Anu Hariharan on Network Effects

Hey, everyone, today we have Anu Hariharan, and we're going to be talking about network effects. So, Anu, you're a partner at yc. Would you mind giving us some background?

Sure, Happy to. I'm an electrical engineer by training and worked at Qualcomm in San Diego. This was the pre iPhone days when we were trying to design mobile video streaming all the way integrated to the stack. After spending many years at Qualcomm in San Diego, I went to business school, worked at BCG for a bit, and then prior to joining yc, I was at Andreessen Horowitz, focused on consumer tech investments. So had the opportunity to work with a lot of great YC companies like Instacart, Airbnb and a few others.

Awesome. You also published this network effect deck at Andreessen Horowitz. So you've kind of become an expert in the field. So just to start out, can you define what a network effect is?

Sure. I think the simple definition of network effect is basically as more users use the product or service, it's more valuable to the people using it. So, simple example, Facebook. The more friends that joined Facebook, the more powerful the network became because it was easier for you as a user to connect with more friends and exchange messages. And now you see the news feed and so you see a lot of content shared. So the more users that join Facebook, the more powerful it is, even for the existing user base.

Cool. And then there are a couple terms thrown around in that space as well. So that virality and growth. Could you define each of those as well?

Yes, I think. See, I mean, if you take a step back, growth is what is confused with network, network effect. And it's very easy to confuse the two, because when you are a startup and if you have a really valuable product that provides inherent value, you're growing really fast. And so what happens is the. The common misconception is you think growth is equal to network. There are different forms of growth, and that's where virality comes into the picture. I think virality is actually about the speed of growth. It's not about the value. What's the difference? It could be word of mouth. In the early days of Uber, for example, when people took the right point to point or arrive at an event, they would rave about the service to other users because it was 10x better than using a taxi, and over time became cheaper. So that helped drive word of mouth growth. And so virality is often about free growth, meaning unpaid growth, which you can drive without marketing because Your product is just much superior to the alternate option available. Network effect, on the other hand, is about value. What does the network offer in terms of its value? So, for example, Facebook, how would you know whether it had a network effect in the early days? It really is about usage, which in other words translates to retention. So it's different from growth. So in the early days of Facebook, even though they were growing quite fast, I would say reasonably fast when compared to apps today, because it was first a web service, the retention improved, meaning if you looked at their daily actives to monthly actives in the early days, as they added more users, that percentage kept improving. And so that retention is the sign of, of network effect because existing users were willing to log in more frequently and be more active because the service was more valuable.

Okay, gotcha. And so what's a good example of virality, like an app that everyone knows about that was viral?

I think a lot of apps out there are viral, if you ask me. Facebook's early days were viral growth. Airbnb didn't have as much virality. I think they did have virality in the sense they had word of mouth growth. But the product itself didn't spread as an organic content of its use. So let me define what that is. So if you look at Facebook, the reason why it grew quite fast was because if you were one user on Facebook, it was not that useful. The early days, the app was launched to see the online student directory. They leveraged the Harvard directory to launch Facebook. But beyond that, it wasn't as useful unless the entire class or the group of students taking that class was on Facebook. So therefore you started inviting other friends who were in the same class. And so the product itself was designed such that it was useful for other users. Same with Skype, right? If you wanted to place a Skype call, you needed someone on the other end. So the product spread as a result of organic consequence of its use. Versus Airbnb, I think, did have vital growth, but it's word of mouth growth because if you look at Airbnb in the early days, it was about renting a room in your house to other guests. So you talked about it because it was less sterile than hotels, especially during events when the hotels were sold out, it was 30 to 80% cheaper than hotels. So you talked about it to your friends and therefore they discovered it. But to actually use Airbnb, you didn't need your other friends on the other end to actually use the app. So I think there's a difference and there are different Forms of growth. And so you need to understand whether your product has an organic consequence of its use, because of which it tends to grow really fast versus word of mouth, referral, anything unpaid attributes to vital growth.

