The mess
Three systems, each correct by its own logic, each producing a different net asset value for the same fund on the same day. The fund administrator priced on its own cut-off. The order management system included trades not yet settled. The accounting ledger booked on settlement. Nobody was wrong, and nobody could agree.
Month-end had quietly reorganised itself around this. The first days went on establishing whose figure to use; the analysis the investment committee actually wanted got the time that was left.
The method
The first week was not technical. It was writing down what “NAV at month end” was supposed to mean — which cut-off, which prices, which trades — and getting operations, finance and the portfolio managers to agree on that sentence in writing.
Once the definition existed, the platform could enforce it: ingest all three sources, map each one to the agreed definition, and reject anything that could not be mapped rather than quietly guessing. The rejections were the valuable part. They surfaced breaks that had been rolled forward from one month to the next.
The result
One agreed NAV at the top of the report, the three source figures under it, and an explicit reconciliation line showing each difference and its cause. When the figures diverge, nobody has to go hunting: the report already says why.
What made it stick
Every break now has a named owner and a due date, and the reconciliation runs every day, not only at month end. The definition document lives next to the code that enforces it, so a change to one is a visible change to the other.