A business translates its website into three languages based on a slide showing market sizes. The pages go live and traffic arrives.
Enquiries follow in all three languages. Nobody in the company speaks any of them. The replies go out in English, apologetically, days later, and the enquirers conclude the business is not serious about their market.
The translation cost a modest sum. What it actually purchased was three markets forming a poor opinion, which is considerably worse than having no presence at all.
Myth: pick the biggest markets
Reality is that market size tells you nothing about whether you can serve it. The right first language is one where you already have demand, some capability and a realistic route to fulfil orders.
Reality: your own data names the first language
Traffic by country, enquiries already arriving in other languages, existing customers abroad and search queries you already rank for. The answer is usually already visible.
Myth: translating the website is the project
Reality is that the website is the smallest part. Support, sales replies, documentation, policies, invoices, returns and complaints all have to exist in that language too.
Reality: support capability is the constraint
A language you cannot answer enquiries in should not be offered. This single test resolves most of the debate about which languages to add and in what order.
Myth: machine translation is not good enough for anything
Reality is that it is now adequate for a great deal, particularly for informational content and for understanding an incoming enquiry, and it remains risky for legal, medical and marketing copy.
Reality: the terminology needs deciding once
Product names, technical terms and anything you would not want translated. Without a glossary, three translators produce three vocabularies and customers notice.
Myth: you can add a language and see how it goes
Reality is that removing one afterwards is visible and damaging. Adding a language is closer to opening a location than to publishing a page.
Reality: start with one and do it completely
Website, key documents, an enquiry route and somebody who can reply. One language served properly outperforms three served partially by a wide margin.
Myth: the effort is a one off
Reality is that every price change, policy update and new service has to propagate to every language, and the maintenance load scales directly with how many you added.
Reality: the legal and policy content is the risky part
Terms, privacy material and anything regulated should be handled properly rather than machine translated, because the cost of an error there is asymmetric.
Myth: the time zone question is separate
Reality is that adding a language usually adds customers in other hours, which is the same coverage problem as weekend and holiday cover questions.
Reality: answering in their language matters more than the page
A visitor who reads English adequately still prefers to ask a question in their own language, and being answered in it is what converts, which is the ground in multilingual support for a saas with global users.
Reality: measure it as its own thing
Traffic, enquiries and revenue per language, against the cost of maintaining it. Some languages will not justify themselves and you need to be able to see that.
Start where the demand already is, and serve it fully
An assistant that recognises the language somebody wrote in and answers in it removes the support constraint that usually decides this question, which is the same first move logic as deciding what to delegate first, and the early signals are visible within days, as in website ai assistant first week. See pricing, run the numbers on the chatbot ROI calculator, and the sector view sits under small business.