Vector search
In short
Vector search is a method of finding the items whose embeddings sit closest to a query embedding. An embedding is a list of numbers produced by a model so that similar things end up near each other.
Vector search is the retrieval technique behind semantic search and many AI features. The vector database is the storage system; vector search is the act of querying it.
How vector search works
- Convert: The query is turned into a vector with the same model used for the data.
- Measure distance: The system compares it with stored vectors using a measure such as cosine similarity.
- Approximate for speed: Large systems use approximate nearest neighbor methods that trade a little accuracy for much faster results.
- Rank and filter: The closest results are returned, often narrowed by fields like country or industry.
Vector search vs. semantic search
Semantic search is the goal: finding results by meaning. Vector search is the main technique for achieving it. Vector search can also match images, products, or companies, not only text.
GTM examples
- Similar accounts: Take your top customers' vectors and find the nearest companies.
- Prospect discovery: Search a company database using a plain-language description.
- Retrieval for AI: Fetch relevant notes or documents to ground a generated answer.
A practical example
Say a team wants more companies like their five best customers. They average those companies' vectors, run a search for the nearest matches, and filter by region and headcount. The team then reviews the top results by hand before adding any to a campaign.
Common mistakes
- Skipping hybrid search: Combine with keyword search when exact names or terms matter.
- Trusting the top result blindly: Similarity is not the same as fit. Check against your ideal customer profile.
- Mixing embedding models: Query and data vectors must come from the same model.
Vector search FAQs
What is vector search?
Vector search finds items by comparing numeric embeddings and returning those closest to the query embedding. It matches on meaning or similarity rather than exact words, and is used for semantic search, recommendations, and retrieval for AI systems.
How is vector search different from keyword search?
Keyword search looks for the exact words in a query. Vector search compares embeddings to find items with similar meaning, even if they share no words. Many products combine both to get exact matches and conceptual ones.
What is the difference between vector search and a vector database?
Vector search is the process of finding nearest embeddings for a query. A vector database is the system that stores the embeddings and runs that search efficiently, usually with indexing and metadata filtering.
What is vector search used for in B2B sales?
Teams use it to find companies similar to their best customers, search company databases with plain-language descriptions, and retrieve relevant notes or documents to feed AI-written account briefs and emails.
