Three case studies from Andrew Chen on solving the cold start problem

So Tinder is. I do love that story also, because it's such an iconic product and it's just so fun what they've been able to do. And so Tinder, in the early years, it was a product that actually basically the first version of it had everything, all the elements that you would think of as Tinder. They just nailed the product idea from day one. And sometimes that happens. The first version of Instagram just nailed it. The first version of Tinder just nailed it. John Bedin, who was their iOS engineer, had a deck of playing cards that he would kind of fiddle with while he was coding. And one day, the very, very, very first version actually just had an X and a check mark, and that's how you would kind of advance and you didn't swipe. But then he was like, oh, you know, swiping's fun. And like, It's a new iOS thing. And like, so he added it in and then, boom, off you go, this iconic thing. So they built basically all the right features. The problem with building all the right features is that then they went and they started to invite their friends and, and they, they went on their address books and they started texting their friends again. This is at a point where online dating kind of is a little bit of a stigma. So imagine this. You're getting, you're, you're, you're texting your friends and you're kind of like, you're almost saying, like, hey, you seem kind of lonely. I feel like you, you need, you need a. You need a boyfriend. You know, it's kind of a little bit like an insult. And then importantly, I think the Tinder Atomic Network is not two or three people. It's also not thousands of people, but, like, it's probably, it's like probably hundreds of people because you can actually swipe through. If you think about how many swipes you can do each day, you can actually swipe like a couple thousand people a day pretty fast, pretty easily. And so as a result, you actually need hundreds of profiles in order to make it interesting. And so that was the problem, that was their cold start problem, was how do they get enough people in it at once? And it just turns out even though they had all the right features, that even when they told their friends about it, it was not working. And I think that that is such an important thing just to stop there for one second, which is that sometimes you actually have all the right features. You just don't have the right network.

Exactly, exactly. Yeah.

And it's very easy that for them to potentially have just stopped right there and just have said like, you know, okay, like this isn't working, okay, let's move on. But instead what they did was they realized that, well, how do we get everyone in the app at the same time, that's all in the same geography and do it all in one go. And what they realized was they could throw a birthday party. They were going to go and sponsor this birthday party from one of the really popular people on the USC campus where they'd gone to school and they had one of their siblings, younger siblings was still in school there. They ended up throwing this party and sponsoring it. And but funny enough, it was like it was at an amazing house. You had to like they bust people from the campus and the whole thing. But then they actually required you to install, install Tinder, show it to the bouncer and then it's in order to get the party. And like you wouldn't use Tinder at the party. But what happened was after the party was done the next day, there was a bunch of people that maybe you saw from across the room that you didn't have a chance to talk to. And so, and so they would just check the app and off they went. So they saw that if they could get 500 people at a party together and this was like 500 of the most desirable people, this was like the Greek system, super like popular, like hyper connected to people, that then they could take over the whole USC campus. And by demonstrating that to themselves then they were able to then pick another school and another school and another school and the next school they picked wasn't a, you know, tiny liberal arts school on the west coast. What it was was they actually started to go after some of the schools that were in some of the southern states. And the reason is because the Greek system is very powerful, very, very strong there. A large percentage of all students are in the Greek system. And so they could kind of do this top down approach and go to the top fraternities and sororities and like kind of sell into there and then like build it and then, and then, and then what happens is once you get USC and UCLA and Loyola and bunch of these other schools, eventually it seeps into LA and you get la and then once you get la, then you get New York and San Francisco and then it goes from there. And so I think it's just this really interesting thing where that is also why, you know, to something Ellie that you just said is that, that's why I don't think of Tinder's network as a super, super broad monolithic network. It's a network of networks of all of these individual schools and neighborhoods. Because even in San Francisco, for instance, you're not going to necessarily date somebody that is from, you know, deep, like South Bay in San Jose, like, that's just too far. That's like an hour and a half. That's too far. Or like you're not going to date somebody that's in Berkeley because that's like, that's like an hour. And so it's a very hyper local kind of experience. And so you can win and lose these hyper local networks. And it was important for Tinder to be able to win kind of piecemeal on all this.

So I said we were going to go to LinkedIn. I may steal a little bit of time and take us to Reddit as a case example, but I just want to pull out a couple of the amazing nuggets that you have there, which. And I'll tee it up for you for both of those case studies. So they sat down and they looked and they had a killer product. They just needed that amazing high quality atomic network and they sat down, they did the work to sort of figure out how to do it. They manually did it, right? Dang. You have for Reddit there, right? Like they manually did it and they found ways to operationalize it. To your point about the zeros earlier, I love this concept, right. If they had gone to a network and somebody wasn't in the Nordics, oh my goodness, that still they would not have the best of experiences. And then the curation element from, from the LinkedIn aspect. So I'm teeing you up for a few more of them. But there's just so many good concepts in the book with these case studies, so. Yeah, yeah, yeah.

