor (client not named); stakeholders in IT, the PMO and Project Controls |
| Situation | Weld, welder and inspection data from a weld-tracking system and budget data from a project-controls system, integrated through an Azure Synapse warehouse feeding Power BI. A legacy data warehouse was being phased out, a job-naming change was underway across the estate, and reports needed new fields |
| Workstreams | The two data integrations; the migration of both from Azure Synapse to Microsoft Fabric; the Power BI reporting estate on top of them |
| Reporting assets | Six built, migrated or rebuilt: a QA/QC report, an executive report, an audit report, a PMO weld dashboard, a weld-data report and a project control budget allocation report |
Unknowns, not stated in the documents cited: user counts, data volumes, and any change in inspection or project-control outcomes attributable to the reporting.
Approach and artifacts
Fix the warehouse before the report. When the client asked for joint location on the reports, the field could not simply be added in Power BI. It was carried from the weld-tracking source through:
- the configuration query and staging
- the weld dimension
- the stored procedures and pipelines
- the reporting layer
A weld classification field went the same way in February. A new defect column followed in June, through staging, the inspection fact table, stored procedures and pipelines, and was then replicated into Fabric.
Rebuild, then reconcile. In February the warehouse was connected end to end. The weld, welder and inspection datasets were validated against source in sequence, and the QA/QC report was rebuilt: relationships, DAX logic, validation, then publish and functional checks. In March the executive report was reconciled against the QA/QC report, and all three major reports were confirmed to match. Refresh failures were diagnosed and fixed in the same month. When the client raised discrepancies — inspection counts, combined versus classified welds, weld-inch totals — each was worked against the client's verified numbers and closed with the fix confirmed.
Synapse to Fabric without stopping the reports. The Fabric warehouse was provisioned. Both integrations' schemas and stored procedures were migrated, with staging, dimension and fact pipelines built and tested end to end. Fabric stored-procedure compatibility issues were resolved at each stage. Capacity was procured and assigned, and a target-state architecture covering both integrations and future sources was issued to the client in April. Changes made in Synapse during the transition were replicated into Fabric.
A third-party dependency, handled in the open. The project-controls integration needed view definitions that turned out to be encrypted and permission-restricted. EPC Group established that, re-raised the request with the client's vendor in May, and the definitions arrived in July. Once they did, five more control-budget tables were staged and new KPI columns were added to the control-budget fact table for the PMO.
Access and release management. A Power BI deployment pipeline was proposed and tested across three stages with a dedicated workspace. A service principal and security group were created for warehouse access, tested through Power BI and handed to the reporting team. When a weld dashboard went blank in June, the cause was traced to a pipeline failure, fixed, and documented in a root-cause analysis.
Work products and record (dated, held privately):
| Item | Date |
|---|---|
| Executed statement of work | November 2025 |
| Written response to the client's year-end priorities (legacy warehouse phase-out, cost management, dashboard validation sequence) | December 2025 |
| Control-budget dimensional model review with the PMO and Project Controls | January 2026 |
| Rebuilt QA/QC, executive and audit reports; three-report reconciliation | February–March 2026 |
| Fabric migration strategy, then the Fabric architecture issued to the client | February; April 2026 |
| Fabric staging, dimension and fact pipelines, tested end to end | March–April 2026 |
| Power BI delivery report; deployment-pipeline proposal | May 2026 |
| Dashboard outage root-cause analysis; warehouse service principal and security group | June 2026 |
| PMO weld dashboard layout approved after client review rounds; self-service sandbox walkthrough | July 2026 |
Results and limitations
| Claim | Basis | Status |
|---|---|---|
| Six reporting assets built, migrated or rebuilt and validated against source | Delivery record | Delivered (EPC-reported; documents private) |
| The three major reports reconciled against one another and confirmed to match | Delivery record, March 2026 | Delivered (EPC-reported) |
| Both integrations built and tested end to end in a Fabric warehouse | Delivery record, April 2026 | Delivered (EPC-reported) |
| PMO weld dashboard layout approved by the client | Client correspondence, July 10, 2026 | On file (EPC-reported) |
| Business outcomes such as fewer weld defects or faster reporting | No document cited | Not claimed |
Provenance
- Sources: the executed statement of work (November 2025), the monthly delivery record (December 2025 – August 2026) and the client correspondence that corroborates it, held privately. Engagement ID: EPC-2026-ANON-CONSTR-PBIFABRIC, an alias unrelated to the client's name.
- Anonymization: industry only, under EPC Group's anonymized-case rule; the client, its third-party systems and vendors, locations, staff names and commercial terms are not published.
- Not claimed: client endorsement, or any measured business result.
- Author: Errin O'Connor, Founder & Chief AI Architect, EPC Group, who set the target-state architecture and approved the warehouse security model on this engagement.
Revision history
- 2026-10-25: anonymized engagement summary published from the engagement's statement of work and delivery record.
Related service
Enterprise Azure analytics platform · Power BI consulting · Microsoft Fabric consulting · Talk to a senior architect
Author: Errin O'Connor, Founder & Chief AI Architect, EPC Group · Published: 2026-10-25.