June 27, 2017
CircleCI Office Hours: Everything You Ever Wanted To Know About Continuous Integration
In this talk, CircleCI's Alek Sharma shares everything you always wanted to know about continuous integration - watch to learn how continuou...
In episode 5 of EnterpriseReady, Grant meets with Edith Harbaugh, CEO and Co-Founder of LaunchDarkly. They talk about Edith’s experiences building advanced enterprise features, and how enterprise sales have changed over the years.
About the Guests
Edith Harbaugh is the CEO and Co-Founder of LaunchDarkly, a feature management platform serving over 100 billion feature flags every day. She is also co-host of the podcast To Be Continuous, and was previously a product director at TripIt.
Grant Miller: All right, Edith. Thanks for joining us. I appreciate it.
Edith Harbaugh: It's great to be here in Vegas.
Grant: Yeah, looking out over the Vegas-- Well, landscape. Not really the strip.
Edith: The beautiful Vegas airport.
Grant: Yeah, exactly. To get things started off I'd love to just have you give the audience a little bit of your background, and how you got into enterprise software.
Edith: I'm Edith Harbaugh, I'm CEO and co-founder of LaunchDarkly. I graduated with an engineering degree in '99, which was the height of the dot com boom. What you did back then was you moved to San Francisco and worked at a dot com, so I did something that was even more derivative than that, I worked as a consultant for dot coms.
This is back then when you'd get two Stanford MBAs that get a couple of million dollars for their idea, and then of course they had to have engineers to build it. So I worked for a company called Xpedior which was basically that we would come in and we'd build your product. This was a really viable career until suddenly it was not.
Grant: When all the VC money dried up?
Edith: Yeah. And that really happened, literally, within two months. I remember us all being fully staffed because we were by the hour, and I remember somebody walking down a hall being like, "I have no bench. Everybody is out on the field. There's nothing we could give you," and then two months later it's like, "We don't have any jobs." All these VCs had stopped funding companies, the companies didn't have any money, and the company literally went from listed on the NASDAQ to delisted.
Grant: Whoa. So, that consulting firm--?
Edith: Completely went under. It went public, I think it was public for two years, and then it was delisted.
Edith: It was good they got delisted, because the name of the company was Xpedior with an "X".
Edith: Expedia had actually filed a lawsuit that they had won, and they were in the middle of rebranding, and then it's like "We're just going under anyway, so it doesn't matter."
Edith: That was my brutal introduction to-- It's really good to actually make money.
Grant: Yeah. That's the beauty of enterprise software.
Edith: If somebody is paying you for a unit of value, that's far better than you depending on VCs giving you another dollop for Pits.com.
Grant: OK. Out of college, dot com boom, went to SF and took this job. Then how did you--? What was the path into true enterprise?
Edith: It was basically as soon as I got laid off, because they shut down our San Francisco office in the process of shutting down the company in November of 2000. I was looking around for a new job, and I really wanted to stay in San Francisco.
Edith: Back then all of the jobs were down in the valley, like they were all in San Mateo at best but really down in Palo Alto, or Mountain View.
Edith: I didn't own a car, so I had to borrow a car for an interview, and I went down for the interview and they offered me a job, and I was like "I don't wanna drive down here every day."
Grant: Yeah, or live down there. Right?
Edith: Yeah. I was a kid, I was 21, I didn't want to live in Mountain View.
Edith: So, it was by default. I knew I wanted to stay in San Francisco, and now it's like--
Grant: There's tons of companies.
Edith: Back then there really wasn't much, it was so-- A friend of mine had gone to a company called Epicentric, and he said "Come over, join us." So I just fell into enterprise by default because I needed a job. At the time he joined as an engineer, and I was a software designer. A spec writer.
Edith: I would design all the systems and write up how they would work. It was a portal management system, so I wrote up all of how we would do delegated administration, and how we would deploy out and bundle up different systems. A lot of the work really held up. I worked with Ed Anuff was the founder of the company.
Grant: I love Ed.
Grant: That's amazing. He went on to--
Grant: Apigee, yeah.
Edith: He was at Six Apart, so he was-- I was the young kid who did the grunt work of actually writing up all the specs.
Edith: We ended up getting three separate patents together.
Grant: That's awesome. It's funny too, how the network persists through the rest of your life. You see Ed at events and things now, still, many years later.
