Hybrid Search

ELI5 - The Vibe Check

Vector search is great at understanding meaning but can miss an exact product code or acronym because it's thinking in vibes, not letters. Keyword search is great at exact matches but has no idea that 'car' and 'automobile' mean the same thing. Hybrid search runs both at once and blends the results, so you get the best of the literal-minded librarian and the big-picture-thinking friend.

Real Talk

Hybrid search combines sparse lexical retrieval (typically BM25 or full-text search) with dense vector similarity search, merging their result sets via a fusion method such as reciprocal rank fusion or a weighted score combination. It compensates for the weaknesses of pure vector search on exact matches, rare terms, and identifiers, while keeping the semantic recall that keyword search alone lacks.

When You'll Hear This

"Pure vector search kept missing SKU numbers, so we added hybrid search with BM25." / "Hybrid search improved recall on our support docs by a wide margin."