Ecommerce Localization: What to Translate First (and What to Skip)

Every guide to ecommerce localization says the same thing: translate your store, adapt your currency, respect local culture. Useful, in the way that "spend less than you earn" is useful. What nobody tells you is the order. You have a catalog of 800 products, a theme full of hardcoded strings, six email templates, and a budget that covers maybe a fifth of it. Translate in the wrong order and you'll spend the whole budget on pages nobody reads in a market that was never going to convert.

What Is Ecommerce Localization?

Ecommerce localization is the process of adapting an online store — product content, currency, payment methods, shipping, and legal copy — so it reads and functions as though it were built for a specific market. Translation handles the words. Localization handles everything that makes a shopper trust the store enough to enter their card details.

The distinction isn't academic. A German shopper who lands on a fluently translated product page, then hits a checkout that quotes dollars, offers no SEPA or invoice payment, and shows a shipping estimate in inches, will leave. The translation was perfect. The localization failed.

Localization vs. Translation: The Difference That Costs You Money

Translation is a subset of localization, and it's the cheapest, most automatable part. The expensive parts are the ones that touch your operations: payment rails, returns policy, tax display, size charts, and shipping promises. Those can't be handled by a translation app because they aren't language problems — they're configuration problems, mostly living in Shopify Markets rather than in your content.

This matters for sequencing. Translation scales linearly with word count and can be done in bulk. Localization decisions are one-off, per-market, and usually need a human who understands the market. Doing the cheap scalable part first, on the wrong markets, is the most common way stores waste money here.

What to Localize First: The Revenue-Ranked Order

Rank content by revenue impact per thousand words, not by how much of it there is. A store's product catalog is 90% of its word count and rarely the first thing that should be translated. The pages a shopper must pass through to buy come first, and there are surprisingly few of them.

The pattern: everything on the critical path to a completed order is worth localizing on day one, even though it's a rounding error in word count. Everything that's merely large — the catalog tail, the blog archive — waits until a market has shown it converts.

The Parts Everyone Forgets

The content that breaks a localized store is almost never the product descriptions. It's the strings that live outside the places translation apps look by default:

• **Theme strings** — "Add to cart", "Sold out", "Estimated delivery". These live in your theme's locale files, not your product data, and a half-translated theme reads worse than an untranslated one
• **Metafields** — size guides, care instructions, spec tables and anything an app writes into custom fields; these render on the product page but are invisible to a lot of translation tooling
• **Transactional emails** — the order confirmation is often the first thing a customer reads after paying, and it's frequently the last thing anyone translates
• **Image text** — banners and lifestyle shots with baked-in English copy will survive every translation pass you run
• **PDF assets** — size guides, care instructions, warranty terms and product manuals frequently live as linked PDFs that no translation pass ever touches; you have to extract the text out of them before any of it can be translated
• **Alt text and meta descriptions** — untranslated, they quietly cap your international SEO ceiling
• **Search synonyms** — a German shopper searching "Turnschuhe" won't match products tagged only "sneakers"

Cross-Border Ecommerce: What Changes Beyond Language

Cross-border ecommerce is selling to customers outside your home market, and language is only the visible layer of it. The layers underneath decide whether the order completes.

• **Payment methods** — cards dominate in the US and UK; iDEAL in the Netherlands, SEPA and invoice in Germany, Klarna across the Nordics. Missing the local default is a silent conversion tax
• **Duties and taxes** — showing a price that excludes VAT to a European shopper looks like a bait-and-switch at checkout; prices should display tax-inclusive where that's the local norm
• **Currency and price formatting** — not just the symbol but the separators and placement; see the multi-currency setup guide for the Shopify-specific mechanics
• **Shipping expectations** — a 14-day estimate that's normal in one market reads as abandonment in another
• **Returns** — EU consumers have a statutory withdrawal right; a policy written for US law is both wrong and a compliance risk

Shopify handles most of this in Markets configuration rather than in content, which is genuinely good news: it's a one-time setup per market, not a per-word cost.

How Much of This Can AI Handle?

