 Hi, I'm Edith Harbaugh. I'm CEO and co-founder of LaunchDarkly. I hope that applause is for me, because it felt good to be back at Slush. I was here three years ago in 2019. And it seems like a lifetime has happened since then. So I'm here today to talk about building your first product, about really getting started at a startup the really early days. So LaunchDarkly now has 4,000 customers worldwide. We recently announced that we crossed 100 million in ARR, annual recurring revenue. But we started with zero customers and zero revenue. And I thought some of the things I'll share today are things that will resonate with the startups in the audience. So from my own background, I started off in engineering. I was an engineering manager at portal software and content management software. And I got very frustrated with releases where I felt like I'd stayed up all night. Me and my team had worked really hard. And when we shipped it, nobody cared at all. And that's the worst feeling in the world if you're an engineering. There's two, I think, parallel fears in your engineering. One that you ship it and everybody hates it, that it breaks the field. The other is that you ship it in absolutely nobody cares. Just silence. So from that, I became a product manager. I had this idea that I was much smarter than my product manager. Once I became a product manager, I realized how hard product management is. The issue with being a good product manager is that there's never a lack of ideas. It's, again, figuring out which one to build, which ones are really going to resonate with your audience, which ones are going to move the needle. So I was a product manager for a while. And then out of necessity, I became a marketer. I was at a plant company, which I'll talk about in a second. And we had a product which sold like a dozen units, like 12, total worldwide. And I was like, this is a good product. How do I make it something that people know about? And then finally, I'm, by the necessity again, became a salesperson. If you're the co-founder or CEO of a startup and there's only two of you, guess what? You're in sales. You are always selling something. You're selling to yourself that this startup is worth doing. You're selling to your co-founder that they want to do it with you. You're selling to your employees to join this unknown company. And that's before you even get to a single actual customer. So the hardest part of any startup is that nobody knows you exist. There's always this myth that you're going to come out of the gate with this brilliant app. You're going to blow up at Slush or South by Southwest. The reality is that you're going to come up with an idea and absolutely nobody is going to care at all. Like if it's something that anybody was already doing, somebody else would already be doing it. And even worse, no one cares. You know, you figure out how many boosts there are here. It's really hard to crack through the noise. The good thing is, all you have to do is find one person to talk to you and you have a potential buyer. So let me talk a little bit about Easybloom, which was a plant sensor company worked at in the 2000 era. So this is literally a picture of what it looked like. I joined as the first product manager after it had already been built and was starting to ship out to users. Basically, it was a sensor with a light on top and the part underneath the ground measured the soil. So it can tell you whether your plants were good too much sun, too much shade, didn't get enough water, or had too much water. And I thought this was great because I had killed every plant I ever owned. My mom got the green thumb. I joke I got the brown one. And I thought maybe if I was working as a company, I could finally keep a plant alive, make my mom proud. And also, as an engineer, I thought this was pretty cool tech. We shipped it at pretty much the absolute worst time in hindsight for a gardening product. So we shipped it in November. So if you look around, and even in November, it's not really a gardening time, even if you're not in Finland. The other issue was we shipped it in November of 2007, which was when the mortgage crisis started in America. So not only was nobody gardening, everybody was just worried about their mortgages. It wasn't putting a lot of money into plants. We had all these nice spreadsheets about how we're going to sell thousands of units. We even had somebody at standby in customer support back then just to pick up the phone if somebody called. The issue was that nobody called. Our actual sales was around, I think, 12 that first month, and just did not move at all. This was really painful because this was our actual physical product. So we literally had like 10,000 of these in a warehouse. On top of that, we made the ill-fated decision to make the box compostable. So the issue is if you have a compostable box and it's sitting in a warehouse, it starts to compost. So we really wanted to get rid of these units. So I just started talking to people. We got a lot of really angry comments on some of our posts, and I listened to them. And the people who didn't like it said stuff like, I don't need a tool to tell me the soil is dry. I could just touch it, literally, touch the soil. Other people said, I don't want technology getting in the way. And people just thought the price point was too high. They thought it was $69, which, to be fair, is a lot of money. We'd done all the pricing based on what the profit we wanted to get was, and $69 seemed optimal. OK, so some quick changes we made even after we shipped was we'd lower the price from $69 to $49. $49 is a psychological barrier in a lot of people's minds, where it's just easier to spend money. If you ever look at a lot of pricey for consumer products and they're all $49 or $99, that's why. It's just this weird, weird, weird thing in people's brains. The