And so at this point, what do you recommend to founders? I mean, are people approaching you asking you, like, I want to build a network effect into my product, what should I do? Like, what do you tell them?

Yeah, I think, you know, if you talk to most founders, even founders that have built really amazing products, I don't think anyone would talk to you in the context of a network effect or viral growth. But the key decisions they took as to what their product should do will help inform how they focused on the network. So this keeps coming up repeatedly as a question from founders. I always go back to the example of Facebook even. I'll also touch upon WhatsApp as an example. When Facebook first launched, they launched in Harvard and it was quite a popular service. They could have easily rolled it out to multiple universities pretty quickly, but they didn't. They actually took a clustered approach. And the reason for that was they wanted to make sure that if you ask the early days, they would say they wanted to make sure that at least 80% of Harvard students were using Facebook and maybe 60 or 70% were using it daily. Why is that important? Because I think if you didn't focus on engagement in the early days, you could end up growing too quickly and lose your older user base. And therefore you're not really focusing on products or improvements that help the overall network. You may end up focusing on features that help you grow in the interim. So I think one of the things Facebook did really well was always pay attention to how the network grew and not just the user group. In fact, in the early days, based on the numbers they've quoted on the S1, if you look at their month on month growth and user base was 16% in today's world in mobile, you might say 16% month on month growth is solid, not great. But I think it's important to focus on are you driving inherent value to your existing core users as you grow? And that's a measure of engagement. In the case of Airbnb, which is a low velocity use case, people don't travel as often, maybe once or twice a year. How do you measure that? The way to measure that is bookings and rebookings, cohort after cohort. And Airbnb did that. That was a measure of do your customers really love you? Because if they do, they will come back and use it more. Often and for more use cases. I think focusing in the early days, especially as you build a product, it's very important to focus on as you grow, how strong your engagement is and how much are your existing users willing to your service.

More and more, what are you seeing now that people are doing that's been effective?

I think these days, especially in recent years, I would say in consumer apps, it's become harder to find the distribution. I think there are two challenges a consumer app should think of. One is, what is my entry point? How am I going to launch this thing? And then two, how am I going to distribute this? When you have powerful networks like Facebook, Twitter, Snapchat, it becomes harder and harder if you launch something as just a feature because these other platforms have the network. So you have to figure out what your distribution strategy is. One app that recently has gained a lot of traction is Musical Ly. And Musical Ly actually pivoted from being an education app. And they first launched as a tool among teens. And the tool was creating short form lip sync videos. And the tools they provided made it really easy for teens to create videos. And it was a fun thing to do. It was cool. And so people talked about it. And so suddenly a lot of students in school started using it. But for the first 14 months or so, even after they launched Musical Ly, it was sort of a slow growth. And I think the founder, Alex has talked about this. They were almost about to run out of cash. And then what they did is they realized that a lot of the Musical Ly videos were being shared on Instagram and Twitter. So it was being used as a tool. So to sort of shape it towards a one they made. One of the things they tried was to make their logo more prominent so that when it was shared on Instagram and Twitter, people knew about Musical Ly and downloaded it. A lot of people created lip sync videos, but they were clearly top users. The people who had a lot of fan base and following and were really good at creating these lip sync videos. They created a Q and a sort of platform on the app for the followers to interact with them. And so that helped shift some of their audience back into the app. And they started focusing on features that helped create the community and network on the app itself. It's still early to say if it will become a network, but I think these are some of the things they've tried to at least form a community. And now you have more users going directly to Musical Ly.

And so if I'm just getting started, how Do I know that when is the time that I can start trying to pull people over from other platforms? Because it seems like you can piggyback onto these platforms to a certain extent, but at some point to create lasting value for your company, you need to have people that you have in your community.

