The sharing economy—where individuals exchange, rent or share goods and services with their peers—has gone global. Airbnb, for example, operates in almost every country in the world. In this episode of Globally Speaking, we discussed the sharing economy with Dan Hill, Director of Product and Growth at Airbnb. Here are the main takeaways.
Airbnb is a pioneering platform that allows travelers to book sleeping accommodations and experiences (day trips, outings, walks around town) with locals, instead of booking hotels or other traditional accommodations. This allows them to deeply experience a city and neighborhood through the eyes of a local.
Dan’s efforts have uncovered new insights about the global traveler’s language expectations and use, which in turn impact Airbnb’s localization strategy. Also, their trial and error with programs such as MT and machine learning have led to similar surprises and adaptations in their localization approach. In this chat, six revelations about their efforts at global expansion really caught our attention.
1. Travelers aren’t that different from country to country
People are far more similar globally than Dan expected them to be. It doesn’t matter if travelers are from China or Cuba or London, they all want the same authentic experiences. Travel expectations and wishes are a “globally shared phenomenon” as Dan says. “They want to meet people; they want to connect; and they want to explore the world.” Also, they have similar attitudes about how language is used when traveling. For example, travelers do not expect to have experiences in their own language when they go abroad.
2. The world is more multilingual than you think
In large parts of the world, many people are bilingual or trilingual. Dan thinks we have all been under-estimating how many people are reasonably fluent in more than one language, particularly across Europe and APAC.
How this manifests at Airbnb is that there are many hosts who can write their listings in more than one language. This broadens their own sales potential without any help or additional cost. This also can mean that some renters don’t need translations either. Though they still may prefer translated content, not having it doesn’t cold-stop a sale.
3. Users don’t expect flawless language
Even if a seller lists properties or excursions in a second language, the customer doesn’t expect it to be perfect. They know their host in another country might not speak their language fully or at all.
Dan says, “I think people understand the nature of this product, and what Airbnb is, is about travel and meeting people; and part of that magic of travel is the sort of slightly awkward language barrier.” Traveling means taking yourself a little bit out of your comfort zone, and struggling with the language—perhaps getting by with just a few words is part of the authentic experience. That language challenge is indeed part of the fun.
4. MT may or may not be useful
Airbnb has used Machine Translation (MT) to translate some of their content, such as user reviews, but with some unanticipated results.
The team surmised that MT would be great for those who could not read the local language; obviously, it helps you better understand the content. Right? But one of their findings was that the user’s experience with MT is not always better than reading a second language with a little difficulty.
Even though getting a translation is easy—an MT option is part of the interface—some renters seem to not need that translation after all. People are using content as-is, untranslated, and still succeeding at booking their stay.
(Despite this, MT still has a place in Airbnb’s program, including, potentially, neural machine translation.)
5. Transcreation is difficult, but worth it
Airbnb ran a campaign around the idea that when two people from different cultures connect, there is “one less stranger” in the world. It is a beautiful idea, central to Airbnb’s values: when people aren’t strangers, when they understand each other, there is more tolerance and inclusivity. However, translating the phrase “one less stranger” into different languages was very challenging. In English, this has a poetic, positive feel. In German, however, the word “stranger” is very close to the word for “foreigner”; it’s not “one less foreigner” that they wanted to say. So, to avoid the message getting misconstrued, they transcreated it instead for each local market. This meant that Airbnb was able to preserve the value and intent of their message worldwide.
6. Machine learning can help with language selection
It’s not always feasible or logical to translate all content into all languages, but how does an enterprise decide which ones to do? Airbnb has started to use machine learning to predict which languages should be included in specific listings based on where past travelers booking in that location are from. For instance, if a host publishes a listing in Notting Hill in London, and if most travelers booking there are from France, Portugal, and Croatia, then the listing should be in French, Portuguese, and Croatian. And this can happen at a grand scale: the data can show the listings all over the world that would most benefit from, say, being in French as well as English.
In this way, languages chosen will cater to most travelers to that area…while saving money by not translating into every language.
As large global industries evolve their language programs, these kinds of revelations from trial and error will surface. Gathering and analyzing data in unique ways can also spur new, unconventional ways of reaching your customers and bridging cultural gaps. Airbnb’s global program is a forerunner: their a-ha moments and the resulting changes to their localization strategy provide insights that may apply to any global business.