tool

Data Catalog

A searchable inventory of datasets with their schema, owner, meaning, freshness, quality and classification.

governancediscoverymetadata

The problem it addresses is that in any sizeable estate, the expensive question is not computing an answer — it is finding which of six similar tables is the one people actually trust, and who to ask about it.

The entries that make a catalog useful, roughly in order: an owner who is a named person; a description of meaning including the grain and the known caveats; freshness — when it last updated and how often it should; classification, so it is visible which columns are personal or regulated; lineage; and usage, since the most-queried table is usually the trusted one and that signal costs nothing to collect.

The failure mode is universal and predictable: a catalog populated once during a project and never maintained becomes confidently wrong, which is worse than absent because people rely on it.

So the design rule is that metadata should be harvested rather than entered. Schema, freshness, lineage and usage all come from the systems themselves. Only ownership and semantic description need a human, and those should be enforced at creation — a dataset without an owner should not reach production, in the same way a service without one should not.