Localising your content may seem intimidating enough even for your ‘standard’ website – but what happens when you’re dealing with huge volumes of content, including colossal amounts of user-generated content such as reviews?
This article examines the approaches adopted by two of the biggest players in the online travel industry and the challenges faced when embarking on content localisation projects of this scale.
Skyscanner’s 3-million word challenge
Travel aggregator Skyscanner prides itself on drawing from a wider base of travel information than other providers to give its 60 million monthly users access to the best travel deals across 30 different languages.
Because Skyscanner takes data from a variety of sources and serves it to a number of different language audiences, there’s a lot of translation work required – over 3 million words a year in fact. Things move fast – new deals are being added over the course of each day, while others expire – so it all needs to be done quickly. It’s also important that translated content is localised correctly for each market and that the overall tone and style works for that audience.
Skyscanner’s approach is to use a combination of machine translation and human translation, allowing translators to focus on elements such as testing, evaluating the quality of automated translation work and maintaining a glossary and style guide for consistency.
While machine translation most of the heavy lifting and helps to deal with otherwise unmanageable volumes of content, translators act as brand guardians, ensuring the accuracy of all the communications across every language version.
For example, linguists will make sure the more common phrases are correctly translated, so phrases such as ‘flights to (city)’ are translated consistently with just the city name updated each time. Skyscanner works with multiple external language service providers for each language or location so they are coordinating a lot of different suppliers and content streams.
They also use local, in-country marketing specialists to ensure they get the best advice on what will appeal to particular markets and audiences.
There’s a real need for human translators to work with machine translation to increase efficiency but also in order to make sure the end product description is suitable for the target market.
As an example, the aggregator tool might recommend a range of hotel options to the user. It’s important to check the vocabulary describing what the hotel offers is not only clear and understandable to the user, but engaging and persuasive at the same time so that users are encouraged to click through and add the items to their bookings.
European and Japanese users have very different cultural expectations for what a bathroom is, for example. In Japan, the bath and toilet areas tend to be separate rooms and it’s important to be clear which is on offer as part of the accommodation.
One of Skyscanner’s biggest problems is identifying when there’s a problem with translated content, as it’s not always possible to see there’s an issue by observing patterns in user behaviour or comparing conversion rates. This means the site has to rely on human moderators to spot and flag issues.
What’s particularly striking about Skyscanner’s approach is the way everyone is involved in translation work, right across the business. This attitude and culture mean that translation work isn’t just the work of the brand’s small internal translation team but a vital activity that everyone in the business shares responsibility for.
And with news of a rebrand and a redefinition of its brand purpose shortly being implemented, having all hands on deck will prove vital for brand’s continued international success.
Expedia’s localisation project
As one of the world’s largest travel brands, Expedia, operates over 200 travel websites in more than 75 markets and estimates that it translates and localises around 200 million words per year. Expedia faces an even greater challenge of scale than Skyscanner.
The brand’s approach is also centred around combining machine translation with human translation. It’s a question of being able to efficiently manage volumes through the use of machine translation but, at the same time, carefully control the messaging by linguists in different localisation teams.
Like Skyscanner, they use machine translation supported and improved by human moderation in a bid to maintain quality. Expedia spends time and resources on creating dedicated local language content for specific markets rather than creating content centrally and then localising and translating it for delivery into the various markets in which it operates.
Like any brand working in a highly competitive market, Expedia has to keep a weather eye on new developments – specifically in translation-related technology. What is interesting is that despite neural machine translation increasingly being adopted by brands hoping to improve the quality of their localised content, the brand claims that neural machine translation isn’t yet the right choice for them.
Their chosen approach is to use statistical machine translation combined with rigorous testing to assess for quality – although they state that they’ll implement new machine translation technology such as adaptive and neural when they are confident it can deliver significant quality and efficiency improvements.
This will be a risk for these brands and they’ll need to track a major change in machine translation technology using rigorous testing. Scaling resources remains a key challenge for Expedia, which doesn’t just rely on MT to scale quickly but also scales up internal teams and external language service providers so quality can be maintained as ambitions expand and workloads increase.
Expedia relies on human translation teams as linguistic strategists, helping to govern the tone of voice in particular markets and managing the brand’s identity via house style guides.
The challenge is compounded by the fact Expedia manages a large number of different travel brands, products and platforms with their own unique brand identities and different target audiences. Maintaining quality and consistency remains an on-going challenge right across the board.
Quality is paramount
Few businesses have to deal with the volumes of content produced by these travel giants. These kinds of projects demand particular solutions, with human post-editing of machine translation and quality assessment being important parts of the mix.
Even brands producing much more modest volumes of translation work need to learn from this approach and recognise the importance of quality and consistency.
Both Skyscanner and Expedia have recognised that their entire business depends on providing users with reliable content they can understand and trust. How localisation is approached is a key part of that.
Both brands also recognise where the ‘danger zones’ are. Launching new products is a good example – that’s when human teams are deployed to conduct rigorous testing. They’ve identified hazardous areas of language – such as Japanese words for bathroom – and found solutions.
With huge audiences to serve across multiple cultural and linguistic markets, both brands have recognised the need to take a structured approach so they can easily scale their offering by having local teams in place and processes and tools such as style guides and glossaries to help them produce high-quality work.
Whether you’re approaching a small or large translation and localisation project, there’s a great deal to take away from these examples to develop your own international expansion strategy.