Most of the word count, and almost none of the judgment. Modern AI translation is strong on product descriptions, collection copy, and blog content — the high-volume, low-risk, tolerant-of-imperfection material that makes up the bulk of a catalog. It's weaker exactly where the stakes are highest: legal copy, brand voice in marketing headlines, and anything where a subtle mistranslation costs you money or exposes you to a claim.

The practical split most stores land on is bulk-translate the catalog with AI, then human-review the ~5% of pages that carry the revenue or the risk. That's the post-editing model, and it's usually an order of magnitude cheaper than full human translation without being meaningfully worse where it counts. One prerequisite: build a glossary before you bulk-translate anything, or your brand and product names will come out different on every page.

A 30-Day Localization Sequence

If you're starting from an English-only store and one target market, this is a realistic order of operations:

1. **Days 1–3.** Pick one market, not four. Use existing analytics — where is traffic already arriving from and failing to convert? That's your market, and it's a much better signal than market-size reports
2. **Days 4–7.** Configure Shopify Markets for that region: currency, pricing, payment methods, shipping zones, tax display. No translation yet
3. **Days 8–10.** Build the glossary. Brand names, product lines, terms that must never be translated, and terms that must always be translated the same way
4. **Days 11–14.** Translate the critical path: theme strings, checkout, cart, navigation, policy pages. Human-review this tier
5. **Days 15–20.** Bulk AI-translate the catalog. Human-review your top 20 products by revenue and nothing else, for now
6. **Days 21–24.** Add hreflang and a visible language switcher. Without these, the work you just did is hard for Google and shoppers to find
7. **Days 25–30.** Translate transactional emails, then measure. Conversion rate by market is the only number that decides whether market two happens

How Do You Know It Worked?

Compare conversion rate for the localized market against your home market, not against its own past. Traffic will move first and mean very little; sessions can double while revenue stays flat if the checkout still feels foreign. If conversion in the new market reaches roughly two-thirds of home-market conversion within a quarter, the localization is working and the catalog tail is worth translating. If it stalls well below that, the problem is almost never translation quality — it's payment methods, shipping cost, or delivery time.

FAQ

What is an example of ecommerce localization?

A US store entering Germany translates its catalog into German, switches pricing to euros with tax-inclusive display, adds SEPA and invoice payment, rewrites the returns policy to reflect the EU withdrawal right, and converts size charts to EU sizing. The translation is one of five changes, and not the one most likely to decide the sale.

What is the difference between translation and localization?

Translation converts text from one language to another. Localization adapts the entire experience — currency, payment, units, imagery, legal terms, and formatting — so it fits the target market's expectations. Every localization project includes translation; not every translation project counts as localization.

Does localization actually increase ecommerce sales?

The consistent finding from CSA Research's long-running "Can't Read, Won't Buy" surveys is that a large majority of consumers prefer buying in their own language, and a substantial share won't buy from English-only sites at all. The effect is strongest in markets with lower English proficiency, and weakest in places like the Nordics — which is a good reason to pick your first market on data rather than on catalog size.

How many languages should an ecommerce store support?

Fewer than you want to. Each language adds ongoing cost: every product edit, new collection, and policy change has to propagate. Most stores are better served by two well-localized markets than six half-localized ones. Add the next language only when the previous one is converting near home-market rates.

Do I need to localize checkout?

Yes — it's the highest-value few hundred words in the store. Shopify localizes much of the native checkout automatically once a market's language is configured, but custom fields, cart messaging, and app-injected strings usually need handling separately. Test a real order end to end in the target language before you consider the market live.

Is machine translation good enough for product pages?

For most of the catalog, yes — modern AI translation is well past the point where it reads as broken. The failure mode isn't fluency, it's consistency: brand names, product lines, and technical terms drift between pages unless a glossary pins them. Bulk-translate with AI, pin your terminology, and human-review the pages that carry real revenue.

Should I use subdomains or subfolders for localized stores?

On Shopify the practical default is subfolders (example.com/de) because they inherit the domain's existing authority and are simpler to maintain. Country-code domains can outperform in markets where local trust matters a great deal, but they start from zero authority and multiply your technical overhead. Whichever you choose, hreflang has to be correct or the choice won't matter.