Introduction
Discover the Discord user growth strategies that helped scale the platform to 200 million users. In this exclusive interview with a former head of product at Discord who spent eight years building the platform, we dive deep into the marketing tactics, product decisions, and business model evolution that turned Discord into a community powerhouse. Learn how Discord identified their target users, built features that drove organic growth, and created a sustainable business model.
Founder Success Story QnA
How long were you at Discord and what was your role technically?
I was at Discord for eight years, and I was the head of product early on. I built Nitro and built the business model.
Can we talk about marketing strategies of Discord early on?
When we built Discord, this was before my time—this is me paraphrasing. But before my time, the team really honed in on specific use cases. For example, we wanted to power MMO guild raids, and we wanted to make sure people could use Discord for games like Overwatch, League of Legends, DOTA, WoW, etc. The team actually went game type by game type, went into the Reddits and the forums of these guilds that people were using, and basically asked people very manually, “Hey, come check out this app. What do you think? What do you like?” Most notably, Final Fantasy was one of the first games that Discord was built for, in part because the founders were playing it. So a lot of the early marketing was just building a very tailored solution to a problem, going into these Reddits and forums, and convincing guild leaders to adopt it and join it.
Did they make it game specific or was that just the marketing itself?
That was the marketing itself. Our philosophy was to always build primitives that solved specific needs for specific games but could be used for other things. So we never made it so specific that we would shoehorn ourselves and say, “Okay, it works here but not there.” Whatever we built in code works for all sorts of games—that’s general functionality. But then when we went and marketed it, it was very tailored.
How did you identify the super nodes like the raid leaders?
Early on, we went to forums, we went to subreddits, we went to gaming conventions like PAX, for example. We would go to streamers, and they usually either would have exposure to these raid leaders or they had a following where these raid leaders hung out. It depended on what approach or channel you used—you had to have a different way of approaching them. And in many ways, as your company scales, as your business scales, you have to evolve it. What worked when Discord was 100,000 users did not work when Discord had 10 million users, did not work when Discord had 200 million users. The strategies you used have to change and evolve, and that’s normal, it’s okay.
Did you ever reward people based on how many users they got?
We certainly gave more merch to people who were, quote-unquote, more successful. But we didn’t necessarily do like, “Oh, you get a bounty for each user that you drive.” As far as I remember, we did kind of tier our marketing where, beyond just the very top, we would pay them; the middle, we made them into Partners; and the smaller and broader you got, we targeted actually a lot of colleges, universities, and schools with our what’s called Hype Squad program. This is all early days—all these things evolved later.
What was the very first thing that was built? Was it just voice chat like group voice chat?
Discord started as text and voice. The voice tech we took from the previous duration company, which was a mobile game. We built a game called Fates Forever. It was a mobile MOBA, and part of the mobile MOBA had to have good voice chat when you’re playing together with other people. So that was the voice tech behind it. And then Stan was one of the lead engineers on the MOBA who then became the co-founder. He had the idea because he had been working on the side on a project called Guildwork, which was effectively like a better way to run a guild forum. He had the idea of combining that voice chat with text chat and with an app called Flow (that was before Slack) to basically come up with what we know now as Discord. So there was always a component of text chat and always a component of voice chat because you need text to organize—”Hey, when are we hanging out?”—and you need voice chat to be able to actually play the game together. We call it kind of like the before, during, and after playing games. Before playing the game, you hang out on text chat and organize it; during, you’re actually just talking to each other (and you’re not in Discord, you’re in the game); and after, you talk about what happened with the game.
Where did the idea for invite competitions come from?
That was actually a community thing that was something that happens organically. Something interesting about Discord—and I don’t think many people realize this—is that a lot of things that happened on Discord weren’t created by us. They are behaviors that happen organically. Sometimes we see a behavior and we’re like, “This should be a thing,” and we make it easier to do. One great example of that is Go Live or screen sharing. We saw people hacking that together where somebody would be on Discord voice chat and also streaming on Twitch and tell their friends, “Go to Twitch, mute, go to voice chat, and I’ll talk to you while I’m playing the game.” And we were like, “This is moronic. Why would you do that? Twitch chat is delayed, Discord is real time. We can make this better.” Other things like invite competitions or a lot of what you’ve seen community servers with—like rules or engagement-type competitions—those are things that Discord moderators came up with organically on their own.
How did you guys build a live streaming feature?
