Going Multilingual: Websites That Sell Across Borders

The orders you are not getting are already searching for you. In Germany, France, Japan — or whichever markets neighbor your home one — buyers type what you sell into search in their own language, on their own seasonal clock, and find competitors who bothered to answer in it. Language is not a garnish on an international website; it is the purchase gate.

This article is the field guide we use with clients going cross-border. The running example is Nordhaus Lighting, a modeled American residential-lighting brand with roughly $6 million in annual online revenue concentrated in the US, Canada and the UK; every Nordhaus figure is illustrative and published research is attributed. We cover market selection, hreflang structure, translation tiers, currency and payments, per-locale SEO, the cost model — and a 90-day plan.

Here is the short version.

Key Highlights

• 76% of online shoppers prefer to buy products with information in their own language, and 40% will never buy from websites in other languages (CSA Research, "Can't Read, Won't Buy").
• Roughly half of all website content is English (W3Techs), while English-speaking internet users are about 26% of the global online population (Ethnologue) — the rest is contested ground.
• Digital wallets carried about 50% of global e-commerce transaction value in 2023 (Worldpay Global Payments Report) — cards alone do not travel across borders.
• The modeled brand's first two locales, Germany and France, cost $43,000 in year one combined; Japan runs $29,000 on its own.
• Every hreflang tag needs a reciprocal return tag, or localized pages compete with your originals and rank nowhere.
• Tiering translation — machine what informs, human what sells — cuts the modeled per-locale cost by about 40% versus full human translation.
• Localized pages on a subfolder reach a modeled 8,400 monthly German sessions by month nine, compounding from there.
• Payment norms decide checkouts: Germany expects invoice and SEPA, France Cartes Bancaires, Japan konbini (market norms).

The Case for Going Multilingual: "Can't Read, Won't Buy" Is Real

The sharpest research here is CSA Research's "Can't Read, Won't Buy" series, surveying thousands of online shoppers across countries. 76% prefer to buy products with information in their own language, and 40% will never buy from websites in other languages at all. Read those together and the conclusion is blunt: a monolingual site is not a smaller version of a multilingual one — it is absent from most buyers' consideration entirely.

The supply side of the gap is just as lopsided. Roughly half of all website content is English (W3Techs), while English-speaking internet users are about a quarter of the global online population (Ethnologue) — so English-language supply overshoots English-language demand by miles. Competitors who localize are buying demand that has waited, unclaimed, for years.

One caution before the enthusiasm: going multilingual multiplies everything — pages, support tickets, tax rules, returns. The companies that win treat it as a market-entry program with a budget, not a plugin. The rest of this article is that program.

Key takeaway: Buyers cannot purchase what they cannot read — language is the first conversion factor in any new market, ahead of price, design and even product.

Step 1: Choose Markets With Demand Data, Not Nostalgia

The best market list is already in your analytics. Google Search Console shows impressions and clicks by country for queries you never localized; support inboxes show which languages customers attempt; shipping records show forwarder addresses and marketplace orders leaking across borders. Nordhaus's modeled data showed thousands of monthly German impressions for queries like "Pendelleuchte Esszimmer" — demand that arrived despite the site never having written a word of German.

Population size flatters the wrong markets. Chinese-speaking users are roughly 20% of the global internet and Spanish speakers about 8%, but for a niche lighting brand the reachable, winnable demand in Germany out-scores both. Weight the shortlist by evidence you already own — impressions, forwarded orders, inbound languages, competitor gaps — and launch two properly rather than five thinly.

The modeled shortlist put Germany first on every criterion: forwarder demand was proven and margins held after duties. France followed; Japan waited for year two — trust there is expensive to earn and worth doing properly.

Key takeaway: Rank markets by demand you can already measure — impressions, forwarded orders, inbound languages — and launch two properly before you launch five thinly.

Step 2: Get the Structure Right — hreflang, URLs and Sitemaps

Before a single page is translated, decide where it lives. Three structures exist: subfolders (example.com/de/), subdomains (de.example.com) and country-code domains (example.de). For most mid-size brands, subfolders on one strong domain win: they inherit authority fastest, consolidate engineering on one codebase and cost the least.

Then wire hreflang, the markup that tells Google which page serves which language and region. The rules that break sites are small: every hreflang tag needs a reciprocal return tag, each page must self-reference, one page carries x-default, and the whole cluster belongs in the sitemap. Get any wrong and Google quietly ignores the cluster — localized pages then compete with your English originals and, often, lose.

The failure is common enough to be its own genre: in our international SEO case study, a tour operator's inconsistent hreflang had Google showing German pages to French users and English pages to German users, while whole markets produced no revenue. Restructuring the clusters came before any new translation — structure decides what each new page is worth.

Key takeaway: hreflang is a promise between pages — every tag needs a return tag, and the structure you choose on day one sets the value of every page you build after.

Step 3: Buy the Right Translation Quality Tier

"Translation" is not one product; it is a ladder of four, and buying the wrong rung wastes money in both directions. Machine translation alone is fast and nearly free; machine translation with human post-editing (MTPE) is accurate with a flat tone. Professional human translation delivers native, SEO-aware quality. Transcreation rewrites the message for the culture, not just the words.

Map tiers to page jobs, not to budgets. Checkout flows, legal pages and brand taglines need humans or transcreation; spec sheets, reviews and help articles tolerate post-edited machine translation; everything that sells — category pages, product descriptions, campaign emails — deserves professional translation with local keyword research baked in. Nordhaus's modeled 12,000 initial words per locale, tiered this way, cost about 40% less than full human translation with no measurable conversion loss on the low-stakes pages.

