A second language in your shop:what actually needs translating.
The language switcher in the header looks like the whole job. It is roughly a fifth of it, and the expensive part only starts once you go live.
We write this from our own practice: our marketing site speaks twenty languages and the shop we build is multilingual from the inside out. This is not a text about why selling abroad is a good idea, but about how much work a second language really costs once the shop is alive and changing every week.
Four layers, not one
The first layer is the shop interface: buttons, basket, error messages. It is done once and usually comes from the platform. The second is content, product names and descriptions, categories, page texts. This is where the real volume starts, and where AI usually ends up, because with a few thousand products nobody writes that by hand.
The third layer is the one everyone forgets: documents and messages. Terms, privacy policy, withdrawal form, transactional emails, delivery texts. A customer who gets an order confirmation in a language they do not speak knows the shop is only pretending to be local. The fourth is commercial: currency, delivery and payment methods that make sense in that country, and taxes. You can have a perfectly translated shop that nobody can pay in.
The real cost is upkeep
A one-off translation is cheap and easy to budget. What is expensive is everything after: every new product, every change in the terms and every promotion multiplies by the number of languages. We have two answers to that. First: AI translation runs on your own key, so you settle it directly with the model provider, with no margin from us and no per-description rate. Second: a missing translation must never break the page but fall back to the source language, because otherwise the first typo in a key kills a whole subpage.
When a second language makes sense
It makes sense when you already have traffic from that country, or a product that wins there on price or availability. It does not make sense as an investment in the hope that traffic will appear by itself. Before switching a language on, check in your analytics where people actually come from, and work out whether you have anyone able to answer in that language when a message arrives. Technically, remember one thing: every language version needs its own address and correct hreflang tags, otherwise Google will treat them as duplicates and pick for you which one to show.
A second language is not a feature you switch on but a commitment you keep. If you can carry it, it opens a real market. If you cannot, one language done properly beats two done by halves.