Lessons Learned Scaling Airbnb 100X

Jonathan Golden
12 min readAug 15, 2017

On my first day at Airbnb, in early 2011, I walked in to see cofounder Joe Gebbia building my desk. I had spent three months slowly convincing the fledgling startup that they needed what was called a product manager. Now, Joe encouraged me to settle in and dive deep into the website.

As I scrolled through the listings I noticed something: Those faces in the reviews? Several were walking around the office. My face became flushed. Had I just made the biggest mistake of my career? Did I join a pint-sized marketplace that was running off of the credit cards of its employees? My anxiety was quickly allayed by CEO & cofounder Brian Chesky who assured me that employee reviews only made up a small fraction of the 10,000 reviewed listings. I caught my breath.

What I’d joined, I quickly learned, was a creative team of problem solvers, wholly committed to making this marketplace work. When reviews proved essential to getting bookings, team-members traveled to get the first reviews of many listings. And as the company’s first product manager — hired when it had just eight engineers — I was going to have to think creatively too.

Marketplaces are notoriously hard to start, but amazing if they work. In my six years at Airbnb, I watched it scale over 100X on every metric: listings, users, revenue, employees, and, yes, sometimes complexity. I joined at an inflection point in the business, when the product was shifting from niche to early adopter, from a marketplace for air mattresses to a truly viable accommodation alternative. As the company boomed, we lived through many more inflection points, even existential challenges. But that creative, “we’ll get through this no matter what” attitude always got us through.

Here the five biggest inflection points I witnessed, and the lessons I learned from each:

1. When jumpstarting a marketplace, set super specific goals — then do whatever it takes to get there.

You may have already read the lore about how Brian and Joe initially got the flywheel going in New York: meeting every host in person, photographing host listings, and leveraging a variety of growth hacks. By the time I joined, those efforts had paid off, and New York made up a significant portion of stays. So how could we replicate the success of New York in other markets?

For Airbnb, liquidity on the supply side is crucial. Cofounder Nate Blecharczyk is highly quantitative and had determined that 300 listings, with 100 reviewed listings, was the magic number to see growth take off in a market. Observing New York, Paris, and a few other top markets, we saw a step-function change in the rate of bookings growth at 300 listings, the point at which guests had enough options to find a listing that matched their tastes and their travel dates.

With a clear liquidity goal in mind, we tried everything to get each market — for us, a specific destination like London — to our magic number. Many of these activities were manual efforts like meetups with local hosts, outreach encouraging targeted guests to try hosting, and on-the-ground host referral campaigns. We also latched onto local events that were bigger than us whenever possible. Online campaigns such as “Make $1,000 in one weekend renting your apartment to Oktoberfest attendees” instead of more generic campaigns like “Rent your apartment to strangers” dramatically improved supply-side conversion metrics. And because one of the most powerful ways to bootstrap supply is to guarantee demand, we encouraged employee travel to unreviewed listings.

Our marketplace was global from day one, but there were numerous benefits to targeting specific geographies over a fixed period of time. We looked for tactics that showed promise, then figured out how to double down and get more sophisticated and scalable. There’s no single technique to credit with Airbnb’s early market growth; the sum of all of these strategies was greater than the parts. And your early course will rarely be linear. At one point, we were ramping up 50 markets simultaneously, and clearly spread too thin. After a moment of truth, we scaled back to a maximum of ten markets at a time.

Years — and over 4 million listings — later, Airbnb doesn’t need to meet hosts in person anymore; the flywheel is officially spinning. But this bootstrapping mentality became a vital part of our DNA. Marketplaces demand creative and scrappy efforts from the start. Make it part of your culture from the get-go.

2. Be willing to take a risk to serve your community.

In July 2011, less than 48 hours after it happened, the story of a host’s trashed home was on the front page of the business sections of the New York Times, Wall Street Journal, and Financial Times. It created a media firestorm with many predicting the end of Airbnb. We had never encountered such an existential crisis. Advisors told us that we were merely a marketplace and should not get involved in dealings between a guest and host. Our founders fundamentally disagreed. They wanted to restore trust to the community.

Brian came to me with a crazy challenge: find insurance for our hosts or else we might not survive the next week! We were in a freefall; bookings started to materially decline. I talked to over 20 underwriters and was introduced to board members of the largest insurance companies in the world. No one was willing to go out on a limb for a new type of business and underwrite our risk.

Three days into the crisis, Brian and I huddled in the office at 1 a.m. with employees sleeping on air mattresses around us. I was out of options. The time had come to take a leap, and I proposed a $5,000 guarantee, backed by Airbnb itself with no insurance to backstop us. Brian, with advice from Marc Andreessen, said, “Can we go bigger?” Brian didn’t want me to edit my imagination even in this stressful time. I tried to calculate all the risk scenarios in my head. But the decision suddenly seemed simple: the risk of doing too little was greater than the risk of doing too much.

