When Bad Data Architecture Becomes a Regulatory Problem

How a clean-slate nostro reconciliation rebuild eliminated manual touchpoints, restored data integrity, and handed a global bank a governed architecture it could own

Most reconciliation problems are data problems in disguise. The breaks aren't real. The exception queue isn't overflowing because the operations team is understaffed. It's overflowing because the data going into the reconciliation engine is badly structured, inconsistently mapped, and has never been properly governed. The engine is doing exactly what it was configured to do. The configuration is the problem.

That distinction matters — because the typical response to poor match rates is more rules, more manual process, more people. None of those fix a data architecture problem. They add cost on top of one.

This engagement started as a data quality issue at a major global financial institution and escalated into a formal regulatory remediation programme. What the regulators could see — inconsistent data across regions, a global exception bucket absorbing items that couldn't be routed, operations teams spending hours each day chasing breaks that were mapping artefacts rather than genuine reconciling items — was the visible consequence of an architecture that had never been properly built.

We rebuilt it from scratch. Across a global branch network, running on SmartStream TLM, with potentially dozens of distinct data source types per regional file and a single general ledger platform whose field structures varied subtly between branches. The work ran across two years and followed a structured six-stage framework: diagnose and stabilise, define Critical Data Elements, design a Golden Source Data Framework, apply AI-assisted data lineage analysis to identify common patterns at scale, implement a two-tier mapping architecture, and fully rebuild the auto-allocation logic.

The outcome was a nostro data architecture the institution's own teams can extend independently. The global exception bucket is gone. Items route directly to the correct business area from point of entry. Entry reference maps to entry reference, value date to value date, currency to currency — consistently, across all regions and all product types.

And the framework, template library, and Golden Source Data Framework are all client-owned on exit.

The fix was never more reconciliation rules. It was going back upstream and building the data foundation properly.

ReconIQ_Nostro_Case_Study_LinkedIn.pdf

Download PDF • 318KB