The Process of Making our Marketing Website Multilingual

Part 1: Languages and Translation

As many of you reading this post know, Transifex recently got a new look for both our product and our marketing website (and we’re, of course, very proud of the result!).

We followed the process we advise our customers to take based on our expertise in the area of delivering multilingual customer experiences quickly and at relatively low cost. It’s a results-oriented approach designed to ensure you deliver a high-quality local-language website that feels authentic to your customers.

So to help other companies, particularly smaller to midsize businesses thinking about going global, we’re presenting a series of blog posts detailing the process of making our marketing website multilingual. Every two weeks beginning today, we’ll publish an installment in the series, covering these topics:

• Deciding which languages you want to launch and how to translate your content
• Determining which content needs to be translated and preparing your translators
• The process for launching the website - publishing process, design considerations, infrastructure set-up
• How we’re maintaining the site and measuring effectiveness

So, without further ado, here’s a look at how we decided which languages to publish on Transifex.com and how we sourced our translations.

Decision factor #1: resources

I almost hate to start here (and when I started this post, I didn’t), but the reality for many businesses, and certainly most (all?) small- to mid-sized companies, is that your available resources will dictate the scope of your expansion efforts. Even if you could show a theoretically positive ROI for delivering a multilingual web experience in the top 40 languages used on the internet, you’re just not realistically going to do that.

So the factors we considered were:

1. 1. How much available budget did we have to purchase translations?We knew that we’d be sourcing professional translations for our key languages, at a minimum, so we needed to go into things with a realistic idea of what dollars were available, not just for launch, but also for on-going updates and maintenance. We estimated our per language launch budget based on the word count of our prior website and the average cost per word we had paid for previous translations.

1. 1. What could our staff handle?We have a small marketing team and no dedicated localization manager. We needed to understand what it would take to not only support translators and translations, but also the infrastructure and visual/display decisions we’d need to make for each localized version of our website.

1. What did our project timeline dictate?We laid out a huge stretch goal of redoing our entire marketing website in just over 2 months. That meant we knew we’d be working on the content right until the bitter end, without a lot of additional capacity to create style guides and train new translators. We’d also have a very short turnaround time to get translations completed. As such, we wouldn’t be able to effectively manage the process of delivering lots of languages for the launch.

The combination of these factors dictated that we’d be looking at delivering no more than 5 translated versions of our site. This small number also confirmed that we’d be taking a language-specific approach, not a country-specific approach. What do I mean by this? Specifically that we’d be delivering English content to English speakers generally, whether they were coming from the US or the UK or Australia. Both our translations and our URL structure would be implemented to serve this approach.

Decision factor #2: supporting local offices

The Transifex headquarters are in Menlo Park, CA, so our main website, like many companies’, is published in English. We are at heart, though, a Greek company with a good-sized development and support team in Athens, Greece. Although Greece doesn’t represent a huge customer market for us, it’s an important locale because we’re part of the vibrant startup community there and we’re growing our workforce; to be an effective local presence and, in particular, an effective recruiter, having our site in Greek is essential. No further analysis required - we’d publish in Greek.

Delivering translated content in Greek
We’re lucky that we have many English/Greek bilingual employees, so in the past, we handled all our translations internally. This ensured that the specific understanding of our business and any unique terms were baked into all our Greek content. However, it was quickly apparent that this approach was not going to work for the relaunch -- translating an entire website in just a few days’ time was absolutely too much to ask from a team already working night and day on our product update.

So plan B? We turned to our existing language service provider, e2f, to determine whether they could support our needs for Greek translation in the timeframe we were requiring. Luckily the answer was yes, they do support Greek and they do have the ability to both translate and review within three business days, our required timeframe.

Going forward, we may transition back to having at least some involvement from our Greek team, either as translators or as reviewers, and in either case, they’ll be able to leverage the Translation Memory (TM) built during this launch phase to support their future work. This is a great option for many companies - using an external translation or language services provider (LSP) for translation, and then leveraging local staff for review, or alternately having local staff translate updates and engaging a LSP to do periodic reviews. The benefit of this type of multi-resource process is not only high quality translation, but also instilling a real sense of ownership within the local team.

