# Retrieval Score Explanations and Term Vectors
Status: public
Confidence: medium (0.685) (verified)
Last verified: 2026-06-03
Generation: ai_structured


## TL;DR

Score explanations and term vectors help agents debug why retrieval returned a document, missed a document, or ranked evidence unexpectedly.

## Core Explanation

RAG failures are often blamed on prompts even when the retrieval layer selected weak evidence. Score explanations and term vectors give agents a lower-level view of lexical matching, term statistics, and document-specific scoring inputs.

Agents should preserve the query, filters, analyzer, candidate document ID, explanation payload, term vectors, shard context, reranker stage, and final evidence order. Explanation output is diagnostic evidence, not a guarantee that the ranking is desirable.

## Source-Mapped Facts

- Elasticsearch Explain API documentation says the API gives information about why a specific document matches or does not match a query. ([source](https://www.elastic.co/docs/api/doc/elasticsearch/operation/operation-explain))
- Elasticsearch Explain API documentation says the API computes a score explanation for a query and a specific document. ([source](https://www.elastic.co/docs/api/doc/elasticsearch/operation/operation-explain))
- Elasticsearch Term Vectors API documentation says the API returns information and statistics about terms in fields of a particular document. ([source](https://www.elastic.co/docs/api/doc/elasticsearch/operation/operation-termvectors))

## Further Reading

- [Elasticsearch Explain API](https://www.elastic.co/docs/api/doc/elasticsearch/operation/operation-explain)
- [Elasticsearch Term Vectors API](https://www.elastic.co/docs/api/doc/elasticsearch/operation/operation-termvectors)