The real A/B test is: custom translation versus automatic in-browser translation. Huge percentage of people use the latter now. Edit: I haven't done careful analysis, but I run a large site with global appeal and 50%/50% US versus non-US traffic. Everyone is using auto-translate, I can detect it easily in the logs. Engagement metrics are near-identical. Sometimes even better, for example in some Asian countries where auto-translation is being heavily used. Regular people use the browser auto-translate now and are quite comfortable with it.
Even bigger question: will you be harming the auto-translators if you don't make your site available in English?
Automatic translators are quite decent at the hundreds of English<->Rare-Languages combinations, but much worse at the thousands of Rare-Language<->Rare-Language combinations. It's all due to the available training sets, and where the most machine learning investment has been made, and that simply this size relationship:
combinations(number_of_languages,taken_2_at_a_time) = i.e. all pairs.
is much greater than:
1 * number_of_languages = i.e. English paired with everything else.
That relation is the power of a lingua franca, at least until translators vastly improve for all the obscure language pairings.
Let's say I can almost always understand automatic English<->German translations. It's decent in that regard. But it really is not a pleasure to read.
Personally I just read the English version, and I know my peers all do the same. I wonder how the use of automatic translation differs around the world.
It may be leaving lots of opportunity on the table but at the early stages of a company - which includes many in the Hacker News community - they're not taking advantage of the opportunities that are already in front of them.. or they're still figuring out how.
Adding an additional market at that stage ranges from a waste of time/money (if it doesn't work) to a poor customer experience (if it does). How do you sell to and support a customer if you don't speak their language?*
This is not a decision you want to make lightly without some deliberate thought and planning with the rest of your organization.
* I work for Clarify.io which although we don't do translation services (and no interest in it), we work with multiple languages on a daily basis. Adding support in even one language the founders don't know is challenging at best. Luckily, between our two founders, we have 5 languages.
Obviously even a kebab shop would benefit from a multilingual website. The question is priorities, and with start ups there is heck of a lot to do before you can think of your potential non-English speaking Chinese customers.
Right, there's always things to do at a startup, this can be up there with them and eventually I strongly believe it will become more of a necessity than a luxury.
The problem is you need the requisite staff to handle support requests in a foreign language, and the logistics in place to ship across the world. That doesn't happen overnight.
Right, almost nothing worth doing happens overnight! In terms of staff though, many do have staff that speak another language for example Spanish, yet they still don't offer Spanish as a language on their website.
Later on, when hiring staff, you can make it a requirement to be at least bilingual which a lot of European and Asian companies have started doing.
Actually, parent is somewhat incorrect here in my own experience. Possibly because he's based in Asia or in the UK. The trend in Europe, from where I stand, is to expect trilingual employees for a growing number of professions (native language, English, and something else) and bilingual employees (native language and indeed English) for a great many others.
"only 27% of internet users are English-speaking."
This metric has probably less meaning than it seems . How many of them are multi-lingual. What level of English-proficiency makes somebody English-speaking? What percentage of purchasing power is within users with English skills? What is the trend of English skills in young internet users?
What languages Internet users in general speak is not relevant unless you are Google. That's like saying 90% of those who have a phone don't speak English, so if you answer the phone you have to cater to them too.
You should care about your customers first, and potential customers second. Unless you actually offer your services, can give support in the language in question, and will take payments from the users, there is really no use translating your web pages.
More importantly, and I will say this in the nicest possible way: Never ever publish anything that has been automatically translated. Never. Don't even think about it.
You don't know the number of times I've had to re-run the translation backwards to English just to guess what the original intention was... and I speak a western language closely related to English. Botching someone's language does not gain you a user.
"Never ever publish anything that has been automatically translated. Never. Don't even think about it. You don't know the number of times I've had to re-run the translation backwards to English just to guess what the original intention was."
This. I've been telling Backblaze for years that their French localization is beyond unusable, and the copy in the very first screen of the installer makes this look like some 12 year old's job, not a company you're about to trust your data with.
Have it proofread by a native or just don't do it, period.