That was a really cool project. As I mentioned earlier, it was a behavior we saw. People were doing it—people would get on Skype to screen share their screen to their friends because Skype supported screen sharing and we didn’t. Initially, we built the first version of it as just regular screen sharing, and we didn’t actually release it because we just wanted to test it. We realized that the screen sharing that we had built wasn’t quite good enough to make it feel like, “Oh, this is like 10x better than what you can do on Skype.” It’s something that we felt was very, very important for just bringing people together that weren’t just playing games together. The purpose of Go Live is actually such that you, as a friend of somebody who’s playing a game right now, can just hop in and talk to them while they’re playing the game. It’s kind of like sitting down on a couch with a friend—they’re playing some console game, and you’re just watching them play and talking about the game. That was the actual human behavior. It took us a long time to get it to be really good with performance. Our engineering team took a long time but kicked ass on making the streaming aspect of it—you could stream in source resolution on a really not-great PC without dropping frame rate. That was a really cool achievement.
Is it not expensive to do that, to let people run it?
It is expensive. It was by far the most expensive infrastructure project. Obviously, at that point, we had already built Nitro, so we were able to burden the cost without needing to raise crazy money, which was good. But it took a lot of time and effort also to make it very efficient on the infrastructure side and make it cost-efficient. We later on signed deals with Google, who was our main CDN provider, to help make that even cheaper for us because we were able to optimize and work closely with their team together. But yeah, it’s not a cheap feature to run, which is why, also, if you want to stream in more than 720p, the cost of such a feature goes up with resolution. The more frames or the larger the frames, the more expensive it is to Discord because everything runs through Discord servers. So that’s why when we released it, we were like, “Well, it’s perfectly fair to ask people to be paid Nitro subscribers if they want to stream in super high resolution and the high frame rates, because that’s very expensive for us.” And it was actually funny enough also one of the top features that made people want to get Nitro because it was so valuable to them.
What is Nitro? How do you answer that question?
Nitro is the premium membership of Discord. And what that really means is that Nitro is the set of features, the set of functionality you get if you want to support Discord but also if you want to show off to your friends. At the end of the day, what Nitro really hooks into is the fact that you hang out on Discord all day, every day, and you don’t want to look like everybody else. It’s a very normal human thing—if you spend a lot of time somewhere, you don’t really want to blend in. Nitro allows you to not blend in by allowing you, for example, to use emoji from other places you hang out with, or it allows you to bling out your profile, allows you to actually buy decorations cheaper, allows you to have access to high-quality streaming. All of the features in Nitro are tied to you hanging out with your friends. If you’re using Discord just for yourself, there’s almost nothing—there almost nothing in Nitro that is worth buying, minus the upload limit, which again is just very expensive to Discord for people to upload gigabytes of files. So that one, yeah, you can use Discord without any friends and still get value from Nitro. But most of the features, the actual value, is you and your friends. It’s actually the thing that you pay for for many people is not just for myself, it’s not just for forming towards myself, but it’s also for my friends. Go Live is a great example—me buying Nitro so my stream is higher resolution is something that my friend appreciates more. I know that my friend appreciates more. It’s a very non-obvious business model.
What was Discord’s original business model?
This is public knowledge: it was going to be a game store. We were going to have the largest group of gamers on the planet using our product, and we were going to sell them games, just like Steam, get a cut on the distribution, just like Steam. We built that a couple years after we built Nitro, and we quickly realized that it wasn’t a 10x problem that we were solving. Discord as an ethos, as a company, as a product, was really good at solving few things really, really well, and we couldn’t build a 10 times better Steam. It was better in many ways—a lot of our users who loved the store would tell us it’s better—but it wasn’t something that I would want to use Discord for instead of Steam. A lot of people who could buy the same game on Steam and on Discord chose to buy it on Steam because that’s where their library still was. There was also, at the same time, Epic who had just launched the Epic Game Store, and the only reason people bought games on Epic is if they were given away for free or if they were only purchasable on Epic. But for anybody who could buy a game on Steam, they’d rather buy it on Steam. People actually want one place to have their games library. People don’t want to have many different apps that have different games. It’s a lot of overhead—imagine your phone having 10 app stores and you have to open each app store individually in order to figure out which one was this app in. Terrible experience. So we quickly realized that we launched a game store, we launched it with a subscription. As part of the game store, we got a collection, a bundle of games that we would put into a higher-priced Nitro. Together with that, we increased the price of Nitro but also included a bunch of games for free in it. This was before Xbox and all of that. We quickly realized that, yeah, we could keep pushing on that, but it wasn’t going to be the runaway success that we had wanted for the business model, and it just wasn’t something that our users really wanted. We kind of went back to our roots, went back to being like, “Okay, let’s solve our users’ problems as the best we can.” We chose to shut it down, chose to sunset the game store over time. Obviously, we couldn’t do it immediately—took us more than a year to do that. But we basically pivoted the business model to just focus on Nitro. We realized, “Hey, why don’t we just let people have more ways to have a better experience on Discord?” And that we’re not required to use Discord—you could still use Discord for free, and you could still talk to your friends on Discord for free—but if you wanted additional cool, blingy features, you can get Nitro. We just went double down. We built server Nitro, which is now called boosting. We built a lot of features around the profile—we built, you know, you can set an animated banner on your profile, you can have a decoration on your profile, you can have custom statuses on your profile. We made a lot of different things in Nitro that all fed into this model of like, “Hey, if you’re using Discord 12 to 18 hours a day, you want to have all the best functionality of Discord,” and that’s what Nitro is—it’s all the stylization and status within a server that you spend a lot of time in.