other thing was to really focus on not the gardeners who wanted to get better, because they hated this, but people who wanted a tool. So Brown Thumbs, like me, who said, hey, I really would like to not have all my plants look dead. We had this other persona I called the Gadget Geek, which was the people who worked at a tech startup and actually really enjoyed technology. I found out later that some of my friends were at Heroku, which was a startup back then, and they had six of them in the office. So I was like, you are my persona. People sometimes just have this one finicky plant that they really wanted help with. And finally, the persona that cracked the nut to move a lot of sales was parent presents. It turns out nobody knows what to get their parents as a present. Nobody. You're like, so we could buy up a lot of Google keywords for stuff like gardening mother present and have a $49 price tag. Boom, done. So this is an example of how even after you ship something, you can still do a lot of clever things with packaging and positioning to change the arc. All right, so let's get a segue into the company that I started. Easybloom was, I was employee number eight, but I'd come in long after the idea had already happened. So LaunchDarkly is feature management. What we do is we allow customers all over the world. I said we have 4,000 IBM, Atlassian, and many startups here to ship out a feature to a user and then without having to run a new release, turn it on and off, and then also do a lot of stuff around segmentation. For example, you can have a different functionality in Copenhagen and Helsinki and do this in milliseconds. It's a true game changer, I think, for the way software is developed because it allows people this freedom of separating out deployment and release and able to quickly react to the field. So I had come up with this idea because I had used platforms like this at other companies and it was always something that we had to build in-house and I just thought that it should be an actual product, it should be a company. So the first person I had to convince was my co-founder and he had worked at Atlassia, he was like, absolutely. And we then had this long phase of trying to talk each other out of it. We're like, why does this already exist as a product? This is so obvious. This is absolutely something that every software engineer, every product manager wants, loves, needs. We couldn't talk ourselves out of it, so we started LaunchDarkly. And again, we had this blind hubris that this idea was gonna be so interesting that everybody was immediately gonna use it. Like I had these spreadsheets of, okay, we have our landing page, we're gonna get 10,000 people land, you know, 100 gonna convert to trial, that this is gonna happen every day, boom, done. What actually happened was, when we started, absolutely, nobody wanted to use it at all. So John, my co-founder, started building it and we had an early engineer and we would just scramble and scratch to try to get people to use it. We did the classic thing that every startup does, what you should do is I went around to every single one of my friends who were software engineers because I had been a software engineer and begged them. Yeah, please try this. And anybody who said anything at all interesting, I would say, okay, can we come to your office and install it? You don't actually need us to come to the office, but that way we know that you'll do it. So that was just the really first painful way of getting an awareness. So just begging your friends, good. We did a post on Show Hacker News. We started putting out our own blogs. We did a product hunt. We tried PR, we did talks, podcasts, articles for our message boards. And now I'm gonna pause and talk through how each of these went. Show Hacker News was actually great for us. We got a lot of feedback that people hated this idea. People who were very clear that this product did not need to exist. I could build that in two hours, it's just Booleans. And what kind of crazy people are they? And in San Francisco, what's in the water there? Why does this exist? But in all those comments, we also got what people thought a system like this would look like. So people were negative, but they'd say, hey, if this was a true production system, it has to be scalable. Or if this is an actual product, it has to have support from multiple languages. The Hacker News threads are still around. They're great if you just wanna go pull it up because in all the negativity, there was a lot of really good product feedback about, this is what a real system would look like. I started writing blogs about what feature management was. I made up the term feature management because I thought feature flagging made it sound like a toy. So I just started writing about why feature management was great. I went around and interviewed people in the industry who work at Netflix, worked at Google, worked at Amazon, and wrote up what their feature flagging systems look like. And then I posted these as like the secret of Amazon. The secret of Amazon is feature management. The secret of Netflix is feature management. So these articles were really good for just getting more distribution for us. We also had our engineers write blogs. And one of our engineers is a really good writer. And seven years later, his blogs are still amongst our best source of traffic. We did product hunt. This is another example where we had these lofty dreams that kind of went down to there. We, again, begged everybody I know to upvote us. Don't knock that. And we were, I think, number four for the day and had expectations that we were gonna get, you know, hundreds of customers again. From this, actually we did get one. One customer, which was Envision. And they turned out to be a wonderful customer and they gave us a ton of feedback. And so that one customer is worth it. I