How SaaS investor Andrew Chen of a16z is helping entrepreneurs use network effects to tackle the cold start problem
How SaaS investor Andrew Chen of a16z is helping entrepreneurs use network effects to tackle the cold start problem
Andrew Chen (00:03):
What is it about companies like Airbnb and Uber that make them so big and so successful? What I find is, they all share this really interesting, really powerful theme, which is that these are products that get more valuable the more users are on them.
Peep Laja (00:21):
I'm Peep Laja. I don't do fluff. I don't do filler. I don't do emojis. What I do is study winners in B2B SaaS, because I want to know how much is strategy, how much is luck, how do they win. This week, Andrew Chen. Andrew is a partner at VC fund, Andreessen Horowitz, a SaaS growth expert, and author of The Cold Start Problem, a book about engineering and using network effects to drive growth. He's no stranger to network effects, as he sits on the board of Clubhouse and AppSumo, and previously led Uber's rider growth product teams. In this episode, we talk about harnessing the power of the network effect to grow your business. We hear about innovative ways companies have gotten around the cold start problem and built loyal, engaged networks of users, and we hear about using data and insights to try and separate global players of the future from the rest of the pack. Let's get into it.
Peep Laja (01:17):
What's your take on different kinds of moats? Is one better than others, and how good of a moat is network effects?
Andrew Chen (01:24):
Yeah. Well, I started writing The Cold Start Problem because after 15 years in Silicon Valley, I always just thought about what is it about the companies here that make them so big and so successful. These are companies like Airbnb and Uber, where I spent many years running some of the growth teams there, companies like Dropbox, companies like Slack, Twitch, YouTube, and what I find is, they all share this really interesting, really powerful theme across all these products, which is that these are products that get more valuable the more users are on them. And I really have come to believe that that is the secret to many of Silicon Valley's successes. And so I think that there are many, many other moats. Obviously through the years, just being able to have extreme technology differentiation. You have design as differentiation from Apple. You have a lot of these types of things, but really, I think network effects is just incredibly, incredibly powerful, so I've made myself study that in a lot of detail and interview a lot of people just on that topic.
Peep Laja (02:35):
Network effects also can be very different depending on the type of company. So in your book, you talk about Slack, for instance. So in there, you create your own networks, right? It's different from, let's say Uber, which is a true two-sided marketplace.
Andrew Chen (02:49):
That's right. I think there... In the broad theme of there being products becoming more useful with more users, that's obviously something that manifests itself in different ways. And I love to use an example from 1905, actually. There's a amazing quote from Theodore Vail, who was chairman of the American Telephone & Telegraph Company, you might know that as AT&T now, where he talks about the idea that a phone is not useful on its own. It's only useful in the way that it allows you to connect with other people. And so when you think about it that way, you're like, "Huh, okay. Yeah. That's what a phone is. It's very kind of peer-to-peer. It allows people to call each other. Sometimes you're the caller. Sometimes you receive calls. Okay, great."
Andrew Chen (03:36):
That is very, very different of a structure of a network, because that is a global network where everyone can call each other. That's very different than, say, a two-sided marketplace like Airbnb, where there are distinct roles like hosts and guests. And guests always want to stay with hosts. Guests generally don't interact with each other. Hosts generally don't interact with each other. The nerdy way to say this is this is a cross-side network effect. But then you have products like Slack, which are really focused around the team that's right around you, or Zoom, which is you're building a network, but it's very ephemeral. You basically are creating a 3% network for one hour and then you're going to call it quits after that, and so... Versus the telephone is something that maybe resembles most closely to WhatsApp, for example these days, where you might reach out and text almost anybody, and the power of it is the fact that WhatsApp has now billions of users, and that's why it's such a valuable and important network.
Peep Laja (04:33):
Who do you think has done network effects or is using them really well on the B2B side?
Andrew Chen (04:39):
The B2B side is where a lot of innovation in network effects is happening right now, and the reason for that is because we went from an era where enterprise software was just bad. It barely worked as a user. You just hated it. If you used older versions of Salesforce or used old versions of a lot of these products, they were just not good at all. And then all of a sudden over the last 10 years, you've seen these very, very user-friendly products like Slack, like Zoom, and there's emojis. It's like they're meant to be fun, and you have this whole movement which is product-led growth. But I think what's underlying a lot of that is these products are meant to be consumer-grade experiences that let you connect with and network with people that are your colleagues at work.
