New — every answer now cites its sources
All posts
RAG6 min read

Why retrieval quality beats model size

Teams obsess over which model to use. The real lever in document Q&A is what you feed it — chunking, deduplication, and metadata.

PR

Priya Raman

Founder · Lumen

Every week a new language model tops the benchmarks, and every week the same thing happens in production: a team swaps the model, accuracy barely moves, and the roadmap slips. When your product answers questions over documents, the model is rarely the bottleneck. Retrieval is.

Garbage in, confident garbage out

A state-of-the-art model given the wrong passage will answer fluently and incorrectly. That's worse than failing loudly, because users can't tell the difference. The fix isn't a bigger model — it's a better index: clean text extraction, sensible chunk boundaries, and metadata that survives the trip from Drive to vector store.

  • Chunk on semantic boundaries (headings, sections), not fixed token counts.
  • Deduplicate near-identical versions of documents — old drafts poison answers.
  • Keep page numbers and section titles attached to every chunk, forever.
  • Re-index on change; a stale index quietly erodes trust.

How we measure it

We evaluate Lumen with a simple discipline: every answer must either cite a source or decline. When retrieval quality went up 18% after reworking our chunking, citation accuracy followed — and 'I don't know' rates fell by half. Users don't need the biggest model. They need the right paragraph.

The best RAG system is a search engine that happens to talk.
Internal rule of thumb

Keep reading

8 min readTutorial

The 30-minute RAG setup for your team wiki

A practical walkthrough: connectors, chunking, and the three questions to test first when standing up document Q&A for your team.

Tomás Laifer · Engineering

5 min readCase study

How we cut support resolution time by 43%

A close look at the numbers behind one team's first 90 days with document Q&A — what moved, what didn't, and what surprised us.

Daniel Reyes · Support Lead

Your documents already know. Start asking.

Join 80+ teams getting source-backed answers from their own knowledge. Free up to 200 pages — set up in two minutes.

No credit card required. We'll never share your email — privacy policy.