just started going out and giving a lot of talks on what feature flagging was. I went all over the world. I went from Oslo to Sydney, just talking about feature flagging. I also did a podcast with my friend Paul from CircleCI and just placed some articles with places like InfoQ, answering questions on Quora and message boards. So from all this, there was a steady stream of feedback of, well, I built that in two hours. It's bullies. No thanks. We've already got one. But from this also I started really collecting a lot of stories. Like if somebody said I already have one, I'd say tell me about it. You know, what do you like about your current system? How does it work? What do you use it for? What do you not like about it? And from this user research, I got a lot of really good ideas. It turned out that the people who had homegrown systems were sometimes really sick of those homegrown systems. Like they'd say we have this system in-house. It breaks all the time. It's brittle. I know we need it, but I'd actually prefer a vendor. So we could, from just from listening to the people who didn't want to buy, learn a lot. So I had a friend at Rainforest QA and he said, well Edith, we have our own system. And I said, great. Can I look at it? And so he sat down and walked me through how he used it. And from that, I came up with a lot of new features. Like he said, I can't use LaunchDark Police System because you don't have a view of users. You just have a feature view. I said, oh, you're right. Let's go build a user view. Let's go do that. Or somebody else said, you know, we have these really sophisticated segments because we like to control what customers see what, not just at a user level, but also combining different ways of looking to users. Great, went and built that. Then we started really uncovering the harder and harder features. What we discovered was that a lot of folks had a system that had universal access. Universal access meant that, for example, a customer support person had the same access as a senior architect. This sounds scary. And we started hearing all these horror stories about, you know, people not understanding what feature flags did. So a feature flag actually allows you to control your functionality. And so if you have somebody junior who logs in and can turn off a major chunk of functionality, that's really scary. The worst story I ever heard was it was a big content management system and somebody toggled a flag that actually turned off uploading. So for about eight hours, nobody could upload anything. And so everybody was just trying to figure out what was broken before they figured out that somebody just accidentally toggled this. They said the fix was just a label and very large letters. Never touched this button again, which worked for a little bit. So we could then, again, a lot of sharkleys start to build the harder stuff. We could build roll-bakes access. We could build audit logs. We could allow you to manage that in a really contained way. So we were getting parity with homegrown. And then we were also getting even more, getting beyond that. And then we did something which I call the hammer technique. So this is an actual hammer. So I got an engineering degree at Harvey Med College and one of the things that we had to do was build a hammer. And not just any hammer, like they literally gave you the schematics and a chunk of wood and a chunk of metal, a chunk of plastic and you had to build it to tolerance. And the really hard part of it was that if you heat treat the metal at the top, metal actually shrinks a little bit. So you'd have milled it precisely so that it would be with intolerance. And then when you go and cook it to heat treat it, sometimes it would shrink too much. And then you would say a lot of stuff that I won't repeat here and start over. So this is not actually my hammer. This is my friend's hammer. I passed the class, but it was so disgusted with this whole hammer situation that I just threw into the trash. I'm like, I'm going to Home Depot. So I think there's two types of engineers. There's the engineer who says, I want to build a hammer. And there's an engineer who says, I want to go to Home Depot and buy a hammer and then build a house. So we got really good at LaunchDarkly about trying to tell people that they had better things to do with their time. And one of the ways we did that was that we just, we pretty much just published our blueprint. You know, at this stage, LaunchDarkly was serving around a billion features a day. And we told people how we were doing it. We told people that, okay, this is actually a complex system. You have to think about scalability. You have to think about a lot of languages. You have to think about role-based access. And this is how, if you're going to do a homegrown system, this is how you should approach it. Because we sold to engineers, we didn't try to, you know, lie or make it sound griner than it is, but it is actually pretty hard to do. And we said, if you want to build this in-house, please do so. Here's our blueprint. Have fun. My other secret thing here was the person who's tried to do something themselves is much more willing to pay money for it the next time. Like if you've ever tried to do a home improvement project, that's kind of not gone as well as you want. The next time you're probably a little bit more likely to hire somebody. So we were very good about just telling people what we'd done. You know, we were getting more and more scale. A thing to look out to when you're selling is pride. As you can tell, I'm a Monty Python geek. This is a, you know, that he's fighting this Black Knight and the whole thing of the fight is the Black Knight is just getting, you know, bloodily cut apart. And at the end, you know, he's literally not able to fight anymore and he's like, come back, I'll bite you. You know, I'm still