Why matching products across stores is the hard part of price comparison

Reading a competitor's prices is the easy half of price monitoring. The hard half is answering one question, thousands of times: is this listing the same product as mine? Get that wrong and every number built on it is wrong too.

Why titles are not enough

Two stores can call the same product different things: "Grip Shift" at one store and "Kraton Grip Shift Grips" at another. And they can call different products the same thing: two brands of "Tubeless Valve", or the same bike frame in two model years. Matching by title alone fails in both directions.

Identifiers are the strong signal

When two stores share an identifier, matching is close to certain:

• SKUs and barcodes that appear on both sides.
• Catalog IDs. Many independent bike shops, for instance, run on a platform that shares one product catalog between shops, and each product page carries the catalog's ID. In a test across two bike shops with about 12,000 and 16,000 products, about 9,600 products matched exactly by that ID. Where the names differed, the ID still said they were the same product.

Even identifiers need care. A shop's own SKUs are often reused across variants, so a shared SKU on its own is not proof. Catalog IDs, which name exactly one product, are stronger than a shop's own SKUs.

Variants and packs

The same product can come in sizes, colors and pack counts, and a listing's displayed price is often the lowest variant. In the same bike-shop test, about 124 products that shared an ID were left uncompared on purpose, because one shop listed a single piece at a few cents and the other a bulk box at tens of dollars under the same ID. Showing that as an undercut would be false. It is better to leave a product out than to show a wrong gap.

Brand as a second opinion

If two titles match but the brands clearly differ, they are probably different products. If the brands agree, or one is simply a longer form of the other ("Nike" and "Nike Inc"), that adds confidence. Treating a known brand mismatch as a veto removes a whole class of false matches.

Approval first

Even with all of this, some matches are uncertain. Tensorbasis keeps ambiguous pairs in a review queue, shows the evidence (shared SKU, same title, similar wording) and lets you approve or reject. Only matches you can trust feed your reports. A pricing tool that quietly guesses is worse than one that asks.

What to check in any tool

• Does it say how each pair was matched?
• Can you correct a wrong match, and does it remember?
• How does it treat packs and variants?
• Does it show prices you can verify on the competitor's page?

If you want to see matching on your own catalog, the free price checker compares your store with a competitor in about a minute.