In short
- The first problem with a second country is not translation, it is internal competition: two versions of the same catalogue, and with the language tags wrong the engine picks for you. Google is explicit here: a localised version only counts as a duplicate if its main content has not been translated.
- hreflang is exactly where this is decided, and exactly where almost everyone gets it wrong: it has to be reciprocal, complete, and point to pages that actually exist.
- Translation always stops too early: product pages are done, but order emails, the confirmation page, product safety notices and reviews are all still in the original language.
- A price is not a currency conversion: it is a decision per market, VAT regime included, and it has to stay consistent with your other channels there.
Opening a second market is the moment a shop discovers how many decisions the first market had quietly made on its own: language, currency, VAT, carrier, vocabulary. All of it was implicit; all of it becomes a choice. Here are the failures in the order we meet them, because the order is what checklists leave out.
Failure 1: your two versions compete against each other in search
The moment the second language goes live, every product exists twice, in two very similar forms. With no clear signal, the search engine picks which one to show, and it picks badly: the version built for your first market surfaces in a country that speaks the other language, or the second-market version cannibalises the first. Two genuinely translated versions are not duplicate content for all that: Google only treats a localised version as a duplicate when its main content has not been translated.
The signal is called hreflang, and its reputation as a headache is deserved for three specific reasons. It has to be reciprocal: if the first-market page declares its twin, the twin has to declare it back, otherwise the whole set is ignored. It has to be complete: every version must list itself as well as the others. Google will tolerate omitting some languages on some pages and will still process the pairs that point to each other, but it ignores the tags outright as soon as two pages fail to point to one another. And it has to point to real pages: an untranslated product that declares a twin which does not exist sends crawlers straight into errors. We have lived through the sneakiest variant on our own site: internal links that crossed languages through redirects, invisible to the eye, counted by the crawlers.
The mistake that costs the most Translating 60% of the catalogue and letting the remaining 40% show the original language on the new market’s URLs. The visitor moves from a page in one language to a product page in another without warning, the search engine receives contradictory signals, and you lose the trust of both. A small range translated in full, checkout and emails included, beats a complete catalogue translated halfway.
Failure 2: the translation stops at the checkout
Product pages are translated, the homepage too, and then the customer clicks “order”: the checkout switches back to the original language, so do the confirmation emails, so does the basket reminder. This is the most common failure precisely because it is invisible from the back office: those texts live elsewhere, in modules, in email templates, in theme strings.
The real checklist for a language does not stop at the catalogue, then: checkout, transactional emails, confirmation page, terms and conditions, returns, and the mandatory notices. Since 13 December 2024, the EU General Product Safety Regulation has required safety warnings and information to be written “in a language which can be easily understood by consumers, as determined by the Member State” where the product is made available, and the obligation applies to the online offer itself. We covered this in our piece on product safety: the second country reopens that obligation across the whole catalogue.
Failure 3: prices converted instead of prices decided
Converting prices at today’s exchange rate gives you absurd decimals on the shelf, margins that drift with the currency, and prices that contradict your other channels in the same market. A price per country is a decision: a sensible psychological rounding, consistency with what your resellers or your marketplaces already charge there, real local delivery costs. This is exactly the backwards pricing problem we describe in the marketplace article, and it applies to your own site every bit as much as to marketplaces.
Before deciding which rate to apply, you have to know which one is due. Since 1 July 2021, Article 59c of the VAT Directive has set a single threshold of EUR 10,000 excluding VAT: as long as the annual total of your intra-Community distance sales and the services concerned stays below that figure, and you are established in a single Member State, the VAT due remains that of your country of establishment. Two traps. The threshold is aggregate: it adds up sales to all other Member States combined, it is not a per-country threshold, and a second market that starts well will break it quickly. And it is assessed on the current calendar year and the preceding one: having exceeded it last year is enough to make this year’s sales taxable at destination. Above the threshold, VAT is that of the country of arrival, rate by rate, without having to register everywhere: the One Stop Shop lets you declare and pay in a single Member State. You can also opt for taxation at destination before reaching the threshold, for a minimum of two calendar years. The shelf price follows from that decision, never the other way round. Outside the Union a second decision is added to this one and weighs just as much on the shelf price: who pays the customs duty, you at checkout or the customer on delivery.
Failure 4: the vocabulary your translator delivered
A correct translation is not a market vocabulary. Customers in the new country are not searching for a literal translation of your existing category names; they have their own words, their own reference brands, their own units. That vocabulary sits in the same place as it did for the first market: real queries, including those from your on-site search once the traffic starts arriving. Translating product pages is the beginning; adopting the market’s vocabulary is what gets them found.
The method, in order
- Decide the range before the language: which products, which carriers, which returns terms, which payment methods the new market expects. A country opens with an impeccable sub-catalogue, not with the whole warehouse.