Absolutely. So I think this is where the network effect definition is important. I think people often say, oh, this is where people get confused about does this company or does the product have really a network effect or not? I think the two quick ways to really test for that is barriers to entry. This is your competitive mode as a product and barriers to exit for the user. I think the second one is the one people usually miss out on. So barriers to entry means you are growing really fast, your product provides inherent value. There are some demand side economies of scale, because of which it's not easy for someone to come and disrupt your network overnight. The barriers to exit for the user is you've built such strong connections and you have a lot of value from this network that it's not easy for you to switch as a user. So, simple example, Facebook and Google. Google actually made it really easy with the circles to import. All your contacts have the same graph, even help break down your friends. Some people have thousands of friends on Facebook. How do you break it down into closed circles versus these circles? You probably touch base every now and then and so on. But it was so hard for users to switch, let alone what the companies tried. As a Facebook user, you found it so hard to switch because you have all these connections, the communication that you've had with these users, you have the news feed set up. There's a lot of things and a lot of value you're already deriving from the platform that it was absolutely hard for a user to switch.

So do these platforms end up being winner take all?

I think network, if you're a business with network effects, you are a winner take all. And it actually helps you tip towards becoming winner take all businesses. And this is why one of the focuses sometimes investors look at is, is your app a destination app and that's a sign for barriers to exit for the user. If more and more users are coming to your app directly and spending a lot of time, and time spent per user is going up. That's a sign that the barriers to exit for the, the user is going.

Up as well, then what ends up happening with Uber and Lyft at this point if it's winner take all?

Yeah. So the right sharing has always been a little bit of a Controversial topic in terms of network effects. And I think you have one school of thought in Silicon Valley and even a broader ecosystem who think that they have very strong network effects. And on the other spectrum of people, they think that they don't have a strong network effect. My personal view is, is you have to break it down in terms of what the product does, what service they offer, and what sort of moat each company offers. I tend to think that the basic point to point ride sharing service does not have a very strong network effect. And the simple reason is it's a point to point write. So as a writer, beyond taking me from point A to point B, I don't have any other value trapped in it. So if there is another service that is equally good and provides the same value, then you are willing to switch. So the barrier to exit for the rider is not as high. So does that mean they don't have any network effect? That's not true. I think that if you look in a city by city basis, they do have some form of network effect, but I think of it as a weak network effect. So what is that meaning? If you have many drivers and riders on the platform, for the rider the promise is that the ETA is less than five minutes. So you know for a fact that you are going to get a ride and there is going to be a driver that's going to pick up in less than five minutes. And for the driver it's the dollar they make per hour. Right? But the problem is beyond that five minutes there's no other connection. So if there's another service that can provide you something in less than 5min, you are willing to switch. Right? And you see people switching because of surge pricing every now and then. However, having said that, can this still be a winner take all market? I think that this companies that have economies of scale, supply side economies of scale, can still have monopolistic share. And that's what I think is playing out in the right sharing market. So two reasons people use Uber or Lyft is 10 times better than taxi and cheaper than a taxi ride. Now the cheaper than a taxi ride, I think this is where you see economies of scale. So the more drivers that are in the platform and the more riders that are in the platform, the cost of unit output that an Uber incurs is probably much lower simply because, you know, if there is a lot of demand on the platform, then the drivers are doing more trips per hour. So therefore they make more dollars per hour and therefore Uber can, you know, charge less per ride for the rider.

So right now it's kind of like a race to like raise all the money. It seems like when self driving cars happen, is it. Do you think it's a race to reach self driving cars? Because that unlocks the network effect ultimately for these companies and then they, then they're the monopoly.

