Hybrid Search

Full definition

Result fusion is typically Reciprocal Rank Fusion (RRF), weighted score combination, or a learned re-ranker. Hybrid search is the standard recommendation from Microsoft Azure AI Search, Elasticsearch, and Weaviate because pure dense retrieval misses rare terms (product SKUs, error codes, person names) while pure keyword search misses paraphrase. In Microsoft's tests on nine BEIR datasets, hybrid retrieval scored 48.4 NDCG@10, against 45.0 for vector-only and 40.6 for keyword-only (Microsoft, September 2023).

Why it matters

Teams that ship vector-only RAG and complain about "the model can't find the right document" are usually one BM25 hybrid step away from a measurable retrieval gain (Microsoft measured 3.4 to 4.6 NDCG points over vector-only on BEIR and customer data), and one re-ranker away from a larger one (a further 11.7 points on its customer datasets). The fix is days of work, not weeks.

Example

An e-commerce search switches from pure dense to hybrid (BM25 + embeddings + RRF) and recovers exact SKU matches that the embedding model was paraphrasing into similar-but-wrong products — conversion lifts 6%.

Next steps

• Get a free workflow map →We map one of your workflows and give you a written verdict: software, a process change or AI. No obligation.
• AI consulting cost guide →Hourly rates and project budgets by firm tier, from independents to the Big Four.

Related terms

• Semantic Search
  Semantic search is information retrieval that ranks documents by meaning rather than exact-keyword overlap, using vector embeddings of query and documents to measure similarity in a learned semantic space.
• Re-Ranker
  A re-ranker is a second-stage retrieval model that re-scores a candidate set (typically top 50–100) from a fast first-stage retriever using a more expensive cross-encoder that jointly attends to query and document.
• Retrieval-Augmented Generation (RAG)
  Retrieval-Augmented Generation (RAG) is a technique where a language model retrieves relevant documents from an external knowledge base before generating a response, which reduces hallucination and enables citation of sources.
• Vector Database
  A vector database is a database optimized for storing and searching high-dimensional vectors (embeddings) by similarity rather than by exact match — the storage layer of every RAG and semantic-search system.

More deployment & operations terms

• Chunking Strategy
  A chunking strategy is the rule set used to split source documents into retrievable units before embedding, which directly determines retrieval recall and answer faithfulness in a RAG system.
• Agent Memory
  Agent memory is the mechanism by which an AI agent stores and retrieves information across turns or sessions, beyond the fixed context window of the underlying language model.