Scaling Growth | Gustaf Alstromer, YC Partner (formerly Airbnb) & Ed Baker (formerly Uber)

What's pretty cool is a few guys who have been living the centre of building up these growth teams kind of for the past, you know, seven or eight years. Edie joined Uber to start the growth team when it was five people, and then over the three and a half years became the VP of both product and growth for kind of all of Uber. Gustav joined the Airbnb growth team; it was him and two other people when he got started a few years ago, and that has scaled to over a hundred.

So, it's fun at a scale-up offsite to talk about scaling. We mostly have been talking about scaling organizations. There's a chance to talk about how do you scale the organization that continues to scale your users. You know what's fun for me is, and I've known Ed and Gustav now for most of a decade kind of in this growth world. We all started back when sort of this idea of a growth team, this idea of a team that's focused on this sort of intersection of product and marketing to build stuff, was pretty novel and almost didn't even really exist.

We were all kind of doing it in a social world. I remember Ed started a company in 2008 called, you know, Friendly. Jimmy Go! And then Friendly, you know, Gustav started a company in 2007 called Haste On and then went over to Boxer. These were like how do we get to millions of users in the early days? So, I'm curious if you just kind of talk about this heyday of social viral growth explosion and sort of what are some of the key crazy things you learned then that have really helped you as you think about growth today.

I don't want to start? Sure. So, yeah, I guess summer of 2007, when Facebook launched the platform, was I think part of that explosion of just crazy, viral growth spammy apps on the Facebook platform. Those were the first apps I built that reached tens of millions of users, you know, within months' time. I think it was a great platform on which to just experiment with a lot of different ideas because it was so easy to build these simple Facebook apps and just run lots of A/B tests.

So much of, I think, figuring out how to grow something is a combination of an art and a science. By running lots of experiments, you start to build an intuition for the types of things that work and don’t.

Okay, what were some of the early tricks you found that made you go super viral? You were pretty famous back in those days.

Yeah, I mean, you were at Facebook back then, right? So, I think Facebook created new rules after I found some loopholes.

Oh yes! Yeah, there was one, I remember. I think the most viral Facebook app, I'm hesitant to even admit this, but it was called Zodiac Photo Album. The way it worked was literally with one click, it would create a photo album with 12 photos in it, one for each zodiac sign, and then show all of your friends who had that zodiac sign in each of the 12 photos and tag all of them. I think we were allowed to tag like 20 people per photo, so it would show up to 20 of your friends per zodiac sign times 12.

So, with one click, you tagged 240 friends, and all 240 people would get an email saying, ‘I've just tagged you in a photo!’ So, that has a very high response rate. People kind of want to see when they're tagged in the photo. They would go, they'd see, ‘oh, it's a Zodiac photo album. Create your own at, you know, the Zodiac app.’

Within, I think, within 24 hours we had tagged tens of millions of people in photos, including Zuck. And then I heard the platform team shut it down shortly. Now the wonder of those days are over, of kind of those crazy viral pops.

Well, and then Gustav, you were really central to a lot of the early experimentation on mobile. I remember we first met when Boxer was blowing up, and you guys had figured out some really unique ways to blow up using notifications and inviting. Can you kind of remind us, like, well how did you get to this world of growth, and how did you kind of figure out stuff in those early days, especially on mobile, which people think is a lot harder?

Yeah, and social was so—before Boxer, I was at a company called Hey Sam, which I co-founded. It was a product, like a messaging product on mobile. We couldn't grow, so I spent a lot of time trying to learn how to grow. When I joined Boxer a couple of years later, the first kind of approach and first thing I kept in mind is like we need to think about growth. At the time, so back to the earlier questions, I think there are a couple of like growth kind of tape that becomes important.

One, you have new platforms. You have the Facebook platform, you have the mobile platform, you have Google SEO. There are a bunch of larger platforms that makes people actually care about investing in them. At Boxer, the iOS platform was key; those things were the address book and push notifications. Those are things that kind of like warranted their very beginning. Once they start happening, people started exploring what you can do with those things.