- Set the URL structure and reciprocal hreflang on that range, and check them the way a crawler would, not by eye: reciprocity, self-declaration, zero pointers to a missing or redirected page.
- Translate all the way to the end of the journey: checkout, emails, confirmation, legal and safety notices. The test is a real order placed in the new language, reading every single email it produces.
- Decide prices per market, VAT and rounding included, consistent with your other channels on the ground.
- Keep checking: language leaks come back with every piece of content you add. An automated check on links that cross languages is the only guardrail that holds over time, and it is the one we run on our own sites.
A fifteen-minute self-diagnosis, if your second market already exists: place a test order in the second language and count the screens and emails that revert to the original one. Then search for three of your products from the target country and note which language version comes up. Those two counts tell you the real state of the job.
The approaches, and what each is worth
| Approach | What you get | Who it suits |
|---|---|---|
| A small range, translated in full | A market you can test in real conditions, with no translation debt. Widens afterwards, product by product | Any first opening |
| Raw machine translation of the whole catalogue | Fast and indexable, but the vocabulary is off-market and the errors show: it converts badly and ranks badly. Acceptable as a transition, never as a destination | Large catalogues, with prioritised proofreading behind it |
| A subdomain or a domain per country | Justified when catalogues, prices and brands genuinely diverge; otherwise you buy complexity and diluted authority for nothing | Organisations built around subsidiaries |
| Opening the country through a marketplace first | Tests demand without building the site: logistics and customer service in the language still have to be handled in full, but the commercial verdict comes quickly | Uncertain markets, comparable products |
The tooling
On PrestaShop, we deploy the module published by Datafirefly Limited, our agency’s sister company, on client sites: Hreflang generates reciprocal, complete alternate tags across a multilingual, multi-store catalogue, which removes the most mechanical and most frequently wrong part of the job. One-off purchase · 12 months of updates.
Method beats tooling: the module guarantees correct tags on the pages that exist. The range, end-to-end translation, prices per market and vocabulary remain decisions, and they are what turns a second country into a market rather than a version.
The bigger picture
Selling abroad online is sold as a tick box, “just add a language”, when it is really a miniature replay of everything you learned on the first market: compliance, pricing, vocabulary, measurement. The shops that get their second country right are the ones that treat it as a launch, with a defined range, criteria and a verdict, not as a translation job.
That is the logic behind our e-commerce and marketplaces expertise: spreading your channels and your markets protects you from dependency, provided each opening is held end to end. Half a market costs nearly as much as a whole one, and returns almost nothing.
Sources
The verifiable claims in this article link to their primary source, accessed on 11 August 2026. We do not cite a source we have not read.
- Google Search Central, Tell Google about localized versions of your page. Open
- Google Search Central, Managing multi-regional and multilingual sites. Open
- Google Search Central, Spam policies for Google web search. Open
- Council Directive 2006/112/EC on the common system of value added tax, consolidated version, Articles 33 and 59c. Open
- Council Decision (EU) 2020/1109 of 20 July 2020 amending Directives (EU) 2017/2455 and (EU) 2019/1995 as regards the dates of transposition and application. Open
- French tax administration, official doctrine BOI-TVA-CHAMP-20-20-30, rules applicable to distance sales of goods in e-commerce. Open
- impots.gouv.fr, using the VAT One Stop Shop (IOSS-OSS). Open
- European Commission, Register to OSS, VAT e-Commerce One Stop Shop. Open
- Regulation (EU) 2023/988 of the European Parliament and of the Council on general product safety. Open
FAQ
Subdirectory, subdomain or a new domain?
A subdirectory concentrates the authority you already have and is enough in the vast majority of cases. A domain or subdomain per country is justified when brands, catalogues or legal entities genuinely differ. The criterion is organisational, not magical: no format “ranks better” in itself. One nuance from Google, though: a country-code domain (.de, .fr) is in itself a strong country-targeting signal, whereas a subdomain or a subdirectory gives none and has to be backed by hreflang or explicit targeting.
Is machine translation penalised?
Not as such: what gets penalised, by humans and by algorithms alike, is text that does not answer the market, wrong vocabulary and visible errors included. Machine translation that has been proofread and adapted to local vocabulary is perfectly viable; left raw, it hits a ceiling.
Same language, two countries: do you need two versions?
If prices, VAT, delivery or vocabulary differ, yes, with regionalised hreflang so the two do not compete. If everything is identical, a single version targeting the language is enough and can be refined later.
How many products do you need to open?
Enough for the range to be credible in its category, few enough to hold the translation end to end: often the best-selling quarter of the catalogue. The commercial verdict from a clean sub-catalogue is more reliable than the one from a complete catalogue left half finished.
Can Dotsland help?
Yes. Scoping the range, structure and hreflang verified the way a crawler sees them, an end-to-end check of the journey in the language, prices per market and continuous monitoring of leaks. It is one of the jobs within our e-commerce and marketplaces expertise. Let’s talk, or start with that test order in the second language.