1.  [](/)
2.  [Diagrams](/diagrams)
3.  Keywords catch what vectors miss

[Back to Diagrams](/diagrams)IR Foundations

# Keywords catch what vectors miss

Hybrid search fuses keyword (BM25) and vector results so each covers the other's blind spots.

![Hybrid retrieval combines keyword and vector search because each fails alone: keyword (BM25) search misses paraphrases with zero word overlap, while dense embeddings blur exact identifiers like error codes. Mature systems run keyword, vector, and metadata signals together and fuse the ranked lists, reciprocal rank fusion or learned weights, so each signal covers the others' blind spots.](/diagrams/foundations/hybrid-retrieval.webp)Full screen

Hybrid search fuses keyword (BM25) and vector results so each covers the other's blind spots.

Every search system eventually learns the same lesson twice.

First: keywords fail. A user asks about 'refund policy', the document says 'returning purchases', zero words overlap, no result. So you add embeddings, and semantic matches start landing.

Then: vectors fail. Someone searches an error code, a part number, a person's name, and the embedding blurs it into a vague neighborhood instead of hitting the exact match sitting right there. BM25, an algorithm from 1994, handles that case better than the 2026 one.

So mature systems run both and merge the results, then filter on metadata. A financial-docs benchmark this month had BM25 alone beating a frontier embedding model on recall, and hybrid plus reranking beating both by a wide margin.

Mixpeek retrievers run keyword, vector, and metadata stages in one pipeline because that's what production search converges to anyway.

Hybrid isn't hedging. It's the answer.

## Where this diagram appears

*   [Glossary: Hybrid Search →](/glossary/hybrid-search)

Run this on your own data

Mixpeek turns video, images, audio, and documents in your object storage into searchable, timestamped results through one API.

[Search your own data](https://studio.mixpeek.com/namespaces/create?mode=managed&utm_source=homepage&utm_medium=diagram_detail&utm_campaign=mvs_onramp)

## Related diagrams

[

![The unstructured-data problem: the large majority of enterprise data is unstructured (video, images, audio, and documents) that keyword systems and relational databases cannot search by meaning until it is turned into embeddings and indexed. The raw files are stored but not answerable.](/diagrams/foundations/unstructured-data.webp)

99% of your data has no rows



](/diagrams/unstructured-data)[

![An embedding turns meaning into coordinates: a model reads 'dog', 'puppy', and 'invoice' and outputs vectors, so 'dog' and 'puppy' land near each other while 'invoice' lands far away. Text, images, audio clips, video frames, and faces all map into one shared coordinate space where distance measures difference in meaning.](/diagrams/foundations/embeddings.webp)

An embedding is meaning turned into coordinates



](/diagrams/embeddings)[

![Semantic search as nearest-neighbor lookup: embed the query into the same vector space as your content, measure distance to every indexed point, and return the closest top-K results with no keyword matching. Exact brute-force compares against everything and is accurate but slow, while approximate nearest-neighbor (ANN) indexes like HNSW and IVF check a smart fraction of the space and return 95-99% of true answers in milliseconds.](/diagrams/foundations/similarity-search.webp)

Search by meaning = nearest neighbors



](/diagrams/similarity-search)