| Renato |
I’m Renato Beninatto. |
| Michael |
And I’m Michael Stevens, and today on Globally Speaking… |
| Renato |
We’re talking with one of the hottest properties on the internet. One of the brands that completely revolutionized travel. |
| Michael |
Yes, business travel, personal travel, you name it, it’s changed. And, actually Renato, I think you had your first experience on this platform with me? |
| Renato |
It was. We went on a business trip to New York, and you didn’t book a hotel, we ended up in this very nice apartment, and we shared a two-bedroomed with a small kitchen, so we’re going to listen from Airbnb how they make this happen, this magic happen, all over the world. |
| Dan |
My name is Dan Hill. I’m a director of product at Airbnb, focusing on growth and internationalization, and localization is a part of that work. I joined Airbnb about five years ago and have seen the company grow from pretty early to where we are today. Day-to-day I work with product managers and designers, engineers and data scientists and marketers to think about how Airbnb grows around the world. |
| Michael |
Dan, I know I would find this hard to believe but we might have a listener who doesn’t know what Airbnb is. |
| Dan |
Airbnb is a platform that allows locals to share their homes, their experiences, with people visiting and it allows guests, travelers when they travel the world, instead of staying in a hotel or staying in other accommodation, to really experience a city and neighborhood through the eyes of a local by staying with them. Airbnb provides a platform that makes a safe, secure, easy way to transact, bring people together, online. |
| Renato |
So, Airbnb is a poster child for an industry that is global by nature, which is travel and hospitality; it’s about tourism; it’s about going places and visiting other areas. It started, naturally, here in the United States, but you have global presence. I have booked, I have stayed at Airbnbs in the Czech Republic, in Finland, in France and Germany—which wasn’t a very good experience—but it’s amazing how widespread Airbnb has become and how it is completely changing the travel business. What was the journey at Airbnb from going from a start-up to a global power and being international almost immediately? |
| Dan |
Yes, so let me first back up a little bit. About 10 years ago, now, I and a co-founder, Steven and I started a company in London called Crashpadder, which we’d never heard of Airbnb, we thought of it just about a week, actually, after they did. |
| Michael |
So, something was in the water for you guys! |
| Dan |
Well, actually, we can talk more about this but around 2007⁄2008 there was a recession, and this was sort of one of several, but one of the motivating factors, people were looking for money; people were thinking about ways they could become micro-entrepreneurs. People who were traveling were thinking “how do I find better value when I travel; how can I afford to keep traveling and discovering new places if I can’t necessarily afford the things I used to be able to?” So, we started this company Crashpadder, based out of London, and over the next three or four years, we became the biggest peer to peer or sharing accommodation provider in Europe. Airbnb, meanwhile, had begun in the US, in San Francisco, and really grown tremendously in New York and LA and these places. In 2012 we got talking to Airbnb and realized that, really, we were better off working together on this new way of travel versus competing. So, Airbnb had acquired Crashpadder, and I moved out to California, San Francisco to join Airbnb. At that point, Airbnb was a largely, I’d say the majority of the business was in the US, still. So, in 2012 the company really began looking at Europe. We opened up offices all over Europe, and we really started getting into localization and translation and really understanding how we could get into the European market. And then, 2013⁄14 was a big year of expanding, or couple of years, expanding into APAC for us. We can talk more about the different challenges along the way, there, but by about 2015, we had really cemented our global position. You look at 2016, and we started to enter some markets like Cuba and China more seriously. Today, now, we’re in basically every country in the world apart from a couple that the US has sanctions against. But, I think we’re one of the few companies, and certainly one of the few Western consumer companies that really does have a very meaningful presence in, basically, every country in the world. |
| Michael |
Yes, you’re operating everywhere. Dan, was there a different mindset between Crashpadder and Airbnb and how it was accepted? Did Americans think differently about Europeans in sharing and, then, I would imagine once you got to Asia they thought differently as well? |
| Dan |
It’s an interesting question, actually. So, I would say perhaps in a generalization, those of us who grew up in Europe were very used to, from a very early age traveling across borders, we’d often take family vacations from London and go to the South of France. And so from a very early age, I think, in Europe and in APAC, cross-border travel is a huge part of our lives. America being such an enormous country, there is so much travel you can do even within the country before people perhaps think about overseas travel. But, that said, one of the really interesting things about Airbnb and the sort of sharing economy or this idea of connecting with real people, wanting local experiences, wanting to actually meet and talk to locals when you travel, we’ve actually found is a very global trend. The differences between a traveler from China and a traveler from Cuba and a traveler from London are actually very few, they all want the same authentic experiences; they want to meet people; they want to connect; and they want to explore the world. In that sense, I think the brand and what we stand for is actually a very globally shared phenomenon. |
