Eight ways to write the same scale, and the listings used one

Scale is the first thing a model builder filters by — it decides whether a kit fits the shelf, the display base and the collection. It is also written inconsistently by every person who types it. The same 1/48 aircraft is searched as “1:48”, “1-48”, “48 scale”, “48th scale”, “quarter scale”, and in the German store almost always with a colon where a US seller writes a slash.

The brand's titles used exactly one of those formats, chosen years ago for consistency. Amazon's search does not normalize them into each other, so the catalog was invisible to most of the ways its own defining attribute gets written — and the brand had concluded that its scales were simply less popular than they are.

We indexed the attribute in every form it is actually typed.

• Notation variant map per marketplace, built from search-term reports rather than assumption — the German and UK habits differ from the US ones.
• Primary notation in the title, variants in the backend, so titles stay readable while the index covers the whole set.
• Subject-plus-scale pairing, since almost nobody searches a scale alone — it is always attached to a vehicle type or subject.
• Campaigns per notation cluster, which immediately showed which variants carried real volume and which were noise.

Impressions on scale-bearing queries rose from 310k to 819k a month, and revenue grew 88% with no change to the kits, the pricing or the number of listings.

Two hundred and sixty-eight notation terms now rank in the top ten, against twenty-two. ACoS fell twelve points because the variant terms are barely contested — most competitors made the same tidy, consistent, costly decision about how to write a ratio.