RAG & Retrieval · Ground answers in evidence
What metadata should every chunk carry?
Short answer
At minimum: source document, section path, page or anchor, version, timestamps and access-control tags. Metadata powers filtering, citations, freshness and deletion.
Metadata is how a chunk stays connected to the real world. Without it you cannot filter by tenant, cite a page, expire stale content or delete a document's chunks when the source is removed. Embedding text alone is not enough.
- Identity: document ID, chunk ID, parent section and position.
- Display: title, URL or page number for citations.
- Governance: tenant, ACL groups, classification and retention.
- Freshness: created, updated and indexed timestamps plus source version.
Follow-ups you should expect
Interviewers use these to check whether the first answer was understanding or memorisation.
- 1How would you enforce tenant isolation with metadata?
- 2What metadata does a citation need to be trustworthy?
- 3How do you handle a document whose permissions change?
Say it in the interview
Every chunk needs identity, provenance, permissions and freshness. Identity lets me cite and delete precisely, provenance gives the page or anchor, permissions let me filter before the model sees anything, and timestamps let me expire stale content. I filter on metadata during retrieval rather than trusting the model to ignore unauthorised text. Without that, RAG becomes a data-leak path.
Use this as a model, not a script. Saying it in your own words is what makes it survive follow-up pressure.
Related questions
Reading is step one
Now say it out loud.
Understanding an answer and delivering it under pressure are different skills. Rehearse this topic in a live RAG & Retrieval mock interview and get scored feedback on depth, clarity and confidence.