Andrew Chen (05:33):
And so when I take the evolution of a product category... I'll give you an example. Word processors are a really, really interesting example of this, because you go back in time and you had products like WordPerfect, Microsoft Word, of course, and these were products that had really very, very light or no network effects at all. You basically used them to type your own stuff and then you'd print it or you'd put it on a floppy disk or whatever and off you went. And then what you saw was Google, in their workspace suite of products, all of a sudden add comments to Google Docs. You saw them add collaborative editing in Google Docs, and it really brought the network into the actual product experience itself.
Andrew Chen (06:23):
I think that what we're seeing is a new generation of products, whether you're talking about Notion or Asana or Airtable or Slack or any of these, which know that you do work with colleagues and you can at mention them and you can comment, and you're seeing this new generation of products come out, and I think these products are just going to be better. They're going to be more successful, because you're not going to need to hire big sales teams in order to sell them. There's going to be really great self-serve businesses, and then you're going to be able to overlay a sales team on top. And I think that's what we're seeing in the market. That why these bottoms-up companies have been so valuable. Who would've ever thought that Zoom and Slack and these companies would be worth tens of billions of dollars?
Peep Laja (07:12):
You never stop tinkering with your business model and your products. Sustainable competitive advantage is a thing of the past. Maybe only a few industries left that have some, like airplane manufacturing. Transient advantage is the name of the game today. Use what you got today to win, while building up the advantage of tomorrow. Strategy expert and Columbia Business School professor, Rita McGrath, argues that sustainable competitive advantage can no longer be the holy grail for companies, because in a constantly changing environment, deeply ingrained structures and systems designed to extract value actually become a liability, and the new path to winning includes taking advantage of shorter term opportunities.
Rita McGrath (07:55):
Our traditional views of strategy treats companies kind of like a castle, and you put a moat around it and you try to defend it to the death, and it's very much about defending the status quo. Well, you know what? In more and more parts of our economy that world goes away, and instead, what we have is what I call waves of competitive advantage, which are advantages that you move through as they develop. You got a whole pipeline of advantages, each of which has a specific lifecycle stage. So, you got the innovation process. How are you going to find those insights that drive the new way of thinking? Then you got a process of ramping those up. How do you get them to scale so that you can actually take advantage of them? Then you got your exploitation phase, which is lovely, but you've also got to be hyper-aware that things may be changing. The advantage may be eroding, and if it does, you need to have a disengagement process that pulls resources out of that old advantage and reinvests in something that's new. So what you want to be thinking about is an ongoing series of waves that your organization is creating to take advantage of transient advantages. By the time you need change management, things have probably gotten too settled.
Peep Laja (09:07):
Can any product build in network effect, and should they try?
Andrew Chen (09:13):
Look, if you are selling mattresses on the internet, I don't know how you add network effects to an idea like that, but if you are in the hotel business and you've been renting rooms out of your hotel for a decade, you're one of these companies like Hilton or whatever, it must be terrifying to then come up against a company like Airbnb where they have no hotel rooms at all. They have no hotels at all. It's just a network. It's just software, and that's what you're competing against. So I think a lot of these categories can be reinvented to incorporate network effects.
Andrew Chen (09:49):
I think there's also the idea that if you're an eCommerce company, and instead of thinking about it like, "Okay, what I'm do is I'm going to dropship stuff from China and I'm going to sell it on the internet, and that's my whole business," the other way to think about it is maybe your job is to actually build a community and a network around everything that you're doing. Like this is an offline example, but a movement and an organization like CrossFit, and the fact that they've been able to put together this really amazing fan base and set of practitioners, it makes it so much easier for them to then create and launch new services, new products, and introduce it into that community.
Peep Laja (10:29):
Network effects is a conscious moat for many companies. Sometimes it's a two-sided marketplace like eBay or Uber. Sometimes it's the ecosystem you're building around your product. This is exactly what Shopify is doing. Here's their VP Growth, Morgan Brown, talking about this in a previous episode of How to Win.