Edith: Yeah. It's funny because the stuff we wrote, people still use.
Grant: That's cool.
Grant: What was the software, what did it do?
Edith: It was for managing portals. It's funny, I was at a dinner last night and somebody had worked at Disney, and they still use some of our software because it's portal management. You could write portlets, which is basically a micro service as you would call it now, and Disney wrote their time card management as a portlet. Disney has 50,000 employees just in Florida, and they still use it for that.
Grant: But there is no company behind it anymore? It got sold, or--?
Edith: Epicentric was acquired by Vignette.
Edith: After I left Vignette, Vignette was acquired by Open Text.
Grant: OK, and then that still--?
Edith: Open Text is a public company in Canada.
Grant: OK. So, the staying power of enterprise software with these contracts, I guess. Right?
Edith: If you have something that's delivering value, people keep using it.
Edith: I remember I would go on site, like the IRS was a customer of Vignette. NASA was a customer. I mean, big companies doing real things.
Grant: Yeah. I think the idea for why enterprise software companies persisted through the bust was because they had these large non dot com customers that actually could continue to pay them.
Grant: So unlike the portal, like web portals that were building all their business off of this VC dot com money that was circular and never really left the ecosystem, an enterprise software company selling software to Disney that's been around for a long time, or NASA, that's been around for 50 years. Those are customers that didn't go out of business.
Edith: I loved visiting the IRS because people there were really happy.
Edith: Yes. They knew that they had a job.
Edith: They could plan these 10 year cycles.
Edith: They're like, "OK. We're going to do this next year, and then this--" So they could have a 10 year roadmap. We would go to another customer, Genentech, and everybody is miserable. Like, "Maybe we'll do a layoff next quarter, maybe not. I don't really want to plan too far in advance." IRS people are happy as a clam.
Grant: And they know they're not going to get out of it.
Edith: I don't know about that.
I loved working with the IRS. I loved working with NASA. We had a lot of really big brand name customers using us.
Grant: At that point you were writing specs for how the software should work, and it sounds like some of this was true enterprise functionality. It sounds like the thing you're describing is similar to ARBAC.
Edith: Sort of, yeah. Delegated administration. Actually, multiple levels of it. A lesson I learned the painful way, is we built this elaborate system and nobody actually used it. It turned out that people wanted a much simpler ARBAC implementation, and we'd went off the deep end of "What if you wanted to share this component to this person, and then they want to re-share to this person but not to this person?" And we couldn't even build a UI to express the API at some point.
Grant: Yeah, sure.
Edith: Then we go on site and people are like, "I just want to put the picture of the horse on the website." So we had built all these elaborate rules about who could do what when--
Grant: Got it.
Edith: That just got in the way, and people were like "I just want to publish this picture."
Grant: Yeah. So the internal joke became, "Let them put the horse on the website," or something?
Edith: Our CEO at the time was like, "We built a mission to Mars without a coffee maker, and people just want that goddamn coffee sometimes." We had overengineered everything and didn't solve the really simple use cases. I think that's why Vignette got subsumed by Blogger and WordPress who did a really good job at really simple use cases.
Grant: Yeah, and over time WordPress has become a pretty massive enterprise software company.
Grant: That's part of that consumerization of IT movement. Creating experiences that are easy for people to use. Now, you might need that advanced functionality in there, but maybe don't expose it to all of your users. Keep it behind a feature flag maybe, or something.
Edith: The lesson I learned was, "Make this simple stuff possible, and then also have all these advanced features. But if you don't have the simple stuff--"
Grant: At the core, yeah.
Edith: Yeah. At TripIt we had a saying, which was, "Time to bacon."
Edith: TripIt, where I went later, was a mobile app.
Edith: Mobile apps are pretty brutal. People download them, half of mobile apps that are downloaded don't get opened once.
Edith: So if people open their app, you need to give them-- We called it bacon. Like, something good very quickly, or they will never open your app again. We obsessed over "What nugget of goodness or strip of bacon can we give people in the shortest possible time so they will reopen the app?" I think the same thing holds true to enterprise software now.
Edith: When I started out in 2001-2002, people would sign these three year enterprise wide site licenses and you're just stuck with whatever your CIO bought. Now with the rise of SaaS, people sign up, they better like it very quickly or they're just going to never login again and let it lapse.
