Case study · Client · Fashion e-commerce · In build

A uniform brand is replacing photo shoots and a paid sizing plugin with its own AI studio.

A premium corporate uniform brand sells online to companies around the world. Every collection means photo shoots, and every order raises the same question: which size? We are building its own AI studio: product visuals on AI models for the next collection, and a size recommender to replace a paid plugin.

Company: Premium corporate uniform brand, 8 people
Function: E-commerce
Status: In build, first collection in November 2026
Catalogue: 169 products, 1,939 variants

67products covered by the size recommender

< 4 cmaverage error on body measurements, across seven tests

1,939product variants in the online catalogue

The size recommender is built and tested, not yet live on the store. The visual studio is in build.

The challenge

A small team, a large catalogue, a photo shoot for everything.

Eight people run a catalogue of 169 products and almost 2,000 variants. Each collection needs studio photos on models, which is slow and expensive. Size advice comes from a paid third-party plugin, and a wrong size means a return.

Behind the store, a small order could take around fourteen emails, and only one person really masters the order system.

What we built

The data first, then the images.

We started with an audit of how orders really flow: three process maps and 35 questions checked with the team. The catalogue audit found 108 variants without a stock reference, to fix before anything else.

The size recommender is the brand's own model, fitted on a public dataset of 6,068 adult body measurements and mapped to every product. The visual studio will generate product images on AI models, from the product photos and a set of approved models, for the next collection.

Audit
Map how orders, catalogue and stock really work, and fix the data first.
Size
An in-house recommender, fitted on 6,068 body measurements, for all 67 products.
Generate
Product images on AI models, from product photos and a set of approved models.
Publish
Images and size advice pushed to the online store.

Before and after.

Before How it works today

A photo shoot for every collection
A paid plugin for size advice
108 variants without a stock reference
Up to fourteen emails for a small order

Target What we are building

Product images generated on AI models, in house
Size advice from the brand's own model
A clean catalogue, every variant referenced
Orders mapped, and the next steps automated

The shift

The brand stops renting its visuals and its sizing from vendors. It owns them.

The business impact.

67products covered by the size recommender
2.0–3.7 cmaverage error on key body measurements, across seven tests
1paid plugin to be replaced by the brand's own model
Nov. 2026first collection planned with AI-generated visuals

Built with Shopify OpenAI image models Python Public body-measurement data

This case is in build. The size recommender is tested but not live yet, and the visual studio has not produced a collection. We will update this page with real results.