Embedding
Also known as: vector embedding
What is Embedding?
A way of turning text, images or other data into a list of numbers that captures meaning. Content about similar things ends up with similar numbers, so a machine can measure how close two items are. Embeddings are the maths behind semantic search, recommendations and retrieval-augmented generation.
Why it matters
Almost every useful thing you can do with unstructured text or images now begins by turning it into embeddings, so the choice of embedding model shapes everything built on top of it. Feed a model trained on general web text your dense legal or medical language and the similarities it finds come out subtly wrong, in ways no error message ever flags. The numbers also go stale. As your catalogue, your jargon or your product range shifts, embeddings generated last year drift out of step with what people actually search for today. Teams that treat them as a fixed asset, rather than something to regenerate, pay for it in search results that slowly get worse.
In practice
A retailer embeds its product catalogue so a shopper searching for a “waterproof jacket” also surfaces a “rain shell” filed under another name. Because new products and seasonal language keep arriving, the team schedules regular re-embedding rather than treating it as a one-off. Keeping the numbers current is what keeps the search matching how customers actually describe things.