With Boxer, when I joined, the most important things about the walkie-talkie app on iOS and Android, you push a button and you start speaking into the other end. We found that the amount of friends that you have when you open the app was the very first time on the app, and how you display them was absolutely key to whether they were going to use the app or not.

So we spent basically the first year optimizing that—those things, like basically how many people do you have when you open up the app the first time, and how do you get to a list like that right when you open the app? Well, you match phone numbers with address books. Spent a lot of time investing in matching phone numbers with address books, not that different from Facebook platform.

We kind of tried to optimize how many address books had matching phone numbers for different people, and we were doing this in the best interests of the users. Like, you wouldn't end up using the app unless you actually saw friends on the app, so that was basically the goal. This was all in the best interest of the users, and sometimes you forget like why you do things.

So, and I remember your company got acquired by Facebook, and you kind of jumped into a little bit more mature growth team at Facebook. You've been kind of virally hacking and building these things on the platform. What was sort of the adjustment to go from kind of being in a sort of experimentation mode to something that was a little bit more mature? What did you see when you observed kind of that mature Facebook growth team?

Yeah, so, well one thing I saw was I was finally working at a company that had an amazing product that people loved, and I just never actually built that myself before. So that was pretty cool, and with that came amazing retention, amazing engagement. Facebook was already at a size where retention and engagement were bigger drivers of the Northstar metric they used than new user acquisition.

So kind of everything I had worked on in the past was all about how do you get those first million users? At Facebook, it was how do we keep the users we've already acquired, keep them engaged, minimize churn?

To me, that was one of the biggest differences. Most of the growth team at Facebook at the time was actually focused on friending because we discovered at Facebook that friending was the number one driver of retention and engagement. There were teams on the Facebook growth team that were basically focused on things like Pym Piquet, people you may know, and other ways of causing people to form more connections.

So I think that was one big takeaway. The other one, we could talk about it later, is just the way the team was structured. I know you got into that a little bit across their different functions as the team grows. Actually, can you walk through that?

I was actually going to think about how that was structured then. I want to get into this transition to these sort of different real world companies.

Sure, sure, the real world companies. Yes, well, so one thing I learned at Facebook was I really liked how the growth team there was structured with all functions. So, product, engineering, analytics, design, and even marketing; those five functions were all critical to working together and, again, all within that growth org.

When I was at Facebook, I think the growth team was about around 500 people or so. Oh wow! But across all those different functions, in fact, all of analytics back then was part of the growth team. I took a lot of those learnings when I went over to Uber and started our growth team there.

It was just initially a five-person team, kind of picked one person for each function, but I think that was one of the big takeaways from Facebook.

Well, and you know that's a perfect transition because when I’ve been talking a lot of people about sort of what the real success of Uber and Airbnb—I mean Joe gave a great talk about sort of why Airbnb has become so important in the world—but a lot of people have said it's kind of we took these hotels and transportation, which were sort of old world real businesses, and applied a lot of these digital growth technology and learning to them.

So, kind of talk a little bit about what it was like going into Airbnb and starting the growth team with two people, and Uber and starting the little growth team, and like what did you see there and how did you kind of start that this sort of practice from what you learned before?

I'm sure, so I think the key to that is red blood. Okay, I think the initial idea to have a growth team is that there is some function of the company that really cares about metrics. They're really optimized for the goals. And some time ago, like this was unusual in startups, as most people did not look at metrics when they made presentations and did not experiment.

So instead of making everybody responsible that you made one team responsible for that, and that ended up being the growth team. I think really great companies today, they have everyone kind of care about metrics, using experimentation, and doing all those things that growth teams do to make decisions. So, I think things have changed a little bit.

And the role of the growth team has kind of changed from being that evangelizing team to being the ones that analyze and wrestle the company, so everyone to drop a new product in a very similar way.

I'd say at Uber there were kind of three big differences from Facebook when I went over there in terms of forming our initial growth team. So the first was, it's a two-sided marketplace. And when we came up with our Northstar metric, I remember having conversations with that small team of five, one of whom Mike Tao is sitting in the audience.

Can you define Northstar metric? You've used that a couple of times; that's actually really important for sort of the growth discussion.