not done yet. So if you're talking to somebody and trying to get them to buy your product, never appeal that they can't do it. So we would try this in the early days of, oh hey, it's really hard, you can't do it. And it would always backfire. If you tell an engineer you can't do this, the engineer says, of course I can. It doesn't matter what it is. They say, just because you said I can't, I'm gonna go say that I can. Again, I'm an engineer. Somebody once told me I couldn't memorize the periodic table of elements and I said, well, yes, I can. Don't know why, but I did. So always watch out for telling people that they're not smart enough to do something. Another thing to watch out for is religious conversion. And this is really hard because if you are passionate about your product, you of course wanna persuade everybody to do it. So we would get in these sales cycles and we would realize that people had never feature flagged ever. And we would start trying to convince them that they should be feature flagging and these sales cycles would never close, like never. So if you're ever in a situation where you have a new product and you're just trying to convince somebody to completely change their way of life, please be careful about that one. What actually really worked for us was trying to find people again who already had a homegrown system and just converting them. So one of our filtering questions used to be, have you now or have you ever feature flagged at a prior job? And this was really good to filter out the people who said yes, I did, I liked the my system, I just don't wanna rebuild it at my new job or I'm brand new and this is gonna be probably a multi-year sales cycle. Another anti-tactic was to get people to move off a better system. I visited a supermarket in Australia and they had a system that they built from the ground up, like basically from the 70s. And it was so cool. Like they had done all this stuff that was custom to Australia and to the supermarket and to the tax laws there and I looked at it so they were kind enough to give me a demo and I'm like, your system is better. You know, you don't need yours. Now I'm happy that about four years later that supermarket chain is coming back to us because we're now at a place where they think we're better and they're tired of maintaining their system from the 70s. All right, here's some really good tactics that you can use when you're trying to think about what to build. If somebody has a repeated objection, you should probably just write a blog post about it. So an objection we got a lot was that feature management would add technical debt. So I literally wrote, so there was this, if you used to Google for feature flagging, the top result was feature flagging is the worst kind of technical debt. I did not like that being the top result. So I wrote an article about how feature flagging, if done poorly, is bad technical debt. So taking what people saw as this disadvantage of feature flagging, making a big collusion of their code and saying, yeah, if you don't have a system that manages it, if you don't have a scalable system where you can understand who's using it, yeah, this is really bad technical debt. But if you have a system where you can actually control and measure, it's good. So technical debt was a real thing that we could help people with. We found that people who had these homegrown systems actually did have a ton of technical debt and they responded really well to this message. Another thing we used was telling people here's how not to feature flag. So if you're doing a new field, you could again take everything that's bad and make it into something good. So we told people, here's the right way to do it. Here's some best practices. And we also even had a website called featureflags.io where we could just explain more. So I think the thing I hope you're hearing from this is early days are super tough. You're gonna hear a ton of feedback that what you're doing isn't valuable and that you're doing it wrong. The key is that these people are actually engaging with you in some way. If they're talking to you, they're interested. And if you have a real opportunity, if you address them respectfully, to actually turn them into converts. Instead of having a flame war, you can actually make a bonfire of good product. All right, so what really worked for us was just again converting all these folks who had feedback into fans. It's really funny, now I go to conferences and one of the things I love most is just running into our customers. When we started, we added literally one customer a month for about four months and I knew every single customer and they knew me because there was no alternative. Now, as I said, we have 4,000 customers worldwide and they have all these cool use cases and they come up and they tell me about them. And the best part about it is every single use case they tell me about whether it be managing a mobile release, maybe they'll also have stuff in different GOs. I'm like, yes, this is all the stuff that we thought about when we started the company eight years ago and it's real now. Like all the diagrams we sat down and wrote, all the use cases that we planned are now actually in existence and that feels really good. So I hope the takeaways that everybody has is early days of startups are super tough. I don't think there's any such thing as overnight success. There's a long, long, long period where you're just gonna have to grind but the benefit is that if you continue every customer you have compounds. Like so I have customers who have started using LaunchDarkly six years and one for example was an IBM and then went to NCR, National Cash Register. He's still a customer. So the groundwork you lay in the earliest age is about listening to people, being attentive to feedback and getting them happy will continue. Thanks everybody.