Well, I'll start with kind of this, this idea. So Paul Graham has a very, you know, one of the co founders of Y Combinator has an amazing essay from years back, still very relevant, called Do Things that Don't Scale. So that title is very compelling. But then he actually has one line that I just want to read out, which is the most common unscalable thing founders have to do at the start is to recruit users manually. Nearly all startups have to. You can't wait for users to come to you. You have to go out and get them. And I think that's a really tough thing for a lot of founders because it ends up feeling like, well, they're very product driven, they're engineers they're maybe a little bit introverted. That's why they're in tech in the first. And then all of a sudden you have to, like, manually, you have to like, throw parties and like, get people like, literally manually to, like, come. I mean, it just seems like it's just like not a very fun thing. But I think what you see in all of these stories and examples is that at the very beginning, to get your first hundred users or customers, it actually is. It tends to be very manual. And you're trying to kind of put in the teams, the startup team, the founding team, is trying to put in that energy at the beginning to even get that to go. And so very quickly on LinkedIn, and I'll spend a little bit more time on Reddit, but very quickly on LinkedIn. When I interviewed Reid Hoffman for the cold start problem, one of the things he broke down is like, look, there's a professional hierarchy. At the very top of the hierarchy, you have folks like Bill Gates that everybody wants to meet, everyone would spend time with. And so someone like that is not necessarily that interested in a service that lets you network with other people. That's not really like a thing. On the other side, there's a bunch of folks that are not as professionally successful and where they're just breaking in and they don't have as much value to add into these entrenched communities. And people need, need reinforcement or they need intros in order to get a hold of them. Neither sets of those users are going to make for an amazing initial community for a product like LinkedIn. And so what Reid and the team did was they really identified this set of the folks that had seen some success. He gave one example at the time, which is like Mark Pincus, who was at the time starting Zynga and wasn't like done with his career, but he was still kind of in the scene, still wanted to meet people, still needed to recruit people and folks like that who are still in the grind. And he was like, well, let's build, build LinkedIn using that base of users and get them going. And then after we get them going, it'll attract everybody else, because once you get enough of those people, then some of the top folks will be very interested just to meet them. And then some of the people who are breaking in will aspirationally want to speak with and connect with some of the folks in the middle. And they literally manually just told everyone in the company, let's all just send links, let's just go through our address books. One by one and just tell our friends they should be on this platform. And it was invite only for the first week and then they opened it up and they went from there. Reddit was, was sort of like a really fun example because in the Reddit case, it was original Reddit, there was no subreddits, it was just one homepage. And Steve Huffman and Alexis Ohanian, what they did was they just said, well, if we don't post content to the homepage, no one's gonna view it. So their hard side of the network was the content creators that put in, you know, and these days it's the moderators and the community people and everything. And what they realized was, well, we need to actually get content up there, otherwise there's gonna be nothing to do. On the flip side, the problem was if you just saw a million links all from Steve, you know, or a million links all from Alexis, that's not good either. And so what they did was they did something, you know, that, that, that, that, you know, now, now it's, now it's kind of like a funny, you know, joke within the thing, which is they basically just wrote these scripts and they would just make all these fake users and the fake users and just, it just, you know, put, put content on the screen and, and, and fill up the homepage. And the funny thing was, you know, first on the days that they didn't run this script, the homepage would be empty and they would like freak out. And so sometimes they go on vacation and they check the homepage and be like, oh my God, Reddit has nothing on it. Like it would freak them out. And then on the other hand, they were posting all these things and it took them months and months and months of doing this. And it started out with just manual posting. Eventually they added all these scripts that would like scrape the top websites online and try to grab cool headlines from CNN and try to put them in, and headlines from Dig and headlines from other places and put them all in. And then eventually one day, one of the things that Steve talked about is he opened up Reddit and he had forgotten to run the script that day. And right there it was completely full of links. And he kind of knew that it was like other. Anyway, so this whole chapter is actually the case study for a concept that is in the book I call Flintstoning, which is Fred Flintstone. If you remember the old cartoon, the Flintstones, they had a car. And if you got in the car, you just use your legs and it was human propelled automation. And really, this is not just another way to say that these were all kind of human in the loop type bootstrapping methods in order to solve the cold start problem. And you only use it for a while and then eventually you get enough users and the product gets far enough along that then you can let go and your community takes it over for you. And so I love the Reddit example.