Defined term
Hybrid search
Combination of keyword (BM25) and vector (embedding) retrieval, often re-ranked.
Hybrid search merges results from a sparse retriever (BM25, keyword) and a dense retriever (embeddings) using rank fusion (RRF) or weighted scoring. A reranker (cross-encoder or LLM) often refines the top results. Hybrid+rerank is the strongest baseline for production retrieval and the default we use in Build engagements.
When it matters
When your corpus contains exact identifiers (product codes, error messages, account numbers) AND natural language content. Pure semantic search misses exact-match needs; pure keyword misses conceptual queries.
Real example
A product support search combining BM25 (catches exact error codes like 'E-2031') with semantic embeddings (catches 'my screen flickers'). Results merged via Reciprocal Rank Fusion. Improvement: +18% MRR vs either method alone.
KPIs to watch
MRR (mean reciprocal rank, >0.7 target), recall@10 vs each method individually (hybrid should beat both by 10-20%), latency P95 (<150ms).
Related terms
Semantic search
Search by meaning, not by keyword overlap.
Embeddings
Numerical vectors that represent the meaning of a text, image, or other piece of content.
Agentic AI
AI systems that can plan, take multi-step actions, and use tools to complete tasks autonomously.
Autonomous agent
An AI agent that completes a defined task without per-step human input.
See it in action
We use this every week
Book a 30-min call and we'll walk you through how Hybrid search shows up in a real engagement we're running.
Book a 30-min call