Martijn van Velzen, Tamas Marton

Most data teams set out to build a well-documented, well-tested, and cleanly governed data platform — few actually get there. NIBC, a Dutch bank, did. In this session, NIBC’s data team will share how they built a dbt Mesh hub-and-spoke architecture on Databricks — a central Cloud Data Platform hub handling all source ingestion, feeding domain-specific spokes for Finance, Risk, Retail, and Listing — anchored by a banking-specific conceptual data model (parties, products, collateral) shaped directly by regulatory and financial reporting requirements. The result is documentation coverage far above the industry norm (where 10% is typical), driven by data delivery agreements that mandate data dictionaries with GDPR and sensitivity classifications at ingestion, plus disciplined use of dbt exposures, clean environment separation, and cross-project governance that most banks are still trying to design.

Attendees will walk away with concrete, replicable practices for closing the gap between “we want to be well-documented and well-tested” and actually being so. Attendees will also learn how a multi-project dbt mesh can keep domain teams autonomous while staying governed centrally, and where NIBC is still evolving even within a mature setup — from tightening production job reliability to exploring native model versioning.