I don't know if you need self driving to be a monopoly, to be honest, because I think that even in the current market dynamics you are seeing, depending on the market, who was first to come or who has executed really well has economies of scale because of which they'll have monopolistic share. Simply because I think the trip efficiency is a very key metric to focus on because that drives cost. Right. I mean you've seen a lot of Uber for X services that have sprung up, but I think that Uber might likely win most of the Uber Forex services simply because they can drive delivery efficiency unlike no one. And if they are able to do two or three trips or four trips an hour, then the cost per trip is so low that they don't have to charge the rider as much and so they can pass on those savings to the rider. So I mean this is exactly what Amazon first party was. This is not new. When Amazon launched their service many years ago, more than a decade ago, it was all first party. And first party is all about economies of scale. Shared warehouses, shared shipping. They made the cost of delivery of each good a lot cheaper. Today they have a big marketplace, so they have both demand side and supply side economies of scale. But Amazon first parties primarily supply side economies of scale and yet they have monopolistic share. So I think it's people tend to focus too much sometimes on network effects and think that's the only winner. Take off businesses. I think that companies with supply side economies of scale especially which have huge operational challenges and who execute really well, can have monopolistic share too.

And are there any other markets right now that you think? Many people consider it to be a network effect, but in reality it's not that strong of a network effect.

I think on demand has been the biggest debate. I think because it's highly low cost, it's not easy to replicate it across, even nationally or globally. So it remains to be seen as to how these networks evolve. Each of them do provide some value, so I don't want to discount that they don't. But it may not be as strong as you saw in social networks.

So then what about messaging? I know you wanted to talk about WhatsApp a little bit. Do you see Them continuing to grow in the US or do you see some of these larger international players coming in and having a network effect here?

I think that messaging has a very strong network effect. It goes back to the simple thing of barriers to exit for the user. I have been a WhatsApp user for very long because I have a lot of family in India. It is the cheapest form of communication. Especially it was a great replacement for SMS. They also had, I think in 2009 when they first launched, they launched it as a status feature. I don't know how many of the readers know that, but it was literally a status feature to let others know that, hey, if I'm in the gym, don't call me or I'm busy in a meeting, don't call me. That didn't take off. But the things that, I mean they still have the status feature, but I think that that things they learned in the status feature really helped them evolve towards a messaging application. Meaning it was one of the first apps at the time that used phone number and address book, did not use a username. And it was very. Today it seems like, well, that's the way to go. But in 2009 when most of them were all about creating user accounts, it was very contrary to how they approached the sign in process. And WhatsApp did not quite a lot of the users asked them to change it to a user sign up, but they decided not to because they felt if you're leveraging the address book, you would know who that person is.

So that's something I always wonder, are all those things that could be considered growth, hacky type things, do they amount to marginal gains or can you in fact build a really powerful network with that in terms of adding it to your product? Because I think a lot of people end up creating things that will never have a network effect and consistently trying to add these little things.

Yeah. So I think it comes back to the same point, which is if you pay attention to engagement, retention of your users, you will automatically know whether it's a network or not. It is the single strongest sign of a network effect. And I think WhatsApp did pay attention, they paid a lot of attention, even though I don't know the exact metric they used to measure it. But they talk about the fact that they always focused on did their core users continue to use the product and if so, how and how can they drive more value to the users? And in fact that's one of the reasons they moved from status to messaging. I think within six months, almost by the end of summer of 2009, they launched messaging and. And they saw a huge spike in growth because it was an alternate version of sms. It started off with the Russian community in San Jose, communicating with their family in Russia, and then it took off in other international markets. By the end of 2009, they launched MMS, basically a way to share photos and videos. It was too expensive to share a photo or video at the time, and this became a great platform to share photos and videos. So now you have this app that started growing their user base, but users, existing users started using it for more things than just sms. And their product features sort of went along that, which is, am I driving more value to my users as I'm growing my user base and with more connections, people were able to connect with more users and they have a very strong network effect simply because the barrier to to exit for the user is really high. I, as the user have all my communication with my family, photos exchanged and videos. It's not easy for me to jump to another messaging app.

So how do you see this playing out with like all of these international messaging apps becoming. I mean, they're just massive. Do they dominate in one country and that's how it plays out ultimately, or will one eventually, like, be the one international?

Yeah, I think that any app that has a very strong network effect, it's very difficult to disrupt them overnight. This is another reason why companies tend to focus on building a network, because it's not easy. Even if you don't do much, it would be very hard to disrupt that network overnight. And so WhatsApp has a very strong foothold in India. WeChat has a very strong foothold in China. So. So if you have a strong network and people tend to spend more time and have more value from that network, I think you will end up seeing a few players rather than one.