Grant: Yeah. That idea that someone signs up and they're using the product, and really they only ever pay a large enterprise wide license agreement once a bunch of people are using it and they know that the utilization is there and the adoption is there, and that the value is there. Then they're coming to you looking for, "How do I get it to the rest of my team, and get value and get control?" But at the start, it's like you gotta get those users to just use-- Get them some bacon.
Edith: Right. Because otherwise it's like, everybody does this. You're looking for bookkeeping software and you're like "I couldn't figure out this one, so let's not go with that." You only have one chance to make a good impression.
Grant: That's a great point. The amount of research people do, especially with the ability to just sign up and try things out and be like, "I couldn't figure it out. It didn't work. Move on."
Edith: Yeah. SaaS completely disrupted the way software used to be sold. It would take 2, 3, 6 months just to get something installed. That's why you had a POC of "Let's even try to get this figured out in your system." Then once it's installed, it's like "It's good enough."
Grant: Bring in the systems integrators to try to hook it up into the rest of your systems, spend a bunch of money there, and then even scheduling that and getting the IT systems also.
Grant: It's a huge time suck.
Edith: What SaaS changed is that the end users are suddenly back in control. Because they didn't have to go to IT and say, "Can you provision a box for me?"
Grant: Right. I think that there's a combination there. It's both the introduction of that bottom up SaaS where you could just sign up and start using it, create this shadow IT thing where no one really knew you were using it. But I think there's also a really important piece which is just the interoperability between applications and the existence of APIs. I think Salesforce still charges to access their API. It's not available at their base level plan, which is kind of crazy, most modern software companies that are started today an API is as common as signing up for a user. You can create a token.
Edith: Salesforce is worth-- They're going to do $10 billion in revenue, so I'm not going to--
Edith: I'm not going to tell Mark he's doing it wrong.
Grant: I think it's just part of how they've always done it, and it's just been how they differentiate their products. Salesforce is also very lucky to be the system of record, so everyone has to integrate into Salesforce.
Grant: That creates gravity for them. So if you're another SaaS application, when you start out no one is going to integrate with your stuff. So you need to have an API and then you need to do some external integration, you need to make that API as available as possible. You just want everything-- You want every integration possible. I think that's a little different today than it was 20 years ago.
Edith: I don't know. I think people sometimes feel like APIs are a magic cure all.
Edith: We had an API at TripIt and the only people who really used it were honestly people who were trying to rip us off, like we're trying to skin it and make a TripIt replicant.
Grant: Yeah. But you were-- It was a consumer company, right?
Grant: So the reason I think that APIs are important for enterprise software is because it lives in this whole world of other applications. It's not just like, you have one app.
Edith: Yeah. I think the mistake many people make, or maybe not mistake, but the learning is to be very deliberate about what is API-able.
Grant: OK. What do you mean?
Edith: It's not just like sprinkling APIs on something, it's like, "OK. How do people want to integrate with my system, application, program interface? What are the functionality that they're trying to get in and out of my system?" Make that very good and not just like, "Let's API this thing over here that nobody cares about."
Grant: Sure. It's an interesting perspective to say, "Really think about what should be available in your API," and I think you have to design it from your perspective to make sure you're thinking about what that use case is. "How will someone use this?" Because maybe you need to have additional fields that you can have initially in order to have a global locator, for them to have a different record and integrate with different systems. You have to think about it globally, not just about your application specifically.
Edith: I think you put it very well. "How does my application interact with the rest of the world?" Because nobody is a silo anymore.
Grant: Exactly right. I think that's maybe the difference between that consumer world, where most of my consumer apps don't really talk to each other and I don't share data between them. My enterprise stuff, if I can hook it up to creative workflow or pipeline I'm gonna do it.
Grant: Then after TripIt, that's when you went to Atlassian? Is that right?
Edith: My co-founder worked at Atlassian.
Grant: That's right, sorry.
Edith: My co-founder John and I met at Harvey Mudd College in LA.
Edith: He went to Coverity and Atlassian, and we had used feature flags at our respective jobs. We used them heavily at TripIt.
Grant: Interesting. You both had been using feature flags independently, and stayed in touch about this concept.
Edith: We just stayed in touch because we're friends. It wasn't like--.
Grant: It wasn't feature flags bringing you together all the time.