Morgan Brown (10:48):
In terms of the moats Shopify has going for it, or things that work really well, one is the counter positioning. Amazon is customer-obsessed. We're merchant-obsessed. That difference in positioning actually creates a whole bunch of interesting strategy trade-offs and sends you down very different paths. If you are a marketplace like Amazon, you're trying to aggregate as many people as possible so that you have the most sellable asset as possible, right, and then you can dictate how much you charge and the rules of engaging that asset. So it's Amazon, Facebook, Google, eBay, LinkedIn. All these marketplaces have very similar dynamics, or you have the platform side, which is what Shopify is, and then there what you're really trying to do is trying to create the maximum amount of value for people on that platform.
Morgan Brown (11:39):
I think other interesting things are kind of the network effects like the app developers, the experts, the agencies, and that type of thing kind of creates this ecosystem of extensibility and expertise that is really self-reinforcing. And then ultimately, I think you get into really some interesting network effects when it comes to the data network effects kind of across all of the different stores and merchants and... All across Shopify is really interesting.
Peep Laja (12:03):
Let's talk about the cold start problem. So if one wants to start engineering a network effect in their business, they have a kind of a chicken and an egg situation, so how do you start overcoming it?
Andrew Chen (12:15):
The cold start problem is the opposite side of the coin of having network effects. People usually talk about network effects like they're going to make it so that you can grow exponentially, you're going to be able to tap into viral growth, you're going to get all these incredible benefits, and that's true, but the problem is if you build a product and none of your users are using it, then the product is not valuable at all, even if you have the right product features. In the early days of Tinder, they actually had built... The V1, they actually had built many of the right features. They had the swiping, big profile photos, messaging. They had all the things that you would need, but when they would invite their friends into the app, their friends would quickly churn out because you can swipe profiles pretty fast, right? And so if you just have ten other users, a hundred other users, very quickly you're just swipe, swipe, swipe, swipe, swipe. You're done. You're kind of like, "This is boring. I'm going to move on."
Andrew Chen (13:12):
And so they figured out that they needed hundreds of people to be in the app at the same time in order for it to work, and so what they did was they realized, "Well, where are hundreds of our customers all hanging out in the same place?" And they realized, well, maybe they should go to the nearby college campus, USC, throw a birthday party for one of the really popular, really social, really connected people on campus, and invite 500 people and tell them, "Look, if you come to the party, and we're going to have bouncers at the party, if you haven't signed up for Tinder, we are going to turn you away. So you have to install Tinder. You have to get in. You have to do this thing." And that's how they got 500 people into the product at the same time.
Peep Laja (14:03):
Tinder co-founder Sean Rad knew that building a network effect was core to the success of Tinder, and for that to happen, the user base had to grow organically within a community once they knew about it and grow towards a specific customer need. Here's Sean explaining what happened at the party.
Sean Rad (14:22):
We knew our goal isn't to sort of create this directory of people to solve a problem where you as an individual might not have access to enough people. We were trying to solve the problem of you knowing who you want to talk to, or sort of going about your life and seeing lot of people that you want to talk to, but not having the context to sort of go up to them and start a conversation. We realized that to have a valuable ecosystem where people are actually engaged and they're actually using it the right way, we needed the product to grow organically, and for two reasons. One, a user who comes in organically through a friend, through word of mouth, is way more valuable because they're more engaged, and when you sort of have this network effect, you want your user to be engaged. And if they're not, when I swipe right on somebody I'm not going to get a response, and sort of the overall value of our ecosystem goes down. So, every incremental user that is organic increases the sort of overall value of the ecosystem. Any user who comes and deflects decreases the overall value. And we realized that to really understand Tinder, it's something that your friend needs to teach you.
Andrew Chen (15:32):
One of the core parts of my theory is that you need to understand the size of the atomic network, that is, a stable, minimal network that you need for your product to be valuable. And so for Zoom, that is two or three people. For Slack that is maybe five or ten people. For Tinder, it happened to be hundreds of people, and so you need to figure that out in order to get your product off the ground. And so, I think every product has to figure that out for themselves.
Peep Laja (16:02):
So if you're a two-sided marketplace like eBay or Craigslist or AppSumo, how do you figure out the number? Is it experimentation? Is there a formula?