Definitely! So, at Facebook, it was monthly active users. It's basically the one metric that, if you talk to— the way I like to think about it is if I talk to anyone on the growth team and ask them what number are we trying to grow, they would be able to say it's that number. And if they're not working on something that ultimately is growing that number, they're probably working on the wrong thing.

So that’s kind of how I like to think about it, and it's helpful to just set goals. You know, we've said every half, we kind of have this is our six-month goal, but then we track our progress every week and see are we on track. Of course, there's seasonality and stuff; it's not just a straight line, but you can kind of see like how are we doing versus our goal.

So at Uber, when we came up with our Northstar metric, we had to consider there are two sides of the marketplace. So it's not just monthly active riders. Ultimately, we decided on trips—weekly trips—as the Northstar metric because for every trip to take place, you need riders and you need drivers.

And so, by having trips as that Northstar metric, you could then break that down into the supply side and the demand side. You need more riders, more drivers, and you can do growth accounting on each side of the marketplace to say, well, to get more riders, you need more new riders, re-engaged riders, and, you know, minimize churn of your existing riders—and same thing on the driver's side.

So that was one thing, was kind of agreeing on this as our Northstar metric in a two-sided marketplace. So that was a bit different from Facebook.

I'd say a second thing was because Uber is a real-world company, there were several parts of that funnel that you can't just measure online. With Facebook, you can track every click, every pixel, all of that stuff. With Uber, there are steps of that funnel where it involves, you know, a driver going in and getting their vehicle inspected or uploading their documents, and ultimately downloading their iPhone app, going into their car, and doing that first trip.

For me, it was super eye-opening when I did my first trip as a driver. I walked out to my car; I'd done everything I was activated. I walked out to my car, and I walked back to my house because I got so nervous about doing that first trip. And that's when it suddenly hit me. This is why, you know, so many of our first drivers who are activated never do that first trip.

So that real-world component added kind of some extra challenges in terms of one, how do you measure it, and two, what are some things you can do to influence the way people behave in the real world?

And then, see, I said three things and now I mentioned two. So, two-sided marketplace, the real-world component to it, and I’ll come back to the third one.

I remember the different orbit, our metrics.

Well, and so as these teams have scaled now, you know, you talk about data analytics, design, product, sort of internet marketing.

Oh, that was the third! By the way, paid acquisition. That's what I was going to actually ask. So at Facebook, we didn't really do very much paid acquisition. But at Uber, we had an entire performance marketing team.

Well, initially it was a one-person team, but ultimately became a team where we would actually spend money to acquire both riders and drivers. One exacto in our social world—we're all about these organic how many people can we organically get into our products? If we spend a dollar on user acquisition, we always said a dollar is too much. And you know the boards are like, how do you get more free virality?

But I think, you know, both Uber and Airbnb, because they're businesses, people actually pay for. How do you think about paid acquisition differently than this sort of organic exhibition? How do you train a team to sort of understand that and think about that differently?

Do you want to start, Gustav?

Oh sure! So, let's see. I would say a lot of those platforms that you are doing growth on, Google, Facebook—they are turning kind of the opportunity of doing free growth there is getting smaller, and the opportunity of doing paid growth better is getting larger.

Most good growth teams should invest in paid marketing. I think that— Mike’s plans are seeing looking at other companies, what they've been doing, large companies do paid marketing that tend to integrate that with the product team and engineering teams. So there are engineers and product managers supporting those teams. They call it a tech person like that, and I'm sure what the Uber name is at that stage right now, where basically most of the leverage you'll have at that scale is going to be through engineering, data science, and product management.

That's a different world than when you start off. When you start off, you have a smaller, smaller team, and yeah. It's probably, yeah, I need somewhere.

I'd say the way I think about paid acquisition is it can just further accelerate growth. So Uber would have kept growing even if we had never done any paid acquisition, but especially in a competitive market it kind of makes sense to pay up to what the rider or driver is ultimately going to be worth when you're in a situation where the faster you grow, the better because as you get bigger and bigger, there are a lot of other efficiencies that kick in.