I'm the Backblaze employee that advocated for this system we have, which is: 1) Developers must add all features fully internationalized, meaning if the locale specific file is present it will display that translation, then 2) The developer must ship a locale specific file which was machine translated, then 3) We have native speakers fix any problems. Usually the first person who complains finds out there is a static text file they can edit to fix the product. I can email you the straightforward installer French locale file if you want to help us fix it!! (Or on Windows run "install_backblaze.exe -unpackonly" and go to the folder it tells you and edit the file called InstallerConfig.xml with Notepad on Windows (NOT Wordpad) or TextEdit on the Macintosh.
This system allowed us to do a large deal in Japan, where the Japanese company helped us improve the Japanese with very very little additional Backblaze developer time involved because everything was "ready to go". We also did a pretty good German version. The Russian is Ok because several employees at Backblaze are native Russian speakers, but the Brazilian Portuguese is probably a train wreck.
I am curious, why does the presence of the badly translated French hurt you? You can always display it in English (one menu click away), or you can help us translate it better. If it is not there, there is nowhere to "start with", why does that make you happier?
I'm USA born, so it might be a cultural thing. It might be it doesn't offend and bother Japanese people to have a bad Japanese translation available (as long as they can fall back to English), but it REALLY bothers the French that bad French is ever seen in any form, or something like that?
I don't think the phone analogy is a good one to the case here. You can write software that serves people in other languages without ever speaking those languages in a way you could never communicate over the phone. Mobile apps are a good example -- it does largely depend on the product and the target audience though but most things should aim to be international. It's not massive effort either in most cases just include a LocalizedStrings.h or whatnot. Not every language has to be supported but simply have the abstraction of two languages usually means more can be added later quite easily.
The problem with translation is that it's rather expensive and only pays off at a certain point.
I make Mac apps, and they are English only, even though my native language is German. Translating the app to German would be straightforward, since I could do it myself. However, it would make future development much more complex; I'm already struggling to keep up with one language. Especially for German I don't think that the effort would pay off; most German speakers understand English reasonably well (epsecially in my target demographic).
However, some countries are different. For example, I frequently get poor reviews from Spain; it's obvious that the people complaining didn't understand the description and purchased the wrong app. It fits my personal experience/prejudice that people in Spain are less likely to speak English. However, a Spanish translation would be more difficult because I'd need to hire someone, and I'm not sure if additional sales in Spain would make up for the up-front translation costs.
But, it's something to keep in mind: Consider how much of your target demographic speaks English even though it's not their native language. There are large variations among different demographics.
Right, there definitely are pros and cons, but you'll never know the power of translating your app to Spanish until you do so.
It's an investment like any other in business, but it can definitely pay off. And once you see some results, you'll be more open to adding even more languages.
I'm not sure if you spend money on marketing your app, but let's say your CPA in English is $15, you may find out that in Spanish it's $7.
Bablic also offers inexpensive translation compared to other ways of building a new website/backend and so on...one line of code and relatively cheap monthly plans.
It depends what kind of website you run, of course this doesn't apply to everyone....and you need to build your own strategy in terms of how it works and how it will work.
But if done properly, the potential is limitless and it's a cost-effective way of reaching new customers.
In my experience you'll get them regardless of wheter your website is translated or not. For simple things you might be able to get along with Google translate (make sure to include both english and translated version in email, so they can get help from an english speaker in case the translation is gibberish)
As someone who has a startup with a multilingual web service (English / Japanese), I normally tell people considering translating their service to hold off as long as possible.
Going from have a unilingual website to a multilingual one slows everything down. Now whenever you make a change to one of the strings in your app, you need to wait to have it translated before you can deploy a new version.
Furthermore, merely translating your service isn't enough, you need to localise it. To do this, you (or someone else) needs to understand your target market. This isn't something you can outsource (cheaply at least).
Finally, adding another language to your application means you need to do customer support in that language. This means the founders can no longer do support themselves, and you'll need to rely on a third party instead.
Think carefully before going down the multilingual path.