How much time were you guys spending on the game store?
The game store was a more-than-one-year project for most of the company to build it out before even launching it. The thought process was just: make the best voice comms for gamers platform, get all the gamers on the platform, and they’ll naturally want to just buy their games on the platform too. That’s right, but the problem is, it sounds great in a PowerPoint slide, it sounds great in theory, but in practice, the act of hanging out with your friends is actually quite separate from where you launched the games and played the games from, unless these games are built into the act of hanging out with your friends. And so we’re actually now releasing—well, I’m no longer at Discord, full disclosure—but we’re now rolling out the ability to build games inside Discord, inside a voice channel. And we’re now actually releasing, or already have released, the SDKs for it where people have started building games inside Discord. And there it actually makes sense because since the game can only be run inside a voice channel while hanging out with your friends, yeah, it makes sense to have a game store, to have a launcher there, and to have you and your friends hop on a voice chat together first and then launch the game together.
Did you anticipate trends when planning marketing?
It wasn’t anticipating, and we wouldn’t plan for it. But what tended to happen is that Discord, because Discord is ultimately a platform to get a bunch of people to talk about something and hang out around something—whether it was a new game expansion that launched, a big game of the year, or some sort of trend happened, whether it’s crypto or Pokémon Go or AI or whatever it is, or Midjourney—every year of Discord’s existence, something happened that we could not predict. Nobody could predict, but something would happen that would suddenly cause a lot of people to just create a bunch of Discord servers to do that thing together. I still remember the summer of Pokémon Go—we called it that. We would suddenly start seeing, like, in every city in the US, we would start seeing tens and hundreds of Discord servers pop up that are all Pokémon Go servers. All the largest servers on Discord were Pokémon Go, and everybody was using Discord to organize raids together on Pokémon Go. This was six years ago or something like that. It was a while, but it was a really cool behavior. We could have never predicted it. The game was a hit, the game had this unique mechanic where you had to get together in person in your local neighborhood to do a raid together, and you had to go spot where the Pokémon are. So you have to coordinate with people, and that is something Discord’s uniquely suited for. You can create a server for your neighborhood and you can post whenever you see something on your Pokémon Go being like, “Hey, this rare Pokémon is over there, let’s all hang out, let’s go raid.” And so the best that we could do when we’re building a platform that ultimately grew because of these trends is to make sure that we built all these primitives that I was talking about earlier such that it was easy for anybody who is familiar with Discord to be able to use Discord for this new behavior. So that invites were super smooth, so that we could have—we built a feature where you can just open up the app and see who’s nearby, and if they both have Discord open, you can just add each other as a friend right there. That helped actually a lot of Pokémon Go users, right, because you would just walk up on the street, open up Discord, and “Hey, we’re all on the same app, right?” So we would build these generic features, and that worked. It’s not built into iOS where you can bump people, but ours worked over the audio jack, actually, in Bluetooth. But we basically built out all this functionality such that when such a thing happened, which we couldn’t predict, Discord was ready to handle it, and we wouldn’t have to scramble and try to figure out what do we need to build for what here.
Was Midjourney using Discord surprising to you?