So the whole network becomes more efficient once you have more liquidity, so there's value to getting there as quickly as possible. It's kind of tricky to know exactly how to calculate the LTV of a rider or a driver in a two-sided marketplace because, you know, the riders pay the drivers and some of that money goes to Uber.

So how much is the rider worth versus the driver worth? And we had a team actually focused on that exact question—a team of data scientists—and it was more complicated than I understood. But basically, there are certain markets that are more supply constrained, or a driver, the incremental driver is worth a lot more. In other markets at a certain time here, it might be more demand constrained, where you might be willing to pay more for the rider, and that varies over time and geography.

A quick comment on that I would say is most companies that grow really, really large—they do it on figuring out one platform. So, they are really good at online marketing, they are really good at SEO, a little bit of a rally, or the Facebook platform or something like that.

It's rare that you have like a figure out all the platforms. So I would think about the proxy as someone who has the product that you have and which of these platforms is it the most suited to. Is it most suited to search? Is it most suited to display? Is it the most suited to tube rallies? It really depends on the park, I would say.

Interesting! And where do you guys see any upside right now in sort of the marketing or growth world? I mean, it seems like a lot of these channels have gotten pretty saturated. Facebook has gotten relatively expensive to be buying on. Google's various platforms—I've heard a lot of people talk about video ads or sort of the new place. Even people are actually going and advertising on TV. You know, where do you guys think about sort of where those growth opportunities sit out there now?

I mean, I come back to build a great product because if you have that product-market fit and the people using your product love it, then all of the channels are out there today that will allow people who love a product to share it with their friends. So, I think that's actually the most important thing, and then everything else will follow.

Yeah, I mean, I would—I don't think there's anything short-term that will remain like successful for a long time. So, the best advice is to invest for the long-term—whether it's SEO, whether it's paid marketing, whether it's a store's quality—take like a couple of years, look out, and see and I want to be a name for, like, look for channels that actually do have to scale, like channels that have like hundreds of millions of people discovering products through those channels.

And then invest for the long-term. Engineering and data science are critical to be successful, in my opinion.

I’m surprised neither of you mentioned Instagram. The tip I got yesterday at F8 talking about Facebook focuses. I think Instagram app install ads are significantly underpriced relative to Facebook. So you guys go back and start buying on Instagram; you might be so surprised.

How does—so one of the unique things, again, in the social viral world, we kind of grew and scaled on the internet. You guys also have to grow city by city. What are some of the things that you did to kind of set up teams for growth that could actually do growth in a city that could sort of customize it for the city, or how much did you try to make it sort of a centralized playbook that you'd control from the mothership?

So, I mean, every city would have its own city team beginning with a launcher. Again, I mentioned Mike in the audience. He was a launcher then, was our general manager in Boston, and then started basically headed up product for the growth team.

So it was really valuable to have the person who was running product for growth be one of these ex-operations people, and working with all the city teams was really important in terms of just making sure that product and operations were integrated.

And Uber, we now have a team at Uber called product ops, and it's basically a team that kind of is that bridge between the product team and the operations team. But as much as possible, we tried to just have a playbook. One thing that's helpful with Uber being a global business, and I'm sure you guys see it as you as much if not more so at Airbnb, is when we would launch in a new city, there'd also be a lot of pent-up demand.

We'd see a lot of people trying to use Uber in a city before we launched in that city, so there were certain things we could do to kick off that flywheel. But I'll share one example of something where we actually did something very different to the product in another city.

And this was back when we were in China. We found that it was hard for drivers to sign up to become drivers in China because of the firewall. There were like latency issues and just like signing up online. A lot of drivers had phones, but didn't have computers, and like it was just hard for them to sign up.

So, one of the engineers on our China growth team built a WeChat bot. WeChat, you know, like a messaging app or the Facebook of—sorry, the Facebook and also the Internet of China, really. And so we built this WeChat bot where as a driver, you could take a picture of a QR code and then it would just start asking you questions—what's your name, what's your birth date, what's your driver's license number?

And people would quickly be able to sign up to become a driver. We'd have these onboarding sessions with hundreds of drivers, if not even more than that showing up in one room, all on WeChat signing up, which was amazing until WeChat started blocking us because they were owned by Tencent, which had invested in our competitor.

