Making an SEO strategy deployable across 2,500 SKUs a season

Bellerose is a premium Belgian fashion brand selling in several languages. Its product content is produced in house by a small team, and until this year every step of that production was a separate tool with a separate review pass.

Industry: Fashion

Headquarters: Brussels, Belgium

Company size: 130-150

A person with shoulder-length hair is wearing a tiger print sweater and high-waisted black leather shorts, holding a black clutch with a gold accent, standing on sunlit stone steps.

Person standing on a city sidewalk in a fuzzy pale pink sweater, wide black trousers, and holding a small handbag in front of an ornate dark building.

Before: five tools, one person, one product at a time

Margaux Breda, E-commerce Merchandising and Content Specialist, started each collection at the collection presentation, taking notes product by product. With the products in front of her, she photographed each one herself, added the pictures into ChatGPT along with the specifications she pulled from the PLM, and prompted it product by product for a description and detail bullets, correcting and placing each draft by hand. The finished texts then went through a professional translation tool against a glossary she maintained herself, and came back for a line by line correction in every language.

SEO came last and cost the most.

I went product by product to make sure the color was properly described, and that the elements essential to the product were in the title.

Margaux Breda

E-commerce Merchandising and Content Specialist, Bellerose

Each meta description was the description's main sentence plus the same stock line on every product: "Discover the latest trends for women."

One person could get through around a hundred products a day. A women's collection runs to 600 styles. Split with an intern, that meant roughly three full days of production per person, taken one day a week over several weeks, before optimization could even begin. Optimization never began.

And the women's collection is only part of the catalog. With the men's collection at around 300 SKUs, kids at around 200, and some 1,400 external-brand products handled by a colleague, product content at Bellerose comes to around 2,500 SKUs a season, each needing a description, details, a meta title and a meta description in every language.

Now: creation, localization and optimization, automated as one system

Newtone automates the whole chain Bellerose used to run by hand. From the product data in the PLM, the imagery in the DAM and the notes taken during the collection presentation, it creates each description and product detail, localizes them directly into every language the brand sells in, and optimizes every meta title and meta description for how customers in each market actually search. Nothing passes through a translation tool: each language is created in its own right, with the brand's tone of voice, glossary and content rules encoded into the system rather than enforced by hand. Verification runs independently of production, which is why the review load fell rather than moved.

We will not remove the human review entirely, for safety and on principle. But it takes far less time. The grammar is correct, and on the vocabulary we used to correct much more than we correct with you. On SEO I do far fewer checks now, where before it was a title per product, one at a time.

Margaux Breda

E-commerce Merchandising and Content Specialist, Bellerose

What that makes possible

With production no longer the constraint, the team's time has shifted from producing pages to deciding what should be on them. Bellerose has begun a full multilingual catalog optimization program with Newtone and its SEO agency, run as a structured experiment: the first batch is the carry-over range, which gives a clean measurement base. The agency defines the rules, richer descriptions built on what is already live and the linking strategy from each PDP to the PLPs, and Newtone applies them across the catalog, so every future collection inherits the same standards automatically. Performance will be read season over season on those same products, across all of Bellerose's languages rather than the master language alone.

Asked whether Bellerose could deploy a content SEO strategy across its product catalog without a tool like Newtone, her answer was unambiguous:

Almost impossible, I think. I had tried before. When you put your files into Copilot or ChatGPT, they do not handle them properly. They make mistakes. The result is nothing like it should be.

Margaux Breda

E-commerce Merchandising and Content Specialist, Bellerose

Margaux had been using Copilot and ChatGPT for exactly this before Newtone, and the results were highly inconsistent. What she produces with Newtone today, full collections created, localized and SEO-ready directly from the brand's own product data and imagery, was not achievable with either.

With Newtone, the question at Bellerose has changed scale. It is no longer whether 2,500 SKUs can be produced in every language each season; it is how much organic traffic a fully optimized multilingual catalog can deliver. The coming seasons will answer it.

AI-powered content creation, translation & optimization at enterprise scale.

Product

AI content creation suite

Industries

Use cases