What’s the difference between “findable” and “usable”?

Findable means people (and agents) can discover the right asset. Usable means they can successfully apply it in a workflow without repeated clarification, rework, or hidden assumptions.

Do we need a data catalog for usability?

A catalog helps, but usability is broader: documentation, examples, semantic definitions, access paths, and interoperability matter just as much.

How much documentation is “enough”?

Minimum viable: purpose, owner/contact, definitions, grain, keys, freshness expectations, known caveats, and 1–2 example use cases.

How do we make data usable for both humans and AI agents?

Use consistent naming, clear definitions, stable identifiers, and machine-readable metadata. Curate what agents can see so they focus on trusted assets.

What consumption formats should we support?

Support the workflows that matter most: BI tools, SQL access, extracts (Excel/CSV) when needed, APIs where appropriate, and semantic layers for shared metrics.

How do we avoid endless bespoke “one-off” datasets?

Anchor work on reusable products first (entities, shared dimensions, standardized metrics) and use platinum-style tailored outputs only for the most critical audiences.

What are the best feedback signals to prioritize improvements?

Combine:

• Usage analytics (queries, dashboard adoption)
• Search-to-success (did people find something and actually use it?)
• Support volume (repeat questions = poor usability)
• Quality incidents caused by misunderstanding (wrong joins, wrong filters)

How do we keep the catalog clean over time?

Treat it like a product portfolio: deprecate, archive, or delete low-value/unused assets and make certification visible.