So if I'm a founder and I am building network, what question should I be asking myself?

Yeah, so I think that the first thing I would say is you should probably take a step back and see are you a network or a marketplace or a platform? And I think it's important to understand this three, not to be super academic about, but the reason why you have to understand whether you're one of those three is because the questions you ask are really different depending on who you are. And a network is basically, I consider Facebook a network or WhatsApp a network. It's a group of connected users and you're sharing information and your product adds value and is Trying to increase the value as the network grows. A marketplace is too heterogeneous. So you have two sites. Think of ebay as a marketplace. You have sellers and buyers. Think of Airbnb, you have hosts and guests. And so the question you would ask if you're a marketplace are different from what you would be, a network. And platform is the third component, which is you have users and developers and a platform. And so think of it as where both groups, users and developers, help build the platform. It can be programmatic, customized, but each group reinforces the value of each other.

And who's a good example of that?

Microsoft operating system is a platform. I think WeChat is evolving into a platform. Facebook is a platform. And the interesting thing is you could be all three over time, and you will see different companies do that. But where you start is important because then the questions you ask are depending on where you start. So if you're a network, I think that the number one question you have to ask yourself is, what is your entry strategy? And I think the entry strategy is changing over time. Before, in the Facebook or WhatsApp days, it was a very clustered approach. I still think a clustered approach works, but is it Silicon Valley or outside of Silicon Valley is a question up for debate. So once you have an entry strategy, you have to think about, how do I grow from there? Because the reason why, both at YC or even outside in the ecosystem, people ask founders to really focus on growth is if you are a product with early signs of network effect, the only way you build a moat is by growing, because you have to get that critical mass. So growth is the second most important thing. The third is what is that engagement trigger? This is something that Facebook established in the early days, which was if you are a new user to the platform, if I could connect you to 10 friends within 14 days, the retention was much higher. And I think it's more important for any company that's building a network to really focus on what that trigger might be. You may need some time and a lot of data before you understand what it is. But paying attention to what is the trigger that sort of creates that aha moment is important. And engagement, it always keeps coming back to engagement. If you think you're building a network, your old existing users or older cohorts have to be using it more. Else you have to really pause and ask why that's not the case. And then the fifth related to all of these entry strategy and distribution platform is people think of the market is wide and it's a huge market opportunity. But where do I start? So which cluster do I start? Or are there, you know, some kind of topology which you could sort of attack first and sort of measure your product market fit before you sort of expand? So those are the questions you'd be asking if you're building a network. For example, Slack is trying to build a network. And a lot of the questions that they grapple even in this growth stages around distribution. And who is that one team member who's going to create this aha moment of getting all the team members on Slack, Right? So if you are a company like Slack or trying to build a network, you have to pay attention to these. If you're a marketplace, the questions are slightly different because now you have two sites and they more often than not are heterogeneous. Even though there's overlap between host and guests and Airbnb, it's not a lot. The question is the chicken and egg problem. So if you scale one site too much, they're too fast, then the other site suffers. And so the, the experience on the platform suffers. So how do you sort of measure that balance and then how do you get to critical mass? Because these kind of marketplaces actually might take longer to realize the network effect. And so, for example, Airbnb took 36 months or more. It was a very slow growth.

So to pause you really quickly, this is probably like the first or second most common question about marketplaces that we get. Like, as a whole, like in terms of questions, do you have any tips for people who are creating marketplace, like, as to how to get over that chicken and egg problem?