Edith: No, we were friends. We went on camping trips, we went to Coachella. It's funny, we were working on another idea around static analysis which he'd gotten his PhD in. I was talking to a lot of developers about static analysis and I kind of realized that I don't think there was a big enough market for us to do a company. But I had been talking to these developers and they were all talking about how they liked feature flags, and they're like "Why do we have to keep rebuilding the same system that we had at our prior job?" LinkedIn had a pretty mature system, people would leave LinkedIn, and people would say "Why are we rebuilding this?"
Edith: At TripIt we had our own internal system because we had 10 million users. We didn't want to launch a feature to all 10 million.
Edith: So we would do a lot of stuff where we would launch it off and then turn it on for selective users who were friendly, who we knew would give us good feedback, and then we would do percentage roll-outs. Also if something went bad in the field we could just shut it off. So, we used it extensively. John had used it at Atlassian, and I said "John I think we should build a system to do this." And John said, "Yes."
Edith: Because there is two questions, "Is there a market?" And "Could John build it?" And John's like, "I think I could build this." And I said, "Great. I think there's a market."
Grant: Your other half is, "I think there's a market."
Edith: Yeah, because static analysis is a hard road.
Grant: Sure, yeah.
Edith: Some people are making a success out of it but I didn't think it was a big enough market for us to really go after.
Grant: But your thesis was that "Everyone is going to build software, and so they're going to need this as a way to to do a modern deployment pipeline."
Edith: Yeah. I mean the joke is I'd worked at Monster, the job site, and even at the time I was frustrated that there wasn't a tool like LaunchDarkly.
Edith: My old boss Antoine says, "Edith you always used to complain that we had to build this in-house. You finally just made a company."
Grant: You hated this problem so much inside of all these companies you worked at, that you had to just solve it for everyone.
Edith: Yeah. Because I remember I was like googling "Why can't we just buy this?"
Grant: Good thing you couldn't.
Edith: It just seems so obvious in hindsight.
Grant: You and John were discussing this idea, and this is what, four years ago? Five years ago?
Edith: This was in June of 2014.
Grant: So, you were talking about it, said "We think this is a business." Just the two of you to start?
Edith: It was the two of us. He was Atlassian, so at this point Concur had bought TripIt.
Edith: I had also gotten all this enterprise experience from Concur.
Edith: TripIt was the classic consumer mobile app, and Concur bought it for a good price. I think it was around $100 million. One of the reasons they bought it is I joined TripIt to start up their TripIt for business.
Grant: OK, cool.
Edith: Which was us selling services to
businesses, and that got Concur's attention.
Before they saw us as this cute app, now they're like "They're going after our core market of business." So they acquired us. I was a part of why we got acquired.
Grant: At this point you were doing product or engineering at TripIt?
Edith: I had been a product manager there, and then Concur acquired us, and Concur is brilliant at enterprise sales.
Edith: I mean they are a juggernaut. They got bought for $8 billion by SAP.
Edith: Because they are very good at enterprise sales.
Grant: What makes them so good?
Edith: They really know their buyer. They sell to CFOs and they have a very good pitch to CFOs, and their entire sales force goes and tells a CFO "This is why you should buy Concur," and it becomes this inevitable thing. People buy Concur because other people buy Concur.
Grant: Yeah. It's like, "You don't get fired for buying Concur."
Edith: It just becomes a self-reinforcing thing.
Grant: I had to use it when I joined a bigger company five or six years ago. I wasn't a big fan.
Edith: Yeah. As I said, they're very good at enterprise sales.
Edith: They know their buyer, which is the CFO.
Edith: The users--not so happy.
Grant: That feels like it's becoming less common, but it's still fairly common in the enterprise software world. That end users don't have as much of a say. Probably far less true in the developer tools world, where I think end users really do have a lot of say, but a CFO picks a technology because it--
Grant: Great reporting, and all these other things. The end user experience is an afterthought half the time.
Edith: Look at Salesforce.
Edith: They sell to the VP of sales, not to the ICs.
Grant: I'm not a big fan of the Salesforce experience, I still remember when I joined that larger company we used Salesforce. I was staring at a dashboard for about three or four months and the sales reps, I was in product, and sales reps were telling me they were selling more of this product than I was building. I couldn't-- I was like, "Why is the dashboard, why are we not seeing any any sales here?" Turns out there's a refresh inside of the dashboard. When I refresh the page it's supposed to refresh, and they're like "No. You have to go to this special button and hit refresh."
