Articles — Product data for shops
Product data for shops
The assistant names a product, a shop, or a place
Before an assistant can recommend you, it has to decide which mentions across the internet refer to the same thing. When your own information disagrees, you get skipped or merged into someone else.
| What has to match | |
|---|---|
| Product | The same title, brand, price, and stock for one item |
| Shop | The same seller name and address, and the same product in that shop |
| Place | The same name, the same pin or address, and the same hours |
One watch, three disagreeing lines
Illustrative example, not a logged test. The Northloop Trail Watch is a fictional product at €300.
| Where it appears | What it says |
|---|---|
| Website | Northloop Trail Watch, €300 |
| Shop listing | Northloop Trail Watch |
| Feed or another page | Trail Watch Black, €249 |
A person can tell that these lines describe one product. A machine has much less to go on, so it looks for the same name, the same brand, the same ID, the same URL, and whether trusted sites treat all of it as one thing.
When those facts agree, the assistant has a single product it can recommend. When they do not, you end up as two or three weak entries, or you get folded into a bigger rival’s listing.
That matching job, which ties a name to one product, shop, or place, is a knowledge graph. You do not need to build one. You need to stop publishing three different stories about the same thing.
A keyword is a text match (“trail watch”). An entity is one real thing: this watch, this hotel, this bottle at this shop.
How it shows up in real asks
“Local walking tours, not the tourist loop.”
The tourism site calls it Old Town Loop, reviews call it OT walking tour, and Maps has a different pin, so the assistant falls back to famous landmarks.
“Supplements that fit me — shops in my city.”
“Creatine” is a keyword, while Brand X, 300g, at Shop Y is a thing. When the names disagree, your local listing loses out to a big marketplace one.
“These seven to ten bottles at this price — worth it?”
Nothing can judge the discount until it knows which bottle and which shop, so a fuzzy identity means it compares the wrong wine.
Brand identity vs this job
Brand identity is about how you want people to feel. This job is duller than that: publish the same name, the same location, and the same basic facts everywhere, so machines can tell you apart from something with a similar name.
One line to check (one product, place, or tour)
Write one line:
Name + where it lives (URL or address) + one boring fact (price, availability, or what the offer includes)
Then check that line against the product, shop, and place rows above, wherever they appear.
That is entity work, and you do not need a vendor tool to do it. Write one dated note of the name, price, and hours you stand behind, then hold the site and the listings against it — the same habit as in Wrong price, wrong shop, fake discount.
Keywords get you into the conversation. Publishing the same details everywhere is what gets you recommended, and what stops a check of one offer from judging the wrong bottle.
Read next if: Your product lives in three places · Structured data publishes facts machines can read