So, that's one example of something we had to do very differently in a different market. I would say the two things that would call out—one is translation—it’s absolutely key to have a way that you can translate content across languages over no profit for over books, and it's like a huge growth driver. A lot of people don’t speak English and need the experience that you’re offering in different languages.

The second thing is what I call good product gaps. So, if you look at your product globally, there are some parts of the product that won’t work in some countries; whether it's authentication, like Facebook connect, whether it’s payments mechanisms, search, SEO—just identify those gaps. They're not that many countries, in my experience, that are dramatically different. There may be five or ten that are quite different than us, but most countries are not that different.

Cool! And just lastly, because growth is all about experimentation and often experimentation against instincts, what is one great case where you and your growth team like really either believed something was going to make a huge impact and then were disappointed when it didn’t, or where someone just did an experiment and even you, you were like, oh, this is never going to work, and then it like blew you away?

So, we have actually instituted—institutionalized this thing I would call it the experiment review. Every few weeks, the growth team gets into a room, and we show experiments that we've shipped and gotten significant results on in the last month, say, and the engineer will go on stage, shield experiments, show the control and the experiment. Before they show the answer, they'll ask the audience, ‘What do you guys think? I'm going to hear things, control. How many think experience, experiment?’

And it turns out that we often disagree. Most of the times we disagree, which is exactly the point you're trying to make by doing experimentation. That predicting product decisions is really tricky. So I would say this happens all the time, and if it doesn't happen, then you're not taking big enough risks.

You should be seeing things that are counterintuitive sometimes when they're very intuitive. Having experience of seeing likes, permutations makes you ask the right questions. Oh, maybe the metrics is wrong this time. That happens as well, but yeah things should be really counterintuitive.

Give one to bid and one to advise—if you have a website and you had people come to your website and you have accounts, don't log them out. Like, keep people logged in for as long as you can. There's really no benefit in automatically logging people out after a couple of months.

I've had a couple of days in terms of growth. So, because I would— I would look for Amazon for some advice in what they have done. So, I mean, one thing I found interesting is, often it's the smallest tweaks that can make the biggest impact. There’s not necessarily correlation between how much work you put in and how much of a step change it makes, or sometimes an inverse correlation.

So I’ll share one interesting example from Facebook. We were looking at the sign-up, basically the funnel, and the viral metrics by region. We noticed that Japan was our least viral country, and when we dug in and tried to understand why, we saw that very few new users were actually sending invites to any of their friends.

We had like a small team in Japan; I spoke with a country manager there. He said, you know, people just think it's rude to send invites, so no one's going to—it’s like a cultural thing. But people don't want to intrude on their friends and send invites.

So we brainstormed a bit, and we decided to literally make a copy tweak. Instead of calling them invites, we called them announcements, and we would say as part of the, you know, sign-up flow you'd import your address book to find your friends, and then instead of saying invite all your friends, we said, let’s Facebook now announce to all of your friends that you’re on Facebook, and this is where they can find you.

And Japan went from like our least viral country to our most viral country, but it was literally just a copy tweak to make people more comfortable with sharing what the product they were using with their friends.

That's great! I've been talking to some mobile developers that even the difference of invite versus add somebody, which is sort of vague, or if you press—if you change your language from invite to add, people will send the SMS out to their friends to invite them to a service at a much higher rate than the word invite.

So these really, you know, subtle things—they don’t expect, they have a time for like one or two questions from the audience if anybody has.

Okay, and you talked about the Northstar, and you said it over— it was about the number of rides. But you could have easily been optimizing for crazy retention at zero growth and new user growth, or new rider growth, or all new riders growth and to news last comment, like horrible retention.

So how did you balance out knowing what your almost subs Northstars were so that you could have the best north star?

That makes a lot of sense, and it actually changed over time. So early on, we were kind of in land-grab mode, where it was all about new user acquisition—new riders, new drivers. But we would look at growth accounting, so you'd say, well, the total number of active riders or active drivers equals last week's active number plus any new users you get minus an equal users that turned plus any users that re-engaged.

