Fitment data

Normalization & synonyms

Making “5x114.3” and “5x4.5” the same thing.

Real catalogs spell the same value five ways. Normalization is how "5x114.3", "5×114.3mm" and "5x4.5″" become one thing — and how "Chevy" finds "Chevrolet".

Two kinds of rules#

Physics ships in code. Unit equivalences are not opinions: 5×114.3mm is 5×4.5″, 1 inch is 25.4mm, and no shop should have to configure that. Unit conversion, decimal tolerance and measurement parsing are baked in and identical for every shop.

Vocabulary is yours. Make and model aliases, brand spellings, finish and compound names, and product-type synonyms are per-shop, stored as a list you can read and edit. Nothing about your catalog is decided by a black box you cannot inspect.

The make and model rules are the ones that matter to every store, whatever you sell — they are what turn "chevy silverado" into a vehicle.

What's canonicalised automatically#

Makes and models: case, spacing, hyphens and punctuation, on both sides of every comparison. F150, F 150 and f-150 reach the same vehicle.
Bolt pattern (wheels): separators (x, ×, -, /), units, and imperial/metric equivalence. Output looks like 5x114.3.
Lengths: inches as decimals; 20", 20, 20.00 collapse to one value, whether that's a rim diameter, a lift height or a blade length.
Offset (wheels): millimetres as signed integers; +35, 35mm, ET35 collapse to 35.
Hub bore (wheels): millimetres as decimals.
Tire size: parsed into width, aspect and rim so 275/55R20 and 275/55/20 match.
Whitespace and case: everywhere, on both sides of every comparison.

Synonyms you control#

Under Search & filters → Normalization, each rule maps one or more inputs to one canonical output. Defaults ship for the obvious ones:

text

chevy, chevrolet trucks   →  Chevrolet
vw, volkswagon            →  Volkswagen
mercedes, merc            →  Mercedes-Benz
f150, f 150               →  F-150

Add your own for the spellings your suppliers actually use. Rules apply at ingest for sources and at query time for shopper searches, so a customer typing "chevy silverado" resolves the same way a tag does.

Finding what needs a rule#

The Sources screen lists every value that failed to resolve, with a count. That list is the work queue — the top entry is usually one synonym away from fixing thousands of rows.

Fix the rule, not the file

Editing 4,000 CSV rows because a supplier writes F150 is a waste of an afternoon that recurs monthly. One synonym fixes it permanently, for every source, including the next supplier who does the same thing.

Ranking is separate#

Normalization decides whether two values are the same. It does not decide which product comes first — that is the ranking tuner.

Still stuck?

Support is answered by the people who wrote this code — not a ticket queue.

Contact support