| Renato |
One of the things that I find very interesting is that the product is local and it’s not owned by Airbnb, you facilitate renting a bed, a bedroom, a home, a house, a farm, but all the content is local. One of the things that I find fascinating is how successful you have been using raw machine translation inside the user experience in your website. All pundits in the industry will say “this is a terrible marketing idea, to use machine translation in your website”, and you just do it, and it seems to work. What is the general experience, why do you think that that goes against all the recommendations of the experts and it still works? |
| Dan |
Great question. Obviously, there is a few angles to this. So, just for context, in the product itself or in the website and the app, we have a couple of different types of content. We have content that Airbnb has written, the text that is part of the application and the products and platform itself. That, of course, we translate—we hope—to a very high quality. We do it through vendors and in-house teams and we strive to make that as local as possible. |
| Renato |
That is templatized, right, you just answer the questions, there, how many rooms you have… not the description but the correct logistics of the location, what are the policies, you can smoke, you can’t smoke, things like that. It’s like check the box so it’s standard for everybody. |
| Dan |
Yes. All that structured data, or just the help center or articles about how it works, are all translated by us and to a high quality. We then have content that is produced by the host; they’re writing about their listing, their location. And we allow the host to write that description in as many languages as they speak. Over 50% of our hosts are non-English speakers, and so a wonderful thing about large parts of the world is many people are bilingual or trilingual. And so we have many hosts who can write their listing in Spanish, and they can write it in English, or they can write it in German and French. And so in that sense we have seen a lot of people, already, are able to write their listings in multiple languages. And then we have a third category of content which is things like the reviews that guests leave. So, I go to stay, perhaps, with an Italian; I don’t speak Italian but I can make my… So, maybe the listing is in Italian but my review is in English, of that listing. So, we have these three parts of the puzzle that we try to match up. 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, perhaps. It’s when you go somewhere, where you feel a little bit outside your comfort zone, and you’re trying to get by with a few words. That’s where I think, actually, people’s expectation of the quality of translation is they are not expecting everything to be in perfect English, or perfect French, perfect Italian. It somewhat loses a little bit of that authenticity, I think, when it feels very standardized and homogenous. So, that’s why I think we have been able to be successful with more machine translation, whether it’s Google’s old model or, as you have in your podcast, some of the newer neural net translation. Actually, that doesn’t, I think, detract from the experience of reaching out to somebody who is different from you and speaks a different language. |
| Renato |
My personal experience, yes, I had to rent an apartment in Stuttgart, and the owner was Iraqi, and he spoke Arabic and German; he didn’t’ speak English. And I don’t speak German or Arabic. And we communicated. We chatted, we negotiated, we set an appointment to meet at the location all through the Airbnb interface; for me everything in English using machine translation, and it was quite efficient. I was impressed. |
| Michael |
Yes. There is something… you used the word authentic and genuine to create that. That creates an issue of trust because you know “yes, I’m dealing with a person, and this person happens to be different from me” and maybe there’s even a greater level of patience in expectations. |
| Dan |
Absolutely. I think the people who use Airbnb…what Airbnb really offers is a chance to explore neighborhoods you wouldn’t otherwise get to be in, to actually connect with local people or local experiences when you travel. And if you look at Paris, for example, if you look at where the hotels are predominantly built, they’re built in the downtown areas, they’re very close together. Airbnbs cover a whole range of neighborhoods and areas. So, what Airbnb offers and what people who use Airbnb love about it is a chance to actually connect with more of the everyday local people in a particular place to discover the things that locals want to do and how they live their lives and what it’s like to be a Parisian in this case. People are very understanding. People get that people speak different languages; have different cultures; and that’s part of, I think, the magic that people actually love about travel, about Airbnb. |
| Michael |
Dan, you shared with me one example that wasn’t exactly a success for you guys, related to a marketing campaign you had, and it was about inviting someone into your house? |
| Dan |
Yes. |
| Michael |
Do you mind sharing that? |
| Dan |
Yeah. I think whilst a lot of the principles of Airbnb and the values that we have are very global, the idea of inclusiveness, diversity, meeting people, connections, local experiences, the way different cultures and different languages express those things is very different; different languages have very different words to mean a lot of these more nuanced ideas. We were internally looking at trying to sort of… we ran a campaign around, it was called “one less stranger”, and the idea was that when two people from different cultures connected there was one less stranger in the world. And when people aren’t strangers, when you understand each other, that’s where we get tolerance and inclusivity. |