Quality assurance is part of the tier: a glossary of product terms, a style guide per language, and a native review of the checkout flow before launch. The one thing never to machine-translate is the part where money changes hands — an error there costs more than every word saved.

Key takeaway: Match the tier to the page's job — machine-translate what informs, professionally translate what sells, transcreate what defines the brand.

Step 4: Localize Currency, Payments and Trust

A localized homepage with a US-only checkout is a brochure. Currency display comes first: prices in the local currency, honest exchange behavior, and duties and taxes shown up front where the market expects them. Then payment methods — and here the global picture is unambiguous. Digital wallets carried about 50% of global e-commerce transaction value in 2023 (Worldpay Global Payments Report), and the local mix shifts sharply by country.

Market norms decide the checkout. German shoppers expect purchase-on-invoice and SEPA direct debit alongside wallets; France runs on Cartes Bancaires with installments common; Japanese buyers lean on konbini cash payments and a visibly local return address. Missing the dominant method does not dent conversion by a few percent — it removes the checkout as an option for the buyers who prefer it.

Trust signals complete the loop: a local contact channel, delivery promises in local couriers, a returns policy that does not require international postage, and privacy terms that respect local law — for European markets, that starts with our GDPR primer. None of these are expensive individually; together they are the difference between "ships from abroad" and "this store is for me".

Key takeaway: Currency, payments and returns are conversion infrastructure — a shopper who sees their payment method and a local return policy finishes checkouts that would otherwise die in doubt.

Step 5: Win Search in Each Locale

Per-locale SEO is a program, not a checkbox. Keyword research starts from local search behavior, not from the English keyword list: "pendant light" becomes "Pendelleuchte", volumes shift, and the competitors are local brands with local links. Metadata, schema, image alt text and reviews need the same treatment, in the target language, on the target market's calendar.

The modeled Germany ramp shows what a proper launch looks like: traffic builds slowly through the first quarter as clusters get crawled, accelerates when commercial pages index and earn first local links, and settles near 8,400 monthly sessions by month nine. Markets that share a language with existing content ramp faster; Japan, with its separate platforms and trust signals, is realistically a year. Budget patience accordingly — per-locale SEO compounds, it does not switch on.

Catalog mechanics matter too: faceted filters, seasonal drops and indexation speed decide whether localized catalogs get seen at all — the same machinery we rebuilt for a fashion retailer in our e-commerce SEO case study. Every additional locale multiplies those pages, which is why structure (Step 2) comes before content.

Key takeaway: Translating pages puts you in the game; local keywords, reviews and links are how you win it — treat per-locale SEO as its own program with its own calendar.

Step 6: Budget the Program Honestly

Cross-border budgets fail in two symmetrical ways: sticker shock over translation, and blindness to everything else. The honest model prices four lines per locale — content, engineering, operations and patience. Nordhaus's modeled first year:

The revenue side, modeled: Germany's ramp reaches about 8,400 monthly sessions by month nine; at a 1.4% conversion rate and $85 average order, that is roughly $120,000 in annual revenue and about $48,000 in gross contribution at a 40% margin — clearing the modeled $23,000 first-year cost in around six months, then compounding. Real programs include false starts, so budget a year before judging a locale, not a quarter.

Ongoing costs are the part most plans skip: annual content refreshes at roughly a third of the initial translation spend, seasonal campaign transcreation, in-language support coverage, and annual re-testing of payments and tax. Nordhaus's plan sets aside about $9,000 a year per mature locale for all four. The platform matters here too — a storefront that makes localized campaign pages cheap to publish, like the headless build we delivered for a fashion label, keeps that ongoing line honest.

Key takeaway: Budget per locale, not per project — translation is the visible line, but payments, tax and annual refresh decide whether the market actually converts.

What Your Business Should Do

Here is the 90-day plan we give clients preparing a first or next locale:

Days 1–30: Read your own demand data.

• Pull Search Console impressions and clicks by country, count forwarder and marketplace orders by destination, and tally support inbox languages.
• Shortlist two markets with measurable demand and draft a per-locale keyword map from local search behavior.
• Decide the URL structure — subfolder, subdomain or ccTLD — with your developers before any page is translated.

Days 31–60: Build the first locale properly.

• Stand up the first locale as a subfolder with complete reciprocal hreflang clusters, x-default and localized sitemaps; validate in Search Console.
• Tier the translation: professional for checkout, categories and email; post-edited machine translation for long-tail content; transcreation for brand pages.
• Wire local currency, the dominant local payment methods, duties and taxes display, and a local returns policy.

Days 61–90: Launch, measure and decide.

• Launch, submit sitemaps, and track sessions, conversion rate and payment mix weekly against the ramp model.
• Fix what the first market teaches — glossary gaps, delivery promises, support language — before launching the second.
• Approve locale two with a per-locale budget table, and set the annual refresh and re-testing calendar.

The Bottom Line

Going multilingual is not translating a website; it is opening stores in markets that were already searching for you. The data is not subtle — 76% of shoppers want their own language, wallets carry about half the world's e-commerce value, and the demand leaking into your analytics is measurable this week. Start with two markets, build them properly, and let the second locale's numbers argue the case for the third.

If you want a second pair of eyes on the expansion math, that is literally what we do all day. Book a free strategy call and we will read your demand data, structure the rollout and scope the first locale — or send us a message with your questions and we will answer within one business day.