Vector database
In short
A vector database is a database built to store and search vectors, which are lists of numbers that represent the meaning of text, images, or other data. Given a query vector, it returns the stored items that are most similar.
Vector databases power semantic search, recommendations, and retrieval-augmented generation. They exist because ordinary databases are built for exact matches, not for finding the closest meaning among millions of records.
How it works
- Embed: A model converts each record, such as a company description, into a vector.
- Store: The vectors are saved along with an ID and any metadata like industry or country.
- Index: The database builds an index so it can search without comparing against every vector.
- Query: A question is embedded the same way, and the database returns the nearest vectors, optionally filtered by metadata.
GTM use cases
- Company similarity: Find accounts that resemble your best customers.
- Semantic company search: Match plain-language descriptions to company profiles.
- Grounding AI answers: Supply relevant facts to a language model in a RAG setup.
- Deduplication and matching: Spot records that describe the same entity in different words.
A practical example
Say a team embeds descriptions of ten million companies. A rep describes their target as "software for independent clinics". The database returns the closest company vectors, and a metadata filter narrows them to a chosen country and size range.
Vector database vs. traditional database
A traditional database answers "which rows have this exact value". A vector database answers "which items are closest in meaning to this". Many teams add vector support to an existing database instead of running a separate one.
Common mistakes
- Adopting one too early: A small dataset may not need a dedicated system.
- Ignoring filters: Similarity alone returns off-target results; combine it with metadata filters.
- Changing the embedding model without re-indexing: Vectors from different models are not comparable.
Vector database FAQs
What is a vector database?
A vector database stores embeddings, which are numeric representations of meaning, and retrieves the items most similar to a query vector. It is used for semantic search, recommendations, and giving AI models relevant context.
How is a vector database different from a regular database?
A regular database finds exact matches on values or ranges. A vector database finds the nearest items by similarity of meaning. Many relational and search databases now offer vector features, so the line is not always sharp.
Why do AI applications use vector databases?
AI applications need to quickly find relevant content among large collections. Vector databases do this by similarity, which supplies language models with useful context in retrieval-augmented generation and powers meaning-based search over company or document data.
Do you need a vector database for semantic search?
Not always. Small datasets can be searched by comparing vectors directly. Vector databases and indexes become useful as the collection grows into the millions and you need fast results with filters.
