The Metadata Your Knowledge Base Needs from Day One

A practical approach

Step 1

Define a minimum document record with a stable ID, source location, content hash, owner and effective or revision date where available.

Step 2

Add fields that affect filtering, such as language, product, audience and access group. Derive access from trusted records rather than free-text guesses.

Step 3

Carry the document and section identifiers through chunking and generation. Make it possible to trace an answer back to the exact source version.

Worked example

A training organisation stores programme guides for students and corporate teams. Both mention onboarding, but they describe different processes. An audience field can narrow retrieval before generation, while a version field helps the reader understand whether the quoted steps apply to the current programme.

A mistake to avoid

Too many optional fields can create inconsistent records that no one maintains. Start with fields that drive a concrete decision. Missing metadata should trigger a defined handling rule, especially when access or currency is uncertain.

Try it yourself

Take ten documents and fill in your proposed metadata schema. Note every field that is difficult to determine, then decide whether it needs an owner, a validation rule or removal from the minimum schema.

Your implementation checklist

• Stable IDs do not change with file names.
• Revision and upload dates are distinct.
• Access metadata comes from a trusted source.
• Answers can identify a source version.