rag medium complexity scale
Hybrid Search
Hybrid search combines semantic vector retrieval with keyword or full-text search.
Decision
Use hybrid search when users mix fuzzy intent with exact product names, IDs, or domain terms.
Use when
- Documentation search
- Enterprise knowledge bases
- Product support
- Technical content with exact terms
Avoid when
- Pure semantic discovery
- Very small corpora
- Systems without search evaluation
- Cases where exact search already solves the problem
Why hybrid search exists
Vector search is good at meaning. Keyword search is good at exact terms. Many real users ask questions with both: “SOC2 retention policy”, “error E102”, or “pricing plan limits”.
Hybrid search combines both signals so the system can find semantically relevant content without losing exact-match precision.
When to use it
Use hybrid search when retrieval fails on names, IDs, acronyms, or specific product terminology. It is especially useful for technical docs and enterprise support.
Common mistakes
- Assuming vector search makes keyword search obsolete.
- Combining scores without testing real queries.
- Forgetting metadata filters.
Next decision
If hybrid search finds the right candidates but rank is still unstable, evaluate reranking.