The question comes round every year, and always in the same words: should the website be in Russian and English too? The answer is usually two flag icons in the header and the site run through a translation engine.
Before that, it is worth answering the question that changes everything else: who exactly is this second reader? There is no such thing as "a second language". There are three fairly different people.
Three readers treated as one
A Russian-speaking customer living in Estonia is not a tourist. They know where you are and what things cost here, and they want the same service your Estonian-speaking customer wants. Nothing needs explaining to them, only the same page in a language they can read.
A tourist wants to know something else: are you open, where are you, and can they book right now. Three pages and one point on a map decide it for them, not a whole translated site.
A foreign business buyer reads in English and is looking for evidence: what you have done before, and whether a contract can be signed with you. Price is often the least important line for them, the work you have already done the most important.
Those three want different pages, at different lengths, in a different order. Until you know which of them arrives on your site, the second language is not yet a decision you can make.
What machine translation does to your price page
Machine translation is genuinely good these days. It is not good in exactly the place where money starts changing hands.
The moment the text turns to prices, terms or guarantees, the machine offers words nobody in your trade uses. VAT, deposit, handover, warranty period: one wrong word changes what you promised.
And the reader notices before they have worked out why. A clumsy price page does not read as clumsy, it reads as though there is nobody behind it who speaks their language. Which is the exact opposite of the reason you started translating.
The machine can produce the first draft, but the price page, the terms and the contact page have to be read by somebody who actually speaks that language. On the rest of your copy, a machine's mistake is cheap. On those three pages it is not.
The most common outcome is a half-translated site
This is what we see most often, and it is worse than one language.
It goes like this. The homepage and the services get translated. The blog, the price list, the terms and the form's confirmation do not, because they were added later. The reader gets through two pages in English, clicks "request a quote" and lands on an Estonian form.
They do not switch language there. They leave. And they leave at the moment they were closest to buying, which is the most expensive place anybody can leave from.
One language done properly tells nobody a lie. Half a second language promises something the site does not keep.
The cost that is not in the quote
The translation invoice happens once. The upkeep does not.
From that day on, every change is two or three changes: a new price, different opening hours, one corrected sentence. What actually happens is that the Estonian side runs ahead and the others quietly fall behind over a year.
A year later the site has two languages, one of which is correct. A wrong price in a foreign language is worse than no price at all.
If you do add a second language, plan at the same time for who keeps it up to date. Not because that is expensive, but because without it the thing decays on its own.
Addresses and hreflang, without the jargon
One technical thing is worth getting right immediately. Each language needs its own address, for example /et/hinnad and /en/pricing, rather than one address that swaps its content at the press of a button: otherwise a search engine cannot index the second language and a person cannot send the link on. Beside that goes hreflang, one line in the page's head telling the search engine that this page is the equivalent of that one. Without it Google picks for itself which to show. A developer sets both up once, and most cheaply while the new website is being built.
The option nobody considers: one page
Translating the whole site and doing nothing are not the only two choices. The third option is one proper page in the second language: what you do, for whom, in which region, roughly what it costs, how to reach you, and one sentence about which language you answer in. Not a small "About us", but a real page.
A page like that costs a fraction, does not decay over time, and answers what that second reader wants to know. If enquiries start arriving through it, your next decision rests on real data. If they do not, you learned the same thing cheaply.
In our experience this is almost always a better first step for a small company than a three-language site that eats the whole budget and that nobody updates a year later.
The test that settles it
In most cases the honest answer is no, or not yet, and you already hold the evidence to decide it.
- Who calls you. Think back over the last fifty calls and messages. In how many of them was Estonian awkward? If you cannot remember a single one, that is your answer.
- What your analytics already shows. Look at visitors by country and by browser language. If nine out of ten are in Estonia with Estonian browsers, there is no market there that you are sleeping through.
- Whether anyone has asked. Has a customer ever said they could not understand the site? One sentence like that in a year outweighs this entire discussion.
- Whether anyone can answer. The simplest question, and the one skipped most often. When an enquiry arrives in English, who replies to it, and in what language? A site that invites people in a language nobody can answer in produces disappointed people.
If three of those four come back as no, your site does not need a second language. It needs something else, and that is where the money should go.
If the answer is yes, then one language at a time, one page before a whole site, the price and contact sections read by a person, and its own address for that language.
And if you do not know which of those three readers is yours, ask us. We will also say so when the answer is that one language, done properly, is enough for now.
Frequently asked questions
Most often the honest answer is no, or not yet, and you already hold the evidence to decide. Look at four things: who calls you and whether Estonian was awkward in those calls, what your analytics already shows about visitors by country and browser language, whether a customer has ever said they could not understand the site, and whether anybody in your company can answer in that language at all. If three of the four come back as no, the site does not need a second language, it needs something else.
Partly. Ordinary prose comes out of a translation engine understandable, but it is not good in exactly the place where money changes hands: words for prices, terms, guarantees and tax often come back as ones nobody in that trade uses, and one wrong word changes what you promised. The reader notices before they have worked out why. The practical distinction is that the machine can produce the first draft, but the price page, the terms and the contact page have to be read by somebody who actually speaks the language.
Because the reader crosses from the translated page onto an untranslated one at the exact moment they were closest to buying. The typical pattern: the homepage and services get translated, while the blog, the price list, the terms and the form's confirmation do not, so a click on "request a quote" lands the visitor on an Estonian form. They do not switch language there, they leave. The same thing happens over time to a properly translated site nobody maintains: every change is now two or three changes, and a year later the site has two languages, one of which is correct.
Yes. Each language needs its own address, for example /et/hinnad and /en/pricing, rather than one address that swaps its content when a button is pressed. If the address does not change, a search engine cannot index the second language and a person cannot send the link on to anybody. Hreflang goes with it: one line in the page's head telling the search engine that this page is the equivalent of that one. Without it Google picks for itself which to show. A developer sets both up once, and most cheaply while the site is being built.
Read next
For a number of Estonian businesses that page genuinely works, and it is worth saying so honestly. Four things still go undone, and one of them decides the rest: you are renting the shop, and the door is not yours.