Multilingual stores: how translation errors quietly kill trust in your biggest new market

A German shopper adds a jacket to her cart, gets to checkout, and the button that should say "Complete purchase" instead reads something closer to "Force purchase." Nobody programmed that. A machine translation engine chose the most literal synonym for a word that had five better options, and nobody checked. She closes the tab. You never see her again, and your analytics show only a cart abandonment with no explanation attached.

The size chart problem

Translation errors on ecommerce sites rarely look like typos. They look like small, plausible-sounding mistakes that a native speaker would catch instantly and a store owner, skimming past in a language they don't read, would not.

Size charts are a classic failure point. Machine translation tools frequently mishandle unit conversions, garment terminology, or regional sizing conventions, producing a chart that reads fluently but tells the customer the wrong thing. A shopper who orders based on a mistranslated chart and receives the wrong fit doesn't blame the translation. They blame the store, leave a negative review, and file a return. The cost isn't a bounce, it's a refund, a shipping loss, and a public complaint in a market where you have zero reputation to draw on yet.

Checkout language causes a different kind of damage. Buttons, error messages, and confirmation text carry an outsized trust signal because they appear at the exact moment a customer is deciding whether to hand over payment details. A checkout button that sounds aggressive, a shipping disclaimer that reads as a threat, or a form error that translates as gibberish all say the same thing to a first-time buyer: this store wasn't built with me in mind. That impression is very hard to undo once formed.

Where machine translation actually breaks

Machine translation has gotten remarkably good at fluent, grammatically correct output. That fluency is part of the danger: broken translations used to look broken, with garbled syntax that signaled "don't trust this." Modern engines produce sentences that read smoothly but say the wrong thing entirely, and fluency without accuracy is worse than obvious clunkiness because nothing about it triggers suspicion.

The failure modes worth knowing:

Short strings with no context. UI elements like "Add," "Save," or "Apply" are ambiguous in English and translation engines guess at intent without knowing whether "Add" means add to cart, add a note, or add an address. Store platforms that auto-translate theme strings are especially prone to this because the translation happens without seeing the surrounding page.

Idioms and brand voice. Phrases like "free returns, no questions asked" or "you've got this" don't translate literally, and machine engines often render them word for word, producing text that's grammatically fine and semantically strange.

Units, sizes, and currency. Sizing conventions, shoe width letters, and clothing measurements vary by region in ways that pure language translation doesn't fix. A translated size chart still needs a localized conversion, and those are two different jobs that get conflated into one.

Legal and shipping language. Returns policies, customs disclosures, and tax language carry legal weight, and small translation errors here create real liability or customer confusion about what they're actually agreeing to.

For more on how checkout-specific failures cost stores money even in a single language, see checkout is where stores silently bleed.

Localization is not the same job as translation

Translation converts words. Localization adapts the whole experience, including currency formatting, date formats, payment methods, address formats, and cultural assumptions baked into imagery or copy. A store can have perfect translation and still fail localization: correct Japanese text on a page that displays prices in a format Japanese shoppers don't recognize, or a checkout that only accepts a payment method rarely used in that market.

Shopify, WooCommerce, and BigCommerce all offer approaches to multilingual selling, from native multi-currency and language features to third-party apps, but none of them fully separate the translation layer from the localization layer for you. That gap is where a lot of stores lose money without realizing it, because the site looks finished in the new language even though it isn't functionally ready for that market's shoppers.

This is also where plugin and app choices compound the risk. Stacking multiple translation and currency apps to cover gaps in a platform's native tools increases the chance that one update breaks another's output silently. If you're already running several apps to patch localization gaps, it's worth reading how many apps is too many alongside your localization plan.

A practical review routine

Full professional review of every page in every language isn't realistic for most stores under $5M in revenue. A lighter, repeatable routine catches most of the damage:

Prioritize by traffic and stakes, not by page count. Checkout, cart, size charts, shipping and returns policy, and your top 20 product pages by revenue matter far more than translating every blog post or FAQ entry perfectly.

Have a native speaker spot-check checkout copy specifically. This is the highest-stakes, lowest-word-count section of the store, so it's the cheapest thing to get a real human to verify even if you can't afford a full translation review elsewhere.

Re-check after theme or app updates. Translation strings often live in theme files or app configurations, and an update can silently revert custom translations back to a machine default. This is the same kind of quiet regression covered in Shopify theme updates: what breaks, just applied to language instead of layout.

Test the actual purchase path in each language, not just the homepage. Add a product to cart, proceed to checkout, and read every string a customer would see, including error messages triggered by an invalid card number or empty field. Those edge-case strings are the ones store owners forget to check and machine translation handles worst.

Watch for silent breakage the same way you'd watch for a broken link. A translation that worked at launch can drift out of sync as you add new products or update policies, the same way links rot over time; the discipline in broken links do not complain applies just as well to stale or missing translations on new pages.

Where this fits into store health

Language errors don't show up in an uptime check or a payment gateway alert, which is exactly why they persist for months without anyone noticing. Cassian™ monitors the store-level signals that tend to correlate with these blind spots, such as broken links, slow pages, and SEO health across markets, so that a new region isn't just launched once and left unattended. The Cassian Score™ won't read your Japanese size chart for you, but a store owner who's watching the technical layer closely tends to catch the language layer too, because both fail from the same root cause: something shipped and nobody checked back.

Multilingual expansion is one of the few growth moves a small store can make without spending on ads, but it only pays off if the new market's first impression matches the quality of the original. A careful, periodic review of the pages that matter most costs a few hours. A silent translation error in your checkout button can cost a market.