Andrew Chen (16:11):
Yeah. I wish there were a formula, because if there were a formula then my job would be a lot easier to fund all of these two-sided marketplace companies. There's a couple, I think, key concepts that these new startups need to focus on. When Uber first got started, the thing they figured out was to launch a new city what they should do is they should go to Craigslist. They should tell drivers, "Look, if you run this app, we'll pay you $30 an hour. Whether you get rides or not, we will pay you the $30." And they were then able to focus on the drivers portion of their network, the supply side of their network, get them onto the platform, and after getting them onto the platform they could then work on the demand side, and then they would do local promotions and all these other things.
Andrew Chen (16:58):
Similarly, you bring up AppSumo. AppSumo is run by Noah Kagan, one of my best friends, and I'm actually on the board of the company and have been involved since the very beginning. For Noah's early, early, early cold start problem, what he needed to do is basically figure out a way to get a lot of software products that would be willing to discount their software onto his marketplace platform, and then enough software buyers who are interested in buying those kinds of software and get them onto the platform as well.
Andrew Chen (17:29):
And so the way he did it, which is just incredibly clever, is he would go to conferences and he would say, "Hey, why don't we make a conference bundle where basically everyone can put their product into the bundle with a discount code, and then everyone will share an email list, and then we'll send this bundle out to everybody?" And in that way, if you are a software buyer, and a lot of these were small and medium-sized businesses, they would be doing this, what they found was he was able to get these bundles sent out to people. People were then buying because of the bundles, and then after that whole experience, he would then walk off with thousands of email addresses of all the buyers and all the software bundles.
Andrew Chen (18:11):
And it turns out if you do that, that's an atomic network right there. And if you're able to do that once and twice and five times and ten times, you can probably run that same playbook fifty times or a hundred times, and then you're able to launch your two-sided marketplace with that in mind, the same way that Tinder figured out you could go college campus to college campus, or Uber was able to go city to city. AppSumo went from conference to conference to conference in a way where it would launch the whole thing.
Peep Laja (18:40):
Is that option only available for if you have deep pockets, you've raised a bunch of money? Is that an avenue available for bootstrapped or do you just go slower or...
Andrew Chen (18:52):
Yeah. Well, I think you have to decide which type of playbook works best for you. The AppSumo case, of course, didn't cost anything at all. That was just kind of hustle from Noah's end to be able to put that together. In later years, AppSumo has done more with money where they've been able to, for example, work with certain software developers and maybe promise them kind of minimum guarantees and saying, "Hey, we'll do a bulk order. We'll buy 500 licenses at a discounted price and resell them." And that's more similar to what the Uber model actually is. So I think using money to solve it is one solution.
Andrew Chen (19:30):
You can also build tools for the supply side of your marketplace. And so if you look at what companies like Convoy, which is a B2B trucking marketplace, they actually fundamentally at their core are operating a brokerage, and then they've layered technology on top. There's another type of solution. In the Dropbox case, they basically made an amazing product that an individual can use for backup and storage, but then also you can share and connect with your colleagues using Dropbox, and then off you go. And so I think these are all different kinds of solutions that you might apply. You'll probably end up trying a whole combination of these. There's no guaranteed solution that's going to work. It's going to be very, very focused on your particular market, your particular case, whether you've raised funding, whether you haven't, et cetera, et cetera.
Peep Laja (20:21):
When we think into the near future and look at the onset of new technologies, the Web 3 and so on, how do we think we can apply some of these lessons on Web 3 companies? Is it to issue token to different sites to incentivize certain behaviors, or what do you think?
Andrew Chen (20:37):
If you are working on Web 3, you are spending a lot of time thinking about network effects and thinking about the cold start problem, because in order to get your NFT drop to be valuable, and for everyone to be excited about your NFT drop, you need to get enough people excited, and you need to figure out how to get the right Discord communities, the right Twitter communities, the right Reddit communities to embrace what you're working on.
Andrew Chen (21:03):
Now, to your point, I think we're going to see a lot of evolution of a lot of the ideas that I mentioned in the book, but they're going to become kind of done in a Web 3 way. Let me give you an example. One of the things that you often see in consumerized B2B products or in consumer products has been this referral program. Dropbox has had this program where you can give people storage space, and then you also get storage space when you refer people, and Uber, obviously, and Airbnb, they also have these ones where you give $10 and you get $10 and you have that.
