Multilingual store: what breaks between translation, addresses and search

Translating a store is a content project. Getting it found in the right language by the right buyer is an architecture project. Businesses almost always fund the first and discover the second six months later, when English-language traffic has not moved despite a complete translation.

On this page

The three breaks in a multilingual store

They do not happen at the same moment and they do not show up in the same place.

The first is the incomplete translation. A store launches its English version with the main pages and the catalogue, then life resumes: a new product, a collection page, an article. Those additions are born in the original language and the catch-up has no owner. Six months later the secondary version is a subset of the store, and the customer who lands on a hole rarely comes back. The cost is invisible because it is counted in sales that never happened.

The second is the malformed address. It is not visible at all from the store admin, where everything looks fine. It shows up in the data, as impressions in one language on the pages of another. The demand exists, the wrong page gets served, and management concludes the market does not respond, then cuts the budget on the wrong line.

The third is the faithful translation. It is the most expensive because it passes every quality check. The text is correct, the grammar is good, and it contains none of the words buyers in that market actually type. The invoice is paid, the risk is untouched.

The order of magnitude is worth stating, because it determines who has to decide. Translating a mid-sized catalogue and maintaining it costs about as much as hiring a part-time resource for a year. That is not a line in a marketing budget, it is an investment whose return is measured in sales in a market you have not yet validated. The decision therefore belongs to management, exactly like the decision to open the market.

These three breaks share one trait that explains why they go unnoticed: none of them appears in a sales report. An incomplete translation does not lower revenue, it prevents an increase. That is the hardest kind of loss to defend in a budget meeting, and the one businesses fund the longest without knowing it.

The addressing rule, and why it exists

Language is not country. That distinction, which sounds theoretical, produces a very concrete rule: on a secondary market, the address has to carry both language and country, of the /fr-ca or /en-eu type. A language-only subfolder, of the /fr type, exists only on your primary market.

The reason is mechanical. On a secondary market, the address does not only pick a language, it serves the right currency, the right prices and the right terms. An isolated /fr does not tell the system which French it is, therefore which market, therefore which prices and which currency.

A store already using language subfolders keeps its existing addresses. It is the markets added afterward that follow the combined rule, which often produces a hybrid structure, consistent but surprising to anyone discovering it without explanation.

On the engine side, three declaration rules apply and integrations regularly miss them. Every version has to declare itself. Every version has to declare all the others. And the declaration has to be reciprocal, failing which the whole set is ignored rather than partly kept. Since Google retired its international targeting report in 2022, no official tool flags those errors at site level.

Translating is not getting found

A buyer does not search for the translation of your word, they search for their own. Those two words coincide less often than you would think, and the gap is widest exactly where your margin is best, on technical products.

Three gaps recur. The trade term first: two English-speaking markets can name the same product differently, and a professional translator will pick the correct term rather than the searched one. Units and formats next, which change how a product gets searched for. The standards cited last: a buyer often searches by the standard they have to meet, and a Quebec standard means nothing to them.

The translation order follows directly from that, and it is not the one people adopt spontaneously. Rank your pages by revenue generated over twelve months, translate the first tenth, measure for a quarter. A business that translates its whole catalogue at once commits the entire expense before getting any answer from the market, and sometimes discovers that demand sits on three product lines.

The correction is not a retranslation. It is a survey of the words searched in the target market, done before the translation, and handed to the translator as a constraint. It costs a few days and it changes the return on the whole project.

Who keeps the translation current

This is the question that determines whether the project ages well, and it almost never has a written answer.

The break point is identifiable: any page created outside the flow of Shopify's Translate and Adapt app, or of whichever translation app you use, escapes the process. A product added by one team, a collection page built for a promotion, an article published by marketing. Nothing flags the absence, and the gap settles in one page at a time, without any sales report showing it.

The rule that works is to treat translation as a publishing step rather than as a project. A page is not live until its secondary version exists, or else it is explicitly accepted as unilingual, with a note for the visitor. Both choices can be defended, the absence of a choice produces a store full of holes.

Before launching the translated version

Market structure is covered in opening a market without duplicating your store, and pricing in what multi-currency settles and what it leaves you. Quebec language obligations are in the language obligations on your website, and the opening sequence in your export plan does not say where the customers will come from.

The architecture choice that comes before these settings is covered in the address architecture for two markets. The decision is taken before the first translation, because it sets the maintenance budget and the room to manoeuvre in the target market.

Frequently asked questions about multilingual stores