Anonymized engagement summary: urgent fix of segment-level backlog errors in a 27-table CARR/ARR Power BI model for a healthcare technology company (2025)
Contracted — countersigned (EPC-reported; documents private)First-pass editionA healthcare technology company (client not named) · 2025 · Project ID EPC-2025-ANON-HTECH-PBIRESCUE
Last updated by Errin O'Connor, Founder & Chief AI Architect, EPC Group
Opening answer. In March 2025 a healthcare technology company's VP of FP&A countersigned an urgent Power BI statement of work with EPC Group. The company's contracted and annual recurring revenue (CARR/ARR) model had started showing wrong backlog values by product segment after the January 2025 dashboard update, while the total backlog stayed correct. The statement of work traces the likely cause to:
- complex relationships across 27 tables
- reliance on many measures
- possible formula inconsistencies
The scope is to find the root cause, fix the DAX and relationships, prevent a repeat, document the fix and hand it over to the company's own team.
Disclosure. This is an EPC-reported anonymized engagement summary. The client is not named. The countersigned statement of work is retained privately and is 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 that document.
Verification class
Contracted — countersigned (EPC-reported; documents private). The client's VP of FP&A signed the statement of work electronically in March 2025. It records the problem as the client described it in a preliminary discussion. No fix record, root-cause report or acceptance is cited or claimed here.
Buyer problem and scope
| Item | What the statement of work states |
|---|---|
| Buyer | A healthcare technology company (client not named); finance-led (FP&A) |
| Symptom | After the January 2025 dashboard update, backlog values by product segment were wrong while the total backlog was right |
| Suspected cause | Complex relationships across a 27-table model, many dependent measures, possible formula inconsistencies |
| Urgency | Time-sensitive; subject-matter experts assigned to the backlog issue first |
| Fix scope | Troubleshoot and fix the segmentation errors; review and optimize the DAX so results hold across product segments; remove unnecessary relationships in the 27-table model; add measures that prevent future calculation breakdowns |
| Reporting scope | A root-cause analysis; best-practice recommendations for data modelling and calculations; a view on restructuring for performance, automation and self-service; the feasibility of moving the backlog calculations to Microsoft Fabric, and of Copilot for Power BI later |
| Hand-over | Real-time walkthroughs and documentation for the company's IT staff on the changes and how to maintain the model; activity reports |
| Flexibility | Remaining time usable for further Power BI support — dashboard enhancements, report optimization and data-model work |
Unknowns, not stated in the document cited: the specific defect found, how long the fix took, and whether the Fabric recommendation was taken up.
Approach and artifacts
Totals right, segments wrong, is a model problem. When a grand total reconciles but its breakdown does not, the arithmetic is usually fine and the filter path is not. The statement of work therefore starts with the relationships and measures, not the visuals:
- an in-depth assessment of the CARR/ARR backlog calculations
- the root cause of the segment-level discrepancy
- the table relationships and data structure, looking for simplification
Fix, then make it hard to break again. Corrected DAX and simplified relationships come with measures intended to stop the same breakdown recurring. A root-cause report records what failed and what was changed, so the finance team is not dependent on memory the next time the model is touched.
Advice that costs nothing extra. The Microsoft Fabric and Copilot for Power BI recommendations are written into the statement of work as advice outside the consulting hours, so the urgent fix gets the time it needs.
Document (held privately):
| Document | Date |
|---|---|
| Statement of work, urgent Power BI consulting and support (countersigned) | March 2025 |
Results and limitations
| Claim | Basis | Status |
|---|---|---|
| The company contracted EPC Group to diagnose and fix the CARR/ARR backlog errors | Countersigned statement of work, March 2025 | Contracted (EPC-reported; documents private) |
| The defect was fixed, a root-cause report issued, or knowledge transfer held | No document cited | Not claimed |
Provenance
- Source: the countersigned statement of work (March 2025), held privately. Engagement ID: EPC-2025-ANON-HTECH-PBIRESCUE, an alias unrelated to the client's name.
- Anonymization: industry only, under EPC Group's anonymized-case rule; the client's name, products, figures, staff names and commercial terms are not published.
- Not claimed: client endorsement, completion, or any measured result.
- Author: Errin O'Connor, Founder & Chief AI Architect, EPC Group, who signed the statement of work for EPC Group.
Revision history
- 2026-10-25: anonymized engagement summary published from the engagement's countersigned statement of work.
Related service
Power BI admin, governance and support · Power BI consulting · Talk to a senior architect
Author: Errin O'Connor, Founder & Chief AI Architect, EPC Group · Published: 2026-10-25.