Andrew Chen (21:40):
Web 3 is fascinating because inherent in Web 3, as soon as you get a token, then all of a sudden you feel the need to promote it. Right? And so if you get your friends to also buy a token, then the value of your tokens will probably go up, which is why the ultimate result, the ultimate outcome of this is very much similar to a referral program, structurally, which is how do you get the people to talk about what you're working on? So I think there's going to be a lot of really, really cool innovations, and we're going to see a lot of different variations like this.
Peep Laja (22:17):
You did a lot of research for the book, and you've been thinking about this problem, so when you compare companies that get network effects right versus who get it wrong, what's the difference, typically?
Andrew Chen (22:28):
Well, I think the biggest danger comes in the fact that once a company is become more established, the DNA of doing new things tends to go away a little bit. All the people that had to solve the cold start problem in the early years and had to do all these clever and creative things, they eventually leave the company and do their own startups or go do something else, and what a lot of companies are left with are a lot of employees that are better at kind of increasing the metrics 5% here, 10% here, and they stop understanding how to solve the cold start problem. To me, this is a flavor of the innovator's dilemma.
Peep Laja (23:11):
In his book, The Innovator's Dilemma, Clayton Christensen states that there are several kinds of innovation. Andrew invests in companies doing disruptive innovation using network effects, bold new ways of thinking which can redefine a whole market. More established companies who have already disrupted the market to their advantage and have a position to lose tend to only offer incremental improvements or sustaining innovations. Here's Clayton explaining the concept.
Clayton Christensen (23:37):
Sustaining innovations, their role is they make good products better, and they are very important in society because if you don't keep making good products better, your margins will fall and the company will dissolve. So it's very important. And when you go out in the world, most of the innovations that we see are sustaining innovations, but sustaining innovations actually don't create net growth, nor do they create jobs. So just imagine that I'm a salesperson for Toyota, and I convince you to buy a Prius Hybrid. If I succeed in selling you that, then you won't buy a Camry. If you buy the next generation of my product, which is better, then you won't buy the last generation, which wasn't as good. And so, sustaining innovations are very critical. They have a role, but they don't create growth or jobs.
Andrew Chen (24:35):
It's really a flavor of why it is that when Google tried to launch Google Plus, they didn't do it organically the way Facebook did, which is to go from Harvard College to all the other schools and to build these atomic networks and to get the engagement to be really, really strong. Instead, what they did was they just linked to Google Plus from the Google homepage and from Google Photos and from YouTube and from Google Maps and all these other places, and the problem with that strategy is, although you got really big numbers... At one point, they were actually claiming they had 300 million active users, which is just amazing. On the flip side, these were not very engaged users. These were not users that were very connected to others, and so what that means is you end up in a situation where inevitably, as soon as the fire hose of users stops, the whole network was very weak actually and it just collapses. The whole thing will collapse, and so I think that is one huge mistake that the bigger companies make.
Andrew Chen (25:36):
I think the other mistake for startups that get into network effects is a lot more obvious, how to build a product that has no network at all, because you have a discipline, like product design, you can listen to your customers, you can do all these other things, and it's pretty clear how to get to product market fit when you're just building this kind of like widget. The thing that is really hard about network effects is that not only do you need to nail the product, but you need to introduce it and build the user base right away. You need to build a critical mass of the user base right away, and you need to do both at the same time, and that is much, much, much harder. And I think that the companies that figure it out, start by basically iterating on the product and building their community all at the same time, and that makes it much more likely that they will hit the right critical mass with the right features and take off as a result.
Peep Laja (26:31):
I love this explanation of why Google Plus failed. If you think about it, Andrew is a hundred percent right. Google drew a bunch of traffic, but it wasn't managed right. It wasn't converted into networks. Google likely didn't have a language to think about atomic networks. I'm learning similar lessons right now, myself, building a two-sided marketplace at Wynter, connecting SaaS vendors with folks buying SaaS or cloud products. Things that I'm looking at, what does an atomic network look like for Wynter? It's not city by city. It's not college by college. So for us, I think it's verticals, business niches where vendors serve a specific audience and set of needs. We're connecting SaaS marketers to more tech vendors, HR professionals to HR tech vendors, software developers to vendors selling software to developers. Each company needs to figure out what the atomic network structure is for them.