Ugh. I can't imagine there being enough benefit to a multilingual website, especially for a b2b or technically proficient clientele, to make the overhead of translation worthwhile -- it's a cost, but much more importantly, a speed/complexity tax whenever you change anything.
If I'm going to get a shitty automatic translation, I'd prefer to do it myself on the fly with Google, vs. having a bad version actually provided by the website.
It seems simple but in my experience there's always hidden complexity and gotchas that don't reveal themelves until the localisation step.
Usually it's when there's some phrase that needs interpolation, or when it's time to run a quick A/B test, or you've emails to send to your user... Etc. Nothing that can't be solved but it adds time and complexity.
I like the advice to wait as long as possible... Look at those users/sales as an incremental that you can capture one-day, after you've picked other low hanging fruit.
Follow your market, pivot when necessary, etc. Be lean. There is an argument for having support, even minimally, for different languages while you find your market. Still, there is the catch-22 of expending resources to support potential markets that may not exist. Take for example the Lean Startup's example of building many integrations for chat applications. In the end, they were using flawed assumptions and didn't need ANY integrations. The opportunity cost of the time they spent was gone.
For myself, I had a side project unexpectedly take off in a big way in Germany although I launched it from the US. It was even latency sensitive (a web-based video conferencing solution). The site wasn't translated before that point, but it was certainly an indication that perhaps it was the time to do so...
Here's how you bootstrap multiple language support.
First, hire bilingual customer support. I have been both a bilingual customer/tech support employee, and hired people for that type of position. You want to hire someone for whom the mother tongue is the target language, and they still speak/write good enough English that they can do English-language support (because you need to train them, and you need them to also do English customer support because your second-language support volume will be too low at first). Don't bother hiring people people who speak your target language as a second language, unless they can demonstrate exceptional proficiency. Your customers need to feel HEARD, and a second language speak can't really provide that.
So you start by offering the customer service in a target language, and then you listen to your staff; they will tell you which are the worst pain points on your site for customers in your target language. Then you can do MVP-style translation, and you go from there. If you're lucky, you'll manage to hire staff that have the ability to do that translation for you, saving tons of money (though then it's worth taking those savings and hire a translation proofer).
30 comments
[ 3.1 ms ] story [ 65.2 ms ] threadEven bigger question: will you be harming the auto-translators if you don't make your site available in English?
Automatic translators are quite decent at the hundreds of English<->Rare-Languages combinations, but much worse at the thousands of Rare-Language<->Rare-Language combinations. It's all due to the available training sets, and where the most machine learning investment has been made, and that simply this size relationship:
is much greater than: That relation is the power of a lingua franca, at least until translators vastly improve for all the obscure language pairings.This is not true. They're not even good at most English<->Common-language pairs.
Personally I just read the English version, and I know my peers all do the same. I wonder how the use of automatic translation differs around the world.
Of course, make sure you have a business in your primary customers' language, I agree!
Adding an additional market at that stage ranges from a waste of time/money (if it doesn't work) to a poor customer experience (if it does). How do you sell to and support a customer if you don't speak their language?*
This is not a decision you want to make lightly without some deliberate thought and planning with the rest of your organization.
* I work for Clarify.io which although we don't do translation services (and no interest in it), we work with multiple languages on a daily basis. Adding support in even one language the founders don't know is challenging at best. Luckily, between our two founders, we have 5 languages.
Later on, when hiring staff, you can make it a requirement to be at least bilingual which a lot of European and Asian companies have started doing.
I'd guess the majority of the people who are bilingual in that setting speak their language and English?
This metric has probably less meaning than it seems . How many of them are multi-lingual. What level of English-proficiency makes somebody English-speaking? What percentage of purchasing power is within users with English skills? What is the trend of English skills in young internet users?
You should care about your customers first, and potential customers second. Unless you actually offer your services, can give support in the language in question, and will take payments from the users, there is really no use translating your web pages.
More importantly, and I will say this in the nicest possible way: Never ever publish anything that has been automatically translated. Never. Don't even think about it.
You don't know the number of times I've had to re-run the translation backwards to English just to guess what the original intention was... and I speak a western language closely related to English. Botching someone's language does not gain you a user.