| Michael |
A beautiful idea. |
| Dan |
Yes. However, translating the phrase “one less stranger” in different languages is actually very challenging. In English, those words, everyone gets them, and they get the poetry of it. In German, we had a hard time, and the word stranger is actually very close to the word for foreigner, and we had a hard time really trying to figure out how do we translate “one less stranger” into languages like German, where it’s not one less foreigner we’re trying to say. So, that’s where it was a real challenge. |
| Michael |
Yes. And you guys ended up, you went forward with the campaign? Or did you make some changes? |
| Dan |
We did. We absolutely localized it. We transcreated the idea into each local market. I can’t remember exactly where we ended up in German but that was one of the toughest ones to figure out how to express that sentiment in that language. |
| Michael |
And in the context, this was a while ago, but the ideas of immigration and all the challenges there, you definitely didn’t want to hit a hot button on that. |
| Dan |
Right, absolutely. |
| Renato |
Dan, one of the things that I love about the business that we’re in, this concept called localization, international, plus the travel element on your case, is unexpected patterns that you find. Airbnb is a huge data collection environment. Was there something totally unexpected that you have found in the behavior of different cultures or different groups of people that you were able to derive from the data that you collect through Airbnb. |
| Dan |
Perhaps one interesting observation that I had, which was counter-intuitive to me was that, actually the opposite almost, people are far more similar globally than I expected them to be. I expected that when we thought about localization and translation, particularly of the broad concept of Airbnb and the product itself, that we would have to do a lot of very, very different things by different markets to be successful. Now, for sure, things like payment methods; things like literal language of the product itself, we absolutely translate and localize. But, as you said in your example, people are actually quite capable of communicating through a shared language somewhere, they can figure out how to communicate. Even tools like Google Translate allow people to really communicate effectively. When we look at, particularly today, things like mobile penetration around the world, iOS and Android, actually almost every country in the world has mobile technology now. So, what was a surprise to me was actually how similar a lot of the world is when it comes to travel and language and thinking about how language is used when you travel. Perhaps some of the things that were actually different, by different regions. Recently, we were doing some work on our machine translation, and how we could improve it. And we had a hypothesis that machine translation, if we think you can’t read that language, is a good thing; it helps you better understand the content. And we got some really interesting results out of that. One of the findings we’re hitting on actually—although we can’t put a number to it yet—many more people are bi- or trilingual than we expected. And so, although the listing may be in French, and we think you only speak English, we actually found that machine translating it from French to English was, often actually, a worse experience than you actually being able to read the original French with a little bit of difficulty. And so, I think particularly as we look across Europe and APAC, we’re actually… I think we have been under-estimating how many people are, actually, reasonably fluent in more than one language. |
| Renato |
So, you will see that they will go in, let’s say, with their Italian user interface, and they read the room description is in Portuguese, or French, or English, and they will not use the machine translation; they will just book without clicking that button. |
| Dan |
Actually, yes, what we found is in that example, we tested automatically translating it from the original source language to the language that your interface is set to. We actually tested it both with machine translation, and we ran a large test with professional high-quality translation. We actually just found that we had assumed that in this case the Italian couldn’t read Portuguese, say, actually in many cases people, I think, are much more capable of reading these other languages than we thought. |
| Renato |
Proving the opposite is always good news, you tested something and you proved yourself wrong; it’s always a positive outcome. That’s nice. How about destinations? I understand that people travel to certain destinations for different reasons. So, we know that the Chinese will travel to certain locations for shopping. I know being a Brazilian that Brazilians would come to the United States, essentially, to shop. And one comparison that I heard is that the Japanese and the Brazilians spend more or less the same amount of money per trip to New York, but the Brazilians do that in 10 days and the Japanese do that in three days, or something like that. |
| Michael |
Dan, I think this leads us into something you’ve written a bit about, and that is, there are a lot of variables when you are trying to price someone’s house or room, globally. How does Airbnb approach that part of the business? And let me add another layer to this question. How do you think we could use your experience and your knowledge in using the data that you have to determine pricing in the language business? Is there any knowledge that we can transfer to this business? |