We had seen similar behaviors on the platform to Midjourney where people would create a bot and use the bot to basically collaboratively have a fun time, collaboratively be creative together. Those bots existed before Midjourney. I think Midjourney really nailed the loop of that bot really early on, to where you would generate an image, it would inside the Midjourney server take some time to generate, it would ping you back to go check out what you’ve generated, you can select one, and you had emoji reactions to go take the next action—maybe regenerate something or you loved something and made you a big version of it. And your friends could use those same reactions to do those same actions. And so it was very easy for people to just go take the Midjourney bot and install it in their friend server and just start riffing on each other’s stuff and having a lot of fun. And so we had seen that behavior in other bots, not nearly as polished as Midjourney, not nearly as cool as what the Midjourney generated, but when we saw it, we basically saw the server popping off, and we got in touch with them, and they actually really wanted us to just uncap their server and support them. You know, at the time, our hard cap for all servers was a million members, and they kept hitting that cap for a while because it was really addicting what people were doing, and people couldn’t get into Midjourney early on. The problem was, Discord was never built to have more than even that many people talking to each other at the same time. Imagine a million people in a single place talking—it’s cacophony, and it’s a really hard infrastructure problem. And so we actually worked with the Midjourney team together—our infrastructure team and their team worked together to basically make it possible for Discord to even support more than a million people inside a server that still works, that still is performant, that still can be used. That was a whole infrastructure project, and the infrastructure team spent a lot of time getting that right. But nowadays, Midjourney is what, like a 15 million, 17 million member server? A number of other larger servers have since popped up, and it’s cool. We would have never done it if not for Midjourney.
How did the bot ecosystem develop on Discord?
That’s a great story. So early on, we’ve always—our product and our infrastructure has always been built on very strong principles. And one of our principles was to have a very robust API early on. We exposed that API so somebody else could have built something on top of it, but we didn’t advertise it, we didn’t have any developer documentation, whatever. We just exposed it. People just started randomly hacking around it, created the Discord server called Discord Developers, and there’s a couple others. Different libraries sprung up around like, “Oh, we can hack around a Discord API and just build our own things that hook into it.” And we spent some time making sure that it still worked and it’s compatible, but we didn’t spend much time encouraging that for ages. It’s funny enough, actually, the person that created the Discord developer server actually joined my team later on as one of our earliest, earliest engineers. She’s fantastic. And we later on realized that we at Discord would never be able to build all the cool stuff the community would need, nor should we, because the bot ecosystem can create so much breadth of functionality that a small team like Discord—and back then, when we built, when we exposed our first API, we were smaller than like 20 people—it was a small team. We would never be able to build all the cool bots people built. And we didn’t have to because we had a community to do it. And that’s actually the cool thing about Discord—you can just organize, self-organize around Discord servers to go build these libraries, to go build these bots. It took us, I think, three or four years after we made it possible to create bots on Discord to actually spend time to solidify all of our documentation, make sure that all the libraries are actually good and worked well. We took ownership of some libraries, we helped other library developers really organize themselves and keep it up to date. And we called it the Bot City team for a while. But ultimately, the goal of that is just to allow people to build functionality and do things that the company should never do.
What were some of the most unique bots you saw?
Well, early on, listening to music in a voice channel was a huge behavior. It’s not—it sounds obvious now, but when we first saw people doing it and just like spending hours and hours just jamming together in a voice channel while hanging out at home, that was cool. That looked really awesome. And that behavior, even though these bots, a lot of them ran into trouble with the rights holders, etc., on YouTube—nowadays, it’s activity, actually. Rhythm released as an activity, there’s another, I think YouTube is an activity, so you can actually do that quite legally, and people do it a lot. Another one that early on comes to mind: someone set up a couple of bots that are effectively like—not pre-LLM—chat bots that will be talking to each other all the time, and you can join the Discord and watch what the hell they’re talking about at the moment. And it was hilarious. Like early Discord, we would just like randomly pop in there and be like, “Hey, look, they’re talking about this some random thing,” and it’s a constant stream of conversation.
Have you considered AI agents for Discord?
We have. We actually built an AI bot. We built Clyde. We launched Clyde. Clyde’s the man. Yes, funny enough, we didn’t choose to keep Clyde. And the interesting thing is, yes, so Clyde for context was a built-in bot inside Discord. It was built into Discord, it worked anywhere. And we were testing it for a long time where you and your friends could basically have an AI chatbot in multiplayer. So imagine ChatGPT but all your friends can chime in. And we worked with OpenAI on that. It was a collaboration. And because back then, like OpenAI didn’t even support a lot of things that we needed to make it work. And so interesting, we found that like Clyde was hilarious for creating really funny memes with your friends, but it became stale pretty quickly, and it actually became kind of cringe after a couple of weeks. If the same Discord server with your friends kept using Clyde, it kind of just became stale and cringe over time. And we could never figure out how to make it feel so natural that it felt like part of your space, part of your friend space. And we saw that when we were testing Clyde, we saw a lot of servers kind of having a lot of fun initially. It was really fun for a couple days, a week or something like that, and then