This. I've been telling Backblaze for years that their French localization is beyond unusable, and the copy in the very first screen of the installer makes this look like some 12 year old's job, not a company you're about to trust your data with.
Have it proofread by a native or just don't do it, period.
This system allowed us to do a large deal in Japan, where the Japanese company helped us improve the Japanese with very very little additional Backblaze developer time involved because everything was "ready to go". We also did a pretty good German version. The Russian is Ok because several employees at Backblaze are native Russian speakers, but the Brazilian Portuguese is probably a train wreck.
I am curious, why does the presence of the badly translated French hurt you? You can always display it in English (one menu click away), or you can help us translate it better. If it is not there, there is nowhere to "start with", why does that make you happier?
I'm USA born, so it might be a cultural thing. It might be it doesn't offend and bother Japanese people to have a bad Japanese translation available (as long as they can fall back to English), but it REALLY bothers the French that bad French is ever seen in any form, or something like that?
I make Mac apps, and they are English only, even though my native language is German. Translating the app to German would be straightforward, since I could do it myself. However, it would make future development much more complex; I'm already struggling to keep up with one language. Especially for German I don't think that the effort would pay off; most German speakers understand English reasonably well (epsecially in my target demographic).
However, some countries are different. For example, I frequently get poor reviews from Spain; it's obvious that the people complaining didn't understand the description and purchased the wrong app. It fits my personal experience/prejudice that people in Spain are less likely to speak English. However, a Spanish translation would be more difficult because I'd need to hire someone, and I'm not sure if additional sales in Spain would make up for the up-front translation costs.
But, it's something to keep in mind: Consider how much of your target demographic speaks English even though it's not their native language. There are large variations among different demographics.
It's an investment like any other in business, but it can definitely pay off. And once you see some results, you'll be more open to adding even more languages.
I'm not sure if you spend money on marketing your app, but let's say your CPA in English is $15, you may find out that in Spanish it's $7.
Bablic also offers inexpensive translation compared to other ways of building a new website/backend and so on...one line of code and relatively cheap monthly plans.
But if done properly, the potential is limitless and it's a cost-effective way of reaching new customers.
Going from have a unilingual website to a multilingual one slows everything down. Now whenever you make a change to one of the strings in your app, you need to wait to have it translated before you can deploy a new version.
Furthermore, merely translating your service isn't enough, you need to localise it. To do this, you (or someone else) needs to understand your target market. This isn't something you can outsource (cheaply at least).
Finally, adding another language to your application means you need to do customer support in that language. This means the founders can no longer do support themselves, and you'll need to rely on a third party instead.
Think carefully before going down the multilingual path.
If I'm going to get a shitty automatic translation, I'd prefer to do it myself on the fly with Google, vs. having a bad version actually provided by the website.
Usually it's when there's some phrase that needs interpolation, or when it's time to run a quick A/B test, or you've emails to send to your user... Etc. Nothing that can't be solved but it adds time and complexity.
I like the advice to wait as long as possible... Look at those users/sales as an incremental that you can capture one-day, after you've picked other low hanging fruit.
For myself, I had a side project unexpectedly take off in a big way in Germany although I launched it from the US. It was even latency sensitive (a web-based video conferencing solution). The site wasn't translated before that point, but it was certainly an indication that perhaps it was the time to do so...
First, hire bilingual customer support. I have been both a bilingual customer/tech support employee, and hired people for that type of position. You want to hire someone for whom the mother tongue is the target language, and they still speak/write good enough English that they can do English-language support (because you need to train them, and you need them to also do English customer support because your second-language support volume will be too low at first). Don't bother hiring people people who speak your target language as a second language, unless they can demonstrate exceptional proficiency. Your customers need to feel HEARD, and a second language speak can't really provide that.
So you start by offering the customer service in a target language, and then you listen to your staff; they will tell you which are the worst pain points on your site for customers in your target language. Then you can do MVP-style translation, and you go from there. If you're lucky, you'll manage to hire staff that have the ability to do that translation for you, saving tons of money (though then it's worth taking those savings and hire a translation proofer).