We decided that $50,000 was our number. I quickly drafted the ten key markets where we’d launch and what we’d cover (theft and vandalism), then started frantically dialing our outside counsel for his approval. Brian wanted to make the announcement in a few hours. On August 1, 2011, we announced the $50,000 Host Guarantee. Assembling a small team of five, we spent the next two weeks fully operationalizing a program that involved legal, marketing, product, and operations.

On August 15, 2011, Host Guarantee launched — and I braced myself for the thousands of fraudulent claims I feared we would get, taking down the entire business. The total exposure risk was over $500 million in potential losses. After getting some sleep, I analyzed every claim. Nothing was blowing up. The announcement had put out the fire, and more importantly, we felt proud of what we had offered to our community. It felt as though we had pulled off the impossible. A few months later, I finally found an underwriter in London that increased the Host Guarantee coverage to $1 million.

As part of our efforts to build trust, we launched a 24/7 phone support line along with the Host Guarantee. There I am on the home page, “answering the phone,” on August 22, 2011.

Our risk paid off. Listing growth accelerated faster than ever after we established more trust, and Host Guarantee became the cornerstone of Airbnb’s relationship to the host community. We also disproved an assumption about marketplaces: it turns out they can effectively step in to provide support when one side gets demonstrably hurt — even when it costs the business a little — due to the infrequency of these negative events.

So take seemingly extreme risks for your users when necessary. Believe in the good of people, not that the worst case will always happen. When you lean into fostering trust, your users will in turn have trust in you. And that is the most important thing a marketplace can build.

3. When competition appears, move faster than you think you possibly can.

No sooner had we learned to build supply than competition came knocking on our door in the form of a European copycat. (And when I say copycat, I mean it. We even found our job reqs on their site — some with Airbnb still listed as employer!) The competitor had just received $90 million in funding and ramped up to 400 employees in two months. Our guests use Airbnb to travel and if we lost Europe, we really wouldn’t be in the business of “travel” anymore. Growing on our own timetable was officially out the window.

We needed to change tactics and go on an all-out blitz to capture as much of the European market as we could. First and foremost, that meant we had to be “local.” It was September 2011, and our site was woefully US-centric. In just over three months, I purchased top-level domains for nine countries, assembled a global translation team, contracted with Akamai to reduce site load time, and restructured the UI on every front-facing interaction to support long text strings (crucial when you go from “average” to “durchschnittliche”). We also shifted to a multi-currency payment platform.

Office expansion in early 2012

Logistics in hand, it was time to build supply and demand. We’d already seen the value of having people on the ground, so we opened offices in eight European cities. Brian cared about the types of people we hired focusing on “missionaries” and not “mercenaries”. Emphasizing culture, we brought in employees who cared about and who wanted to grow the community. We also bought Accoleo and Crashpadder, two smaller players in Europe, to seed supply in Germany and the UK. On the demand side, we had to think broader, with potential travelers coming from all over Europe. In January 2012, to officially launch in Europe, we did a PR blitz to drum up as much press as possible with campaigns like “Rent the Country of Liechtenstein on Airbnb.”

This pace might sound too fast — and it was. The night before the international launch, we were still trying to get the homepage translated! And on the home front, we were undeniably neglecting operational challenges. But it was worth it. When competition comes after you, move ridiculously fast. Marketplaces are normally winner-take-all markets. If we had lost ground to European competitors in 2012, we may have never gotten it back.

4. When it comes to edge cases, understand your tipping point (because they’ll bury you at scale!).

As we moved into peak summer travel of 2012, organic growth was kicking into high gear — and operations were getting strained. We had 200,000 listings now. Edge cases that had once happened a few times a day were now happening 50 or 100 times a day. Reservation cancellations (and alterations, refunds, resolutions between guests and hosts) were all still handled by email, then manually implemented by a customer support rep. At one point, we had more customer contacts than actual reservations, insanely low efficiency.

This wasn’t a surprise. We knew inefficiencies existed, but we’d kicked the can down the road in favor of critical initiatives like the Host Guarantee and European expansion. We’d finally reached a tipping point, though: we needed to move fast to automate the most burdensome processes, or we would have to hire over 1,000 people that year. The cost of this second option would not only affect the bottom line, but also the complexity and culture of the organization. Instead, we focused on operational efficiency.

