Module 09 — Model Y

Categorisation.
Solved.

Six thousand Google categories. Two-hundred million unique GTINs. Ninety-four percent accuracy. Better than humans. Faster than rules.

Wrong category, wrong customer. Wrong category, wrong supplier match. Wrong category, wrong everything.

The solution

One model. Six thousand categories.

Model Y assigns every product on the platform to the right place in the Google Product Taxonomy — with confidence scoring, parent-child resolution, and multilingual coverage across nine European regions.

Capabilities

What categorisation should do.

01

Google Product Taxonomy

The standard the largest commerce surface in the world uses. Six thousand leaf categories.

02

Confidence scoring

Each prediction comes with a probability. High-confidence assignments commit; low-confidence flag for review.

03

Parent-child fallback

When the leaf is uncertain, Model Y commits to the deepest confident parent — never wrong, never empty.

04

Multilingual

DE, FR, IT, EN, ES, PT, NL, PL, TR — same model, same accuracy, no per-language retraining.

05

Batch and real-time

Process millions in a nightly batch. Categorise a single new GTIN in under a second via the API.

How it works

Extract. Predict. Assign.

01 — EXTRACT

Read the product

Title, description, brand, supplier hints, retailer hints — every signal the platform already has.

02 — PREDICT

Score the taxonomy

Probability over six thousand leaves. Top candidate plus next two retained for audit.

03 — ASSIGN

Commit with confidence

High confidence commits to the leaf. Low confidence commits to the deepest confident parent.

Proof

Better than humans.

0%
Accuracy on validation set
0
Categories supported
0M+
Unique GTINs classified

Run Model Y.

Send us ten thousand products. We return them categorised, with confidence scores, in under a minute.

Talk to us