Peep Laja (27:28):
How do you look at this now as an investor versus being in-house like you were at Uber? How do you look at this problem differently?
Andrew Chen (27:37):
So my day job is as an investor at Andreessen Horowitz. We've been lucky to work with many, many interesting companies that have network effects. There's sort of two ways that I end up working with companies. The first is on the way in, which is we end up seeing something like ten or twenty thousand business ideas a year, and because of the way that everything works, I should maybe be only doing three or four or five investments per year. So that means I have to say no to a lot of things, and I have to figure out, "How do I get excited enough to say yes to a few things per year?" And so the cold start problem really helps articulate my framework for picking the best companies. It has shaped the way that I ask for data from companies.
Andrew Chen (28:28):
I'll give you an example. When you have a product like Clubhouse, you are often trying to figure out, "Is it really working? Is it really working in a lot of places or is it just this like nerdy tech thing?" And when we did the second round in the company, when we did the Series B, one of the questions I asked Paul Davison, the CEO, was... I said, "Well, can you help me just divide up all of your data by region? Give me your retention curves for Sweden. Give me your retention curves for Nigeria. Give me your retention curves for the US. Give me your retention curves for all the other regions in the world, so that I can evaluate whether or not the retention is just good in the Bay Area and in the US, or is it really, really good in all these other places." And going back to the atomic network idea, what I was really trying to do is I was trying to pick. I was trying to understand, "Has he built 5 or 10 or 20 atomic networks that are standing on their own, or is this really one really good atomic network in the US and then a bunch that haven't worked at all elsewhere?"
Andrew Chen (29:32):
So one of the things that I often will do is, when a product is able to be disassembled, their network is able to be decomposed into smaller networks, and I always ask for data on a network-by-network basis. And what you want to see is that the markets that are more successful, have had more growth and have more users, have higher retention rates, and that what that tells you is, that lets you make the connection between network density and growth and the network effect that is being created. I mean, I love to invest primarily in companies that have network effect somehow. And so part of writing this book, selfishly, is to be able to hand over a document to the founders I work with to be able to say, "Hey, I think you're in the tipping point phase of this whole thing, and these are some things to think about," to help them operate, to help them grow, and to help them grow, not by buying ads on Google and Facebook, but instead to help them figure out how to amplify the network effects, how to build the right features in their product, and to go from there.
Peep Laja (30:43):
So what's Andrew's advice on how to win with network effects? One, look for smart opportunities to take advantage of the power of networks.
Andrew Chen (30:52):
Look, if you are selling mattresses on the internet, I don't know how you add network effects to an idea like that, but if you are in the hotel business, it must be terrifying to then come up against a company like Airbnb where they have no hotel rooms at all. They have no hotels at all. It's just a network.
Peep Laja (31:12):
Two, figure out the minimum network size you need to have to get to a viable product.
Andrew Chen (31:18):
You need to understand the size of the atomic network, that is a stable, minimal network that you need for your product to be valuable. And so for Zoom, that is two or three people. For Tinder, it happened to be hundreds of people, and so you need to figure that out in order to get your product off the ground.
Peep Laja (31:38):
And three, look at the data to assist whether you are creating a true organic network effect or whether you need to adjust your strategy.
Andrew Chen (31:46):
I always ask for data on a network-by-network basis, and what you want to see is that the markets that are more successful, have had more growth and have more users, have higher retention rates, and what that tells you is, that lets you make the connection between network density and the network effect that is being created.
Peep Laja (32:08):
A final takeaway from Andrew.
Andrew Chen (32:10):
It is a lot more obvious how to build a product that has no network at all. The companies that figure it out, they start by basically iterating on the product and building their community all at the same time, and that makes it much more likely that they will hit the right critical mass with the right features and take off as a result.
Peep Laja (32:32):
And that's how you win. I'm Peep Laja. For more tips, follow me on LinkedIn or Twitter. If you like the podcast, please subscribe and leave a review. I read every single comment. Thanks for listening.