Everyone I talked to had a different problem to solve. I had to be a ruthless prioritizer, and developed strict evaluation criteria:

  • How often is this happening? (number of contacts)
  • How many users are impacted? (number of users)
  • How important is this to the user? (severity of the contact)
  • Can manual efforts continue to address this? (complexity of solution)
  • How many resources would it take to solve? (engineering hours)
  • Can we leverage the work on this solution and apply it to other problems? (impact of solution)

Framework in place, it became a matter of moving down the list. Which reservations were associated with the most customer contacts, and why? Was one customer initiating a lot of contacts about one difficult problem? That, while certainly not ideal, wasn’t a priority for automation. If we were hearing from a lot of people, how important was the issue? We addressed these cases one by one, taking big wins at the beginning and working our way down.

How do you know, though, when to pull resources away from other growth initiatives to address these edge cases? My rule of thumb was when a problem was occurring at least 50 times a day, it was time to solve it more holistically. At a time when we were growing anywhere from 300%–600% per year — and edge cases were growing at least as fast — that’s when the potential explosion of problems proliferated.

It’s tempting to think, “I’ll just tackle edge cases as they arise — get ahead of the problem.” But use caution. If you don’t solve problems manually first, you could lose crucial insights into what’s causing a problem and why. Moreover, it may not be the best use of your time early on. We knew that our cancellation flow was flawed, for example. But we also knew winning Europe was more important.

Wait too long to automate, though, and problems grow so enormous they’re harder to fix. And if you don’t have your basic processes well in hand when you take the next big plunge — a next-generation product, for example — the complexity could be defeating. In the early days of a marketplace, there’s always a struggle between growth plans and mundane maintenance. Keep an eye on your edge cases so you’ll know when it’s time to pull resources away from expansion to clean things up closer to home.

5. When the fires are out, don’t rest. Focus on streamlining the user experience.

Eventually, Airbnb moved out of triage mode. But we didn’t let our guard down — we turned our focus to studying how people were using the product, and what we could do better.

In 2014, we took a deep dive into our data around negative booking experiences. And one clear bottleneck emerged: for a guest, the worst experience was getting no response to a booking request. In fact, they would try again less than a quarter of the time, far less than if the host had just said no. We were losing guests to a false impression that there was no liquidity in the market for them. For years, we’d offered an alternate, but largely unused, booking model called Instant Book. We were initially against promoting it because we wanted to give hosts flexibility but realized that the hosts who implemented Instant Book engaged with the product more. It was time to dust it off.

First, though, we needed to change the behavior of hosts in two ways, one psychological and one tactical. First, we needed to convince our hosts that guests accepted through Instant Book would be safe. We began building Verified ID, the product we now use to confirm guest identification in greater fidelity. Using measures including ID scanning, name match to an offline piece of information, and fraud detection machine learning models, we built more trust in the platform, providing confidence to hosts that their guests would be on good behavior.

On a practical level, we needed to convince hosts to update their calendars. It sounds trivial, but we discovered that a huge number of people really hate dealing with their calendars. Hosts were now more active on mobile than on the web, so we met them where they were, completely redesigning the mobile calendaring experience and launching notifications to remind hosts to continually manage availability.

With these pieces in place, we educated our hosts on the benefits of Instant Book to nudge adoption up. Slowly but surely, they realized that it was substantially better for them, as they booked more stays with less management. We started to see a significant migration to Instant Book, from single-digit percentages to more than half of hosts using the feature.

Two and a half years later, 2 million listings have Instant Book and overall booking conversion has increased by more than 60%. An existing feature, which we finally decided to pay attention to, generated more growth than any other feature to date! Building something isn’t always enough — you need to understand the emotional and operational needs of your users to get the most out of a feature. Indeed, sometimes the heaviest lifting in product development is telling a story for your users, showing them a better path forward.

Closing Thoughts

I’ve learned something at every step of my Airbnb journey, but the crunch times have been the most educational. Don’t fear them, and know that you will get through them. Airbnb’s inflection points are when the team discovered its ethos and when we created unique solutions to seemingly immense challenges. Airbnb is a wildly collaborative place that has pushed my thinking about what is possible, working backward from the ideal solution. I’ve come to realize that the far out ideas are possible, it just takes dedication to solve any problem that comes your way. Tackling the right problems at the right time ultimately determines the growth trajectory of a marketplace, and the most crucial contribution of any startup leadership team is understanding what problems to address and when. So always take lessons from former and current marketplace successes like Airbnb, but find your own path too.

When I joined I couldn’t have imagined that Airbnb would have the impact on the world that it does today. I am so excited for what the future holds.

Follow me on twitter @jpgg.



Jonathan Golden

Partner @NEA, Previously Product @Airbnb, @Dropbox and @Hubspot