Yeah, I think that too. If you ask the successful marketplaces or marketplaces that have scaled really well, they would almost always say they focus on demand first and, you know, let's. And it's for various reasons, it is the right strategy. So I'll give you the example of Instacart. When Instacart first launched, I think they launched it as a, you know, just like an online grocery aggregator, meaning I could search for a tomato or I could search for an egg, right? And within six months, they changed the format to retail stores. So you select whole foods and then you go launch. And the simple reason was because a lot of the customers that were ordering via Instacart or did not repeat an order said, hey, I really want to know which store I'm getting the groceries from. And, you know, that really helped fuel their growth. So they focused on the demand side to really understand what is the value prop that they are sort of offering and then you saw. And then if you know, for Instacart, it's a city by city rollout. So it took almost, I think almost a year or a little less than a year for them to get their playbook right in San Francisco before they sort of rolled it out, out to another city. Now it doesn't mean you will take the same amount of time and hopefully you shouldn't take the same amount of time. But all the things that you learned in rolling it out to San Francisco will be helpful for you to reach a similar scale or even halfway there much faster in other markets. And that is sort of the test you're trying to do by rolling it out. And so what happens then is when you're launching in a new city, you need to understand how much demand you may anticipate because you have some, some from your own history, you have some understanding from San Francisco, some understanding of the Chicago market, and you staff some amount of supply on standby. Right. Meaning how many shoppers do I have now? You will never, if you ask any marketplace, they never get it right. Always the first time. So one of the things I know Instacart used to track in the early days was lost deliveries, meaning these were deliveries they couldn't fulfill or people who couldn't complete the order simply because they didn't have times available. And that's because they didn't have shoppers available. So you always learn as you roll out to multiple markets. But the reason you focus on demand early on is because it goes back to everything that we teach in yc which is focus on whether customers love your product, do they really want your product and what is the pain point that you're trying to solve? And usually in marketplaces supply always goes to where demand. So if there's enough demand, there will be more people who sign up on the supply side. Right.

People like to make money. And then I guess to wrap up that point, people who are building a platform, what questions should they be asking of themselves?

Yeah, I think with platform the question is a little bit about. It's very similar to the two sided marketplace question, which is like which side is more important. So for example, if you are Nintendo and you have a gaming console, you need a lot of third party game developers. Third party game developers will come to your platform if they know there will be more users who would be playing those games. The questions are a little similar. I think the more important thing that platform companies also have to pay attention to is is this a Market where they think there will be a single platform winner or would they be multiple? And. And the strategy might be a little different depending on if it's single or multiple, because there's cost involved. The other thing is demonstrating commitment to the platform. For example, Facebook, after they acquired Oculus, they showed an enormous strong commitment to Oculus, which then gets more gaming developers to come to Oculus to develop games for VR. So I think there are the questions that are similar to Marketplace, but you have to show certain other elements of strong commitment, because now you have two communities committing to working on that platform. There's another element which is common among marketplaces and platforms is the old saying, which is. Which is your money side and which is your subsidy side. I don't know if you've heard of that.

I don't know.

There's always one set of views, usually not always. There's one set of users who are price sensitive and there's one side which is not price sensitive, but is looking for value. Simple example is, I think Adobe in the early days, they rolled out PDF for document readers free, because if you had charged document readers, their user base today would be a lot less. But the document writers were ready to pay for PDF creation or other tools because they wanted access to. To the readers. So you should try and figure out which side is more sensitive and which side is the hardest side and try and subsidize the harder side. If you know you can drive a lot of value for the money side.

That's a great point. So just kind of wrapping up. You've spent a ton of time researching this stuff. If I am a person who wants to learn more about it, what would you recommend I do? What resources do you like, what books do you like, what podcast, whatever. And we can just link it up in the interview.

Okay. Yeah. So I think that you should read the deck we put together at every scenario, but there are lots of great papers and articles. I think the best way as a founder you really learn is you look at all these YouTube videos of founders who scaled these amazing products in the early days and try and understand their product philosophy and what decisions motivated them them. That really helps you understand how they built the network and what questions they focused on. I think that was the single most important resource I've seen and it helps inform what their thinking was, even if they can't articulate those in academic terms. But you would see that all these points were hit.

That's great. Yeah, we can link up to a bunch of those. Cool. All right. Well, thanks for coming in.

Thank you. That.