Shoesy.AI

Sizing setup

Sizing is the part of the assistant that handles the highest-stakes shopper question — do these run small? — and the part most likely to get returned items if it gets it wrong. Shoesy AI ships with a curated cross-brand size database and reads your catalog's own variants; there is nothing to configure to turn it on.

Pre-loaded brands

Shoesy AI ships with charted size data for three dozen footwear brands, and recognises several hundred more by name. Confidence is at least 0.6 on every mapping before the assistant will recommend a size; below that it falls back to a hard "we don't have enough data" copy rather than guessing.

The brands with full charted size data are:

Adidas, Allbirds, Altra, ASICS, Birkenstock, Brooks, Clarks, Columbia, Converse, Crocs, Dr. Martens, ECCO, Geox, Hoka, Hush Puppies, K-Swiss, KEEN, Merrell, Mizuno, Native Shoes, New Balance, Nike, On, Onitsuka Tiger, Palladium, Puma, Reebok, Salomon, Saucony, Skechers, The North Face, Timberland, TOMS, UGG, Under Armour, and Vans.

Beyond these, the assistant recognises several hundred additional footwear brands by name. A recognised brand with no charted data resolves through the engine's size-system grid at a conservative true-to-size default, with the assumption disclosed in the reply — still held to the same 0.6 confidence floor.

Each charted brand carries a size-conversion chart (US / UK / EU number mapped to a foot length) plus a fit profile — how the brand tends to run, with model-level exceptions where a specific silhouette runs long or short. If your catalog mixes brands with different fit conventions (e.g. a "true-to-size" Nike vs. a "runs small" Asics), the cross-brand size conversion is handled deterministically — the assistant never asks an LLM to compute a size.

How sizing works at runtime

A sizing question routes to a deterministic function call, not an LLM completion. Three things determine the answer:

  1. Your catalog — the variant axes Shopify reports for the product (US/UK/EU number, width letter).
  2. The cross-brand database — known fit-vs-true offsets per brand and per model family.
  3. The shopper's reference — if they say "I'm a 10 in Nike Pegasus," that becomes the anchor point for the recommendation.

If the confidence score lands below 0.6, the assistant returns a fallback message instead of a number — linking to your store's size guide when you have configured one (Knowledge → Assistant behavior → size-guide URL, a Growth feature); with no size-guide URL the fallback simply omits the link. This is intentional — bad sizing recommendations cost more than missing ones.

There is no per-store sizing-rule editor (yet)

Sizing is fully automatic in the current version: it runs off the built-in brand database plus your catalog's variants. There is no dashboard form for adding your own size offsets — so if you have read elsewhere about entering a "runs small / runs large" rule, that surface does not ship today. (The dashboard's Custom rules tab is unrelated to sizing — it controls banned phrases and free-form assistant instructions; see the Settings tour.)

The single most effective thing you can do for sizing accuracy is keep your catalog clean:

What the assistant will not do

If a sizing answer feels wrong, email support@shoesy.ai with the transcript — the cross-merchant fit graph we are building grows directly from that feedback, and it is how brands and models get added to the charted set.