Search in e-commerce once began very simply: the search engine looked only at words that a Web site visitor typed in. This is called keyword search (word search). For example, if someone typed “red coat,” you only got results with exactly those words: “red” and “coat.” So the system didn't understand why someone was doing that search, or what they really meant.

This works fine if you know exactly what you are looking for, but it is limited. In fact, website visitors use all sorts of words, synonyms, or make mistakes. Consider:

• “rainjacket” instead of “raincoat”

• “white sneakers” instead of “sportshoes”

• “outfit for a wedding” instead of. “party dress”

What does Vector Search do differently?

Vector Search uses AI to understand what someone means, not just what it says literally. Instead of searching for exact words, the system converts words into vectors - these are strings of numbers that represent meaning in an imaginary space.

Products and searches then become points in that space. And the closer two points are to each other, the more similar their content is. So you are not searching by words, but by meaningful distance.

Using Vector Search requires the search algorithm weighting .

Vector Search is part of the Smart Search solution and therefore depends on your contract whether you can use it.
Are you interested in Vector Search? Please contact your Customer Success Manager.

Example:

• The search phrase “summer dress” is close to “light dress” and “beach dress.”

• But further away from “winter coat” or “jeans”

At what place in the search algorithm do you set the Vector Search algorithm?

The recommendation is to set the Vector Search after the first Word Search and in any case before the Autocorrect, because the search phrase will be distorted which in turn affects the operation of the Vector Search. You can only add one Vector Search in the search algorithm.

Through accuracy, you specify a threshold for a product to be found through Vector Search.

This is underground a calculation that works as follows for distance:

Higher accuracy and thus smaller distance to dimensions

Lower accuracy and thus greater distance between dimensions