Edith: Salesforce, like I said, they make a ton of money but their user experience is very dated.
Grant: I think Lightning is better from what I hear, but it's ultimately-- It just comes down to the fact that there's an inefficient market in enterprise software. This is why you can have huge commissions as well.
Edith: It's inefficient or efficient, depending on how you look at it.
Grant: OK. How is it efficient?
Edith: They're very efficient at selling to their buyer, which is the VP sales.
Grant: Sure. It is efficient from Salesforce perspective, it is inefficient from a market perspective, in my opinion.
Edith: That the market would have better--?
Grant: Yeah. I also believe that any market with large commissions is inefficient.
Grant: That's just a general thesis because if you think about brokerage, 20-30 years ago before information was freely available, stockbrokers and bonds brokers had information that other people didn't have. So they were able to create this spread, and that spread is commission. So I think you see the same thing as there's a huge information asymmetry in inefficient markets. As the information becomes more symmetrical, commissions shrink, the bid/ask gets closer, and you end up not needing huge commissions. So I think that's a sign. But that's just a random aside.
Edith: I'll random your random aside.
Edith: My sister lives in New York and the idea of broker commissions on apartments, I just can't wrap my head around it.
Edith: So when she moves, she has to pay two months rent to a broker.
Grant: It's just the market there, like everybody does it. It persists itself. Sometimes it's laws, like in California you have to be a real estate broker and you have to do two years of studying and have all these hours and then they end up persisting these entrenched systems.
Edith: I'm going to stand up for commissions.
Edith: And sales.
Grant: Yeah, great.
Edith: I'll say I had a mix of experiences. I'd worked at big enterprise companies like Vignette. Vignette, we did not take deals under half a million, because it wasn't worth the cost of sales.
Edith: Because we had to do a POC, you had to send people on site, and then SaaS came along and changed everything. Then I was also at Concur, but even then you had sales people.
Edith: Atlassian famously does not have sales people.
Grant: Right. I say it with air quotes.
Edith: When we started LaunchDarkly John and I had this idea that we would be this really bottoms up, feature flags for all, self serve thing.
Grant: I think every technical founder has this dream about their software company. We did it at Replicated, too.
Edith: We made this really nice self-serve checkout path.
Grant: Integrated with Stripe, or something.
Edith: Yeah. It was all ready to go. Then our first big customer, who by the way sat next to me at our dinner last night which was awesome--
Edith: He wanted a contract because he worked at a big public company and they needed a year contract, and they didn't want to do self-serve. I kept saying "Just put in your credit card."
Grant: Right. Easy.
Edith: He's like, "No. The way our company works is it needs to go through legal and security and procurement." And I'm like, "Just put it on a credit card." And he's like, "No we need a year contract." So that was my brutal introduction to, we were actually enterprise sales.
Edith: And that you need a salesperson a lot of times to actually help you navigate your own bureaucracy.
Grant: Yeah. Point taken. I'd say that potentially the commissions can also be attributed to the fact that these are not commodities that are being sold. So the commoditization of technology would also force down prices and force down commissions. But one software solution versus another, they're not interchangeable. They're not wire compliant. Some has different features, different security models, different things. So in that case a complex sale in a complex application could just create the need for larger commissions.
Edith: Even beyond that, our own experience will have a champion at a company, and they're like, "I love this. Can you come and give a demo to the rest of my team and get them on board?"
Edith: I'm like, "That's a salesperson."
Edith: The hilarious thing is that we sell to developers, so they'll say "We need somebody to come on site and give a demo and work with procurement, but not a salesperson. Because we hate salespeople." But, that's a salesperson. We'll call it another name, but like--
Grant: That's what they do.
Edith: They're like, "Can you help? Send somebody on site to help us understand your software and see the value, and convince other people, and work with my purchasing?" And I'm like--.
Grant: You just described a salesperson.
Grant: Yeah, it's funny.
Edith: I think the word "salesperson" has gotten a really bad rap back from the early days I talked about it, like 2001 where you couldn't see the software for six months. They were kind of flim flam PowerPoint chicanery, for lack of a better word.
Grant: That's interesting. So your perspective there is that the role of the salesperson has changed. I'm guessing-- That's actually an interesting question. In your sales-- I think about this, I differentiate between companies sometimes in the enterprise software world as "sign up," or "contact us." What's the main call to action on your website?