And you can look at each of those three things, kind of the magnitude of each, and early on in the growth curve when you're kind of at the bottom of that S-curve, the new user acquisition number is probably going to be much larger than the churn or the reengagement.

And so it makes sense to focus the most on growing that number, but as you get bigger and bigger—as your active user base gets larger—the churn as an absolute number goes up because it tends to remain like, you know, constant percentage or roughly like constant percentage of the absolute number.

And then at the same time, new user acquisition gets more difficult because you've already reached out to so many new people. So, if you just monitor, look at the growth accounting, and kind of see like those are the levers which ones can have the biggest impact. Early on, it might be new user acquisition, but later on, it's probably going to be retention and re-engagement.

But that said, you kind of need that retention from the beginning because if you're just like—I’ve run plenty of things in the past that haven't had retention, and it's fun while it's going like this, but if that retention is not there, ultimately, it's not really going to help that like Zodiac calendars.

Exactly! That’s that kind of stuff. You earlier told a story of like the challenge of getting people to go from registering as a driver to actually like overcoming the nervousness to go out and do the first trip as a driver.

Interested in what are some of the experiments you ran there, what you learned about how to actually get people to do that?

Yeah, definitely! So, you know, I'll just mention one example is we saw that sometimes like a driver, a new driver, would log in and might not get a trip request right away. Sometimes that driver might just be waiting for a really long time because they're a brand-new driver, and maybe they're waiting somewhere—maybe they are waiting in front of their house and there just isn't them, but a bunch of stuff happening right there.

Or there’s another driver who's more experienced and who knows if you wait a block away from here, they're more likely to get a request. So, like, one of the experiments we've run was what if we make it like—make sure that that first experience is better?

Are there things we can do to kind of get that driver hooked? So that's one example. You know, we also experimented with are there ways we can get potentially existing drivers or other people to like get on the phone with a driver and talk them through that first trip?

So the team's experimented with a bunch of different things like that.

Quick just to kind of probe on something you alluded to regarding growth and organizationally, do you guys think that the right steady-state outcome is for all teams to become growth-oriented and growth people inside of them, or to have a separate standalone team that's kind of fully vertically integrated?

And if the latter, what conflicts have you seen between say marketing—you know, who's focused on owning the homepage and the language, and the brand kind of goals or lifecycle—who thinks that they under tension versus growth?

And likewise, with product groups that feel like they own a product and a customer experience that they're trying to get with you guys kind of hacking away at that.

So I would say that ultimately, the company becomes a growth game in the sense that you use data to make your decisions. Like you don’t make—you basically look at the data because it gets so hard to predict what are the right decisions to make.

Once you make an optimization, ultimately all of the larger teams are using data to validate experiments, and that's kind of the idea of a growth team for us, means we're using the funnel. You’re on the top of the funnel, you're finding users that are not already using Uber. In other companies, growth teams are a smaller organization that kind of evangelizes that to their sort of product team.

But everybody, really, the entire crack team is now a growth team, and the goal of the data team that had to build these tools is to get everybody to use them. Like they are literally measured by how many people use the experimentation frameworks.

We actually merged the product team and the growth team about a year ago at Uber for largely those reasons. Especially as we started focusing more on engagement and retention within the growth team, and that's kind of what the core product team was working on as well; it just made more sense to bring it all together.

Now, we just added—talk about your lifecycle and your homepage copy, teams and everything and brand—they actually, when they merge together, and they sort of have the orientation where really growth and marketing are sort of one thing as opposed to two.

You understand sort of everything we do is in the spirit of growth, even if it is driving our brand or other stuff, you know, rather than trying to keep them separate and contentious.

Is for them with that merging a good time for one more question? What questions for a super aggressive in China and growing? Were there any really crazy things you guys discovered or hacks there that maybe people could discuss?

People could use back here?

Listen, make sure there were! But I probably can't. I mean, that's actually—I don't have any like super crazy ones that are coming to mind right now, but I probably shouldn't talk about any of the stuff that worked really well about this since we're probably still doing it if it did.

Okay, sorry!

The awesome! Well, thank you! Hopefully, you'll be scaling your orgs and your user bases together. Thank you, Ed, Andrew, and Gustav!
[Applause]