One of the largest apartment developers in the country had a financial planning model that took 11 hours to refresh. By the time the numbers landed, the day they described was already over.
The problem
A decade of accumulated history had left 1.5 billion legacy records inside the company's Vena environment. Every refresh dragged all of it. The close cycle inherited the delay, FP&A inherited the close cycle, and the finance team had learned to plan around an 11-hour blind spot. Deleting history is easy. Deleting history without breaking a single report, reconciliation or audit trail that depends on it is the actual job.
The approach
- Archive, not amputate. I led a large-scale archiving initiative: map every dependency first, preserve the history in queryable form, then remove it from the hot path. 1.5 billion records moved out of the working model with zero broken reports.
- Fix the process around the data. The REFINE initiative (Real Estate Finance Excellence) aligned pro forma deal models with live financial reporting, so investment decisions and reported results finally spoke the same language. Kanban boards and weekly status reporting gave executives real-time visibility into the work itself.
- A roadmap, not just a rescue. I designed a cloud-native modernization plan on lakehouse principles: Vena, pro forma, live financials and FP&A flows mapped into Bronze, Silver and Gold layers, with Power BI over Microsoft Fabric and OneLake to support predictive, CFO-level analytics.
The outcome
Refreshes went from 11 hours to 2, which materially accelerated the financial close. Partnering with finance, FP&A and investment teams, deal modeling workflows moved into centralized reporting structures with better governance and cleaner data. And the company left with an architecture roadmap that makes AI and predictive analytics a next step instead of a rewrite.
Performance problems are usually governance problems wearing a disguise. The 11 hours were a symptom. The missing archive strategy was the disease.
What this means for you
If your model refresh has become a planning constraint, the fix is rarely more compute. It is deciding what belongs in the hot path, preserving what does not, and doing it without breaking trust in the numbers. I can find your 11 hours.