Synonyms

Closing the distance between the customers' words and the catalogue's, by hand or through the suggestions that come out of searches with no results.

Customers type "trainers" and the catalogue says "sneakers". Nobody wrote anything wrong, and the search returns zero. Synonyms are where that distance is closed.

It lives in Search → Synonyms, and it's a plan feature: without it, the screen doesn't appear in the navigation.

Creating by hand

Pick the engine, type the term and the list of synonyms. A synonym written by hand takes effect straight away, with no reindexing and no waiting.

The rule works in one direction only, and that's on purpose. The term is what the customer types; the list is what the engine goes looking for in the catalogue in its place. With "coat → jacket, parka", whoever types "coat" now sees all three; whoever types "parka" still sees only parkas.

Other engines treat synonyms as groups of equivalent words, all standing in for one another. Not here, and it's better this way: in that model, "parka" would start pulling in the whole coat collection, and a narrow term would contaminate every broad search in the store. Damage like that shows up in no report at all, because the search doesn't fail, it just returns worse.

Write the reverse pair when the two words really are the same thing with the same breadth, and both show up in the searches: "trainers → sneakers" and also "sneakers → trainers". Outside that case, one direction is what you want.

If a synonym is itself the term of another rule, the two join up on their own on the way to the engine. "lovesense → lovense" and "lovense → tech vibrators" written separately work as if they had been written together, with no need to repeat the list.

Two more pieces of advice that save work:

• Write the way the customer writes, not the way marketing writes. "Jumper", "sweatshirt", "hoodie": all three, if all three show up in the searches.
• Don't use synonyms for typos. That's already handled, on its own, and a list full of typos becomes impossible to maintain. See how correction works.

The suggestions

A model analyses the searches that returned zero results and proposes lists of synonyms. The suggestions appear marked AI, in violet, and sit waiting for approval at the top of the list: nothing reaches the engine without someone clicking Approve. The gold badge in the navigation says how many are waiting for a decision.

It's the more valuable half of the screen, because it starts from the real problem: these aren't synonyms somebody imagined at a desk, they're the words your customers typed and didn't find.

Before you approve Read the list with the question "does whoever types this really want that?" in mind. A wrong synonym is worse than no synonym at all: it starts returning results to people who were looking for something else, and that never shows up in any report as a problem.

Where you see the effect

In the playground: besides the results, it shows which synonyms were applied to the typed term. It's the way to confirm that the new list really is acting, and to understand why an unexpected result showed up.

Stores in more than one language

Synonyms are per engine, like almost everything in search. "Trainers/sneakers" in the Portuguese engine doesn't apply in the French one, and it's just as well: foreign words that enter one language's vocabulary don't enter the others at the same speed.