Retrieval Quality
Also called Retrieval Effectiveness
How good your search actually is at surfacing the right stuff — the ceiling on any RAG answer.
Think of it like
A cook can only be as good as the ingredients delivered; bad groceries, bad dinner.
Example
A RAG bot gives wrong answers not because the model is dumb but because retrieval keeps handing it the wrong paragraphs.
How it actually works
Retrieval quality is the composite of recall, precision, and ranking — do the right passages show up, and near the top? It is the most common and most overlooked failure point in RAG: teams tune prompts for weeks when the real problem is that the relevant chunk never gets retrieved. Measuring it with real queries is the fastest way to find where a system is actually broken.
For product teams
Fix retrieval before blaming the model; it is usually where quality problems really live.
For engineers
Aggregate of recall@K, precision@K, and rank position; the upstream bottleneck for grounded generation.
Related
- Recall at K — Measured by recall and precision at K.
- RAG — Caps how well RAG can perform.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome