TL;DR
Automated cataloging tools often fail because they treat governance as a software problem rather than an operating model problem.
TL;DR
Automated cataloging tools often fail because they treat governance as a software problem rather than an operating model problem.
2026-09-21 - Governance, Metadata, Operating Models
There is a recurring pattern in enterprise data architecture. An organization realizes its data landscape has become an untrackable maze of tables, reports, and pipelines. Leadership buys a shiny metadata management platform. Everyone expects that once the tool crawls the ecosystem and automatically maps lineage, trust will follow.
Six months later, the catalog is full of unassigned tables, outdated descriptions, and tags that nobody maintains. The tool works fine. The governance failure persists.
I have watched this play out from inside the room. The software can tell you where a piece of data came from, but it cannot tell you who owns its business meaning. When organizations treat governance as a tooling procurement exercise instead of an operating model design challenge, they end up with expensive digital filing cabinets that nobody uses.
Controls come first. If you plug automation into an ungoverned environment, you simply accelerate the creation of noise. Automated line-drawing tools map every single transformation, turning a messy data warehouse into a massive, interconnected spiderweb that is even harder to understand than before.
Tools are force multipliers. If your underlying operating model has clear ownership boundaries, defined accountability, and established review processes, a good metadata platform makes those processes faster. If your operating model is chaotic, the tool just catalogs the chaos.
Metadata tools love to ask for a point of contact for every asset. In practice, this often defaults to whichever engineer built the pipeline or whoever happened to touch the table last. That is not ownership. That is just an audit trail of technical labor.
True platform ownership means treating data outputs with the same rigor as any other balance-sheet asset. A table or a dataset needs a business owner who decides when it is deprecated, who is allowed to consume it, and what quality thresholds it must meet. Until an organization assigns those responsibilities as formal parts of someone's role, the catalog remains empty of real human accountability.
When governance depends on manual cleanup sprints or individual heroics from dedicated data stewards, the system is already broken. Sustainable platforms run on systems that make compliance the path of least resistance. If creating an undocumented table requires extra work or breaks a deployment check, people stop doing it.
We need to stop looking to software vendors to solve our human process problems. Fix the operating model first. Define who owns the data, establish what quality means for your specific ecosystem, and only then bring in the tools to scale those rules.