| Dan |
Great question. To give a two-minute summary of the pricing and are still doing, we are essentially on the host side on the supply side, what Airbnb is really doing is enabling micro-entrepreneurs, people to take an asset they already have, their home, or a passion they already have, they do, and turn it into a bit of a small business. For many people the part they are amazing at is providing hospitality. It’s being welcoming people into your home, into your neighborhood. But, some of it is not obvious, and it’s hard, like “how much should I charge for my time or my space?” A few years ago we were trying to understand how to help people list their space on Airbnb and people are very easy at filling out their name, their address, how many bedrooms they have in these things, but the question of price was one people had never really had to think about before, “how much is my place worth? What is a fair price; do I compare to hotels; do I compare to my neighbors; do I compare to what I’m trying to get out of this; how do I think about it?” So, we started using machine learning. We obviously look across millions and millions of past bookings and start to provide recommendations for people based on this type of listing that you have; based on the location; this is what we think a good price would be; this is what the market rate is for that type of listing. And since then we’ve developed that into think about things like seasonality; to factor in major events that may happen; we’ve launched things that let hosts more input what they consider their lower and upper bounds, and then we can optimize within that. So, there is a variety of stuff we’ve done, there, to help people think about how to set their price. |
| Michael |
I guess the demand would probably be influenced a lot by that. Less people from the US traveling to Canada because the dollar is not doing as well, and so they just go somewhere else where it could be better; the algorithm’s not as effective. Got you. |
| Dan |
There is some work we’re doing right now, actually, which is in the localization space, translation space, which is somewhat similar. We have actually started to try and predict which are the most important languages for your listing to be translated to, based on where we see people visiting that location. So, we’re building a model that we’re hoping will tell us “great, if you publish a listing in this neighborhood in London, it should absolutely be in French, Portuguese and Croatian. If you publish this listing in Tokyo, it really needs to be in English, simplified Chinese and French”, for example. So, this is where we’re trying to use big data and looking at machine learning to actually help us really understand, rather than thinking about translating everything into every language, what are the actual main corridors, and what are the languages barriers that are most difficult to overcome, and which are the ones that don’t need the additional help of translation? |
| Renato |
That’s fascinating! You are not going to translate into Turkish to a place where you never get Turkish people. But, that’s also you take away the opportunity of Turks who want to go there, to go there. But, if you have the trends and if you have the majority, that’s what you want to cater to. I think it’s fascinating. |
| Dan |
Yes, so for us it’s really of course we want everybody to be able to travel anywhere. And for us this is perhaps a way of helping prioritize that effort and that work. Instead of thinking about it country by country, we can think about it in a much more granular and more specific way, and will it identify particular listings, locations, and machine learning can allow us to do that on a large scale, so that we can actually be like “great, these 10,000 listings are the ones all over the world that would most benefit from being in French as well as English”, say. Now, of course, our long-term goal is to make everything available to everybody but it’s a way that we can actually start to think about that problem. |
| Renato |
Well, it’s like a fit for purpose localization. Localize almost on demand. Well, you have on demand right now with machine translation, but you can optimize and increase as a host the likelihood of renting your place by localizing into the people who go into that area. That’s great. |
| Dan |
Exactly. So, one of the things I’m hoping we might get out of this is the ability to say to hosts “hey, if you were to add a French description to your listing, you would see X% more business; that would really help get a million more people interested and be able to understand you and your placing.” So, I think there’s a lot we could do with this. We’re still early in this but I’m excited by it. What I really love about Airbnb and the work we’re doing around localization and translation is it’s this wonderful melting pot of what it means to both… the goal of translation is to help people who couldn’t understand something understand it in a different language. But then, of course, we have this offline component that when you actually meet the person or you have to communicate with that person, you actually can’t pretend that everyone speaks the same language. So, it’s this wonderful intersection of both what does it mean to enable people to recognize that people do speak different languages and can’t always communicate as easily, but with the technology we can bring to make that a much more seamless experience online. And so I just find this very interesting melting pot between online and offline and what language really means in the real world every day to people, but also what it means online in a product and a technology, how we think about those things. |