Tips for your decision on supporting distributed business office(s)
If you have multiple foreign offices, the language decision might not be quite as clear cut as ours. Factors for every business to consider when it comes to local office support:

• Strategic importance of each office to your overall business
• Size of the local workforce and on-going recruiting needs
• Market opportunity based on either current returns (revenue, profit) or potential growth (more on this in the next section)

Decision factor #3: current market coverage and market potential

Our next step in determining our target language set was to look at the geographic breakdown of our customer population and the geographic source of our web traffic. Not surprisingly, the lion’s share of our revenue and traffic comes from the US. The interesting analysis was for the remaining geographies.

From a revenue perspective, our key countries/languages (e.g., the next 10 after US English) are largely Western European -- UK/English, France, Germany, the Scandinavian countries -- with the exception of Israel and Japan. From a traffic perspective, our main geographies are slightly different, with Russia, Brazil, China and Spain bubbling up, while the Scandinavian countries, Israel, and Japan don’t make the top 10.

From here, we analyzed the potential of the locales that are providing traffic but haven’t reached a critical mass of sales yet: Spain is growing (plus we have experience with translating into Spanish), so Spain would be in; Russia hasn’t shown any real sales traction, so it didn’t land on the priority list; and China and Brazil are both definitely growing but from a very small base, so they got placed on a secondary priority list, to tackle when we expand to our next set of languages.

This left us with the following languages (in addition to Greek):

• French
• German
• Spanish

We knew from our prior experience that the sales we’re generating in those locales provides us the return we need to continue to justify these translations. But we didn’t just look at the cold, hard numbers. We also assessed how these choices would impact the markets where we do have customers but don’t localize. On the positive side, we felt that these languages would provide adequate (maybe not ideal) support for the Nordic countries, where English is widely spoken, and many of the other Western European locales where French, German, and English are spoken by a critical mass of citizens. Even Israel and Japan might be acceptably supported with English. Perhaps the one question mark was Chinese; however, we haven’t localized our site to date for any Asian markets using double-byte characters so it seemed prudent to handle this in phase 2.

Delivering translated content in French, German and Spanish
For these markets, choosing to go with our professional translation provider was an easy decision (as stated in our upfront resource assumptions): we knew we wanted to deliver high quality translations, we had translated the content for the prior version of our website into these languages with our provider, e2f, so they have experience with our content and expectations, and we had the data that our financial return would more than justify the translation cost.

Decision factor #4: community

The final element of our analysis was to consider countries where we have a large network of involved users even if we don’t have a minimum level of sales. This is a factor that is fairly unique to Transifex that most businesses won’t need to consider, but it was part of our process, so we wanted to document it as well.

We’re lucky to have this large community stemming from a combination of our open source roots and the distributed nature of our business, where translators and developers from all over the world are involved with delivering translated content for our existing customers. That community enabled us to publish the prior version of our website in Russian, Turkish, and Bulgarian, handled completely by dedicated community members free of charge. We love this participation and enthusiasm from our community!

For our new site, we wanted to take a best-practices approach for delivering our marketing messages in other languages, both by working more closely with our translators to ensure we are customizing our top-level messages with care as well as ensuring that we always deliver a completely translated experience across all our languages. We decided that due to our short timeline and our lack of any internal staff with language capabilities in these tongues to manage the activities of the community, we couldn’t launch using community delivered translations and meet our branding requirements. We may return to crowdsourcing some content in these languages from our community, but that will be when we are able to manage the process more actively.

Tips for leveraging your community / crowdsourcing
We have multiple customers that leverage crowdsourcing very effectively across an extensive set of languages, typically for product interfaces and in-product content. Don’t be fooled, however, by the lure of “free translations” -- delivering good results working with your crowd requires a different type of resource investment, including active and constant management of the community. To learn more about the proper techniques for implementing an effective crowdsourcing program, read Getting Started with Crowdsourcing.

Our launch plan was set -- we decided to launch our new site with 5 languages: English, Greek, French, German, and Spanish. Our next step was to review our content/translation requirements and prep materials for our translators. We’ll discuss this next step in two weeks.

In the meantime, if you have comments or questions, please add your thoughts here or email us at hi@transifex.com!