ING's Credit Risk department maintained a research database covering the bank's full credit portfolio — every loan, exposure, and counterparty relationship across a book of approximately €4 trillion. The team I led was responsible for that database: loading it, maintaining it, and producing the reporting services that Credit Risk depended on.
The work had two distinct dimensions. One was operational: keeping the database current, accurate, and available. The other was regulatory: delivering the stress-test scenario reporting that Basel II required.
The Context
Credit Risk is a demanding data environments in banking. The portfolio is large and complex: multiple asset classes, counterparty hierarchies, exposure types, and risk factors. Regulatory requirements specify exactly how scenarios must be applied and what the outputs must look like. The quant team that designed the stress-test models needed clean, consistent input data to satisfy demands of regulators.
When I joined, the data quality processes were ad hoc, following the ad hoc nature of the stress tests. There was tension between the technical team and the business stakeholders.
The Data Quality Programme
We designed and built a structured monthly data quality programme. The programme had two layers.
The first layer validated business logic against the data definitions and dictionaries: field-level checks, referential integrity, conformance to regulatory classification schemes. These were deterministic — a record either met the definition or it didn't.
The second layer applied statistical plausibility checks: volume trends across the portfolio, distribution shifts in key risk metrics, anomalies in counterparty exposure. These weren't hard failures but signals — patterns that warranted investigation before the data was trusted for reporting.
The reporting run did not proceed until quality was confirmed. This meant the regulatory outputs were always backed by verified data, there was no path to submission that bypassed the checks.
Team Development
Leading the team of four, the technical competence was never in question. What was missing was client orientation. The team had operated as a back-office function — receiving specifications, producing outputs, minimal direct engagement with the Credit Risk analysts who used those outputs.
I changed that deliberately. Over time, each team member developed the ability to sit with a stakeholder, understand what they were trying to analyse, translate that into a data request, and deliver it — without me in the room. By the time I left, that was standard practice. The team had raised its own seniority level by taking ownership of the full chain from business question to data output.
If your organisation needs data quality infrastructure that can support regulatory reporting demands, let's talk.