Most Reconciliation Problems Are Data Problems in Disguise

We recently completed a major Nostro reconciliation transformation for a global financial institution — a clean-slate rebuild of the entire data architecture across an international branch network. Here's what we found when we got under the bonnet.

The source system had a global data layer that was similar but not identical across regions. Each data feed also contained dozens of distinct data source types, each with its own structure — all being pushed through a single flat mapping into generic reference fields, with no consistency across regions or data sources.

A critical identifier could be buried inside a longer text field, or fragmented across two or three separate reference fields. The reconciliation engine had nothing reliable to match on.

Unmatched items weren't routing to the right business areas. They were falling into a global exception bucket — requiring manual re-allocation before anyone could even begin investigating the break.

Operations teams were drowning in work that wasn't real work. And eventually, regulators noticed.

The fix required going all the way back to the data.

Part 2 coming next week — how we rebuilt it.