Edith: We're still doing a hybrid because we sell to developers.
Some developers are just allergic to salespeople. They are like, "Just leave me alone. I will put in the credit card."
Edith: And others are like "Please. I want a rep." And then they'll write again like, "The rep hasn't got back to me quickly enough. Please. I want a rep." The nice thing is people love our salespeople. Not to brag too much about LaunchDarkly, but--
Grant: No, go ahead.
Edith: They see us as a trusted advisor. We went to New Zealand to visit customers and the customer had a gift for our salesperson.
Grant: That's great.
Edith: Yeah. He got him a really nice New Zealand gin.
Edith: He said, "Thank you so much. You've helped me so much. Here's a present."
Grant: That's awesome.
Edith: Which is not how people think about salespeople.
Edith: We closed the deal with another big bank in Australia, and I'm the CEO, so I had just literally signed the contract and I got an email from the customer. "Your salesperson was amazing. He helped us so much. We're really happy, looking forward to working with you more."
Grant: That's great.
Edith: Yeah. I think salespeople have been tarred with this awful image of pushy people, instead of "This is somebody who could help you."
Grant: I think that's what gives us as technical founders a bad taste for sales. You're right. Maybe it's just a sign of the maturing model. Now we can show you the software faster, we can help you understand we're actually trying to communicate value and communicate how this fits into the rest of your tools and how this should work, and there's education. The role of a salesperson--
Edith: It's not to trick you.
Edith: It's not to trick you, it's not to hoodwink you, it's not to bamboozle you because people can cancel a contract anytime.
We have 140% retention because we're like, "We will sell you good things that you will like."
Grant: You made this transition, first had it as a sign up flow and you could pay with a credit card. Your first customer wanted a contract. Did you go off and hire a salesperson or did you do the early sales yourselves?
Edith: I did all early sales myself. I'd like to say it was from some noble urge, but I kept thinking they were one- offs.
Edith: Then I realized that I was the bottleneck. Literally, people would be like "We're ready to sign. Where is the order form?" And I'm the founder, and you know how it is. You're doing--
Grant: A million things.
Edith: Somebody is like "I'm really ready to pay you money, where is this order form?" And that's when we hired our first salesperson.
Grant: OK. Then how far into the company's life cycle was that?
Edith: It was about two years.
Edith: I did basically the first million in sales, I became the bottleneck, we hired somebody.
Grant: At this point, you had raised, how much money?
Edith: We'd done our seeds. We'd done a $2.6 million seed round, we'd been selling for about a year. It took us a year to basically build anything sellable.
Edith: We had people using it, but it was about a year before anybody would really pay us money for it.
Grant: You were bringing on these early customers, getting them to try it out and use it and give you feedback, and eventually you felt like you had something that you could charge for and people would pay you for. You were selling those deals for the next year.
Grant: Then got up to about a million and realized, like--
Edith: I'm the bottleneck.
Grant: Yeah. What did the team structure look like at that point? Was there mainly engineering, was their product, was there marketing?
Edith: We were an 8-person company.
Grant: You were an 8-person company? OK.
Edith: The department thing sounds a little bit grand.
Grant: I think about it like, "Did you have someone who was responsible for marketing?"
Edith: Yes. The joke is I was our first AESDR. John Kodumal my co-founder was our CTO and SE.
Grant: OK. Sure.
Edith: So, the early deals, you got us.
Edith: Whether you liked it or not.
Edith: It's really fun now because now we have over 500 customers, but the early ones, they know me and John and I know them. We talk to them.
Edith: We then hired two engineers, we hired marketing very
early. A designer was our fifth hire, marketing was our
sixth. I actually credit Tom
Grant: From Heavybit?
Edith: Yeah. I'm a good marketer, and I kept thinking I could just keep doing it. He's like, "Edith. You need a scale to millions of dollars. You're the CEO and doing a lot of things, you need to hire somebody do this full time." Except for Tom said this far more harshly with more British swear words in it. He's like , "You've got to scale."
Edith: So a marketer was our sixth hire, and I think that was really smart.
People sometimes say "I'm going to wait till my "A" and then hire marketing," those companies usually don't get an "A".
Grant: Somebody has to do the marketing. Either it's gonna be you're doing it and you're putting all this effort into it, or you have to hire it in order to get there.
Edith: I was trying to do it off the side of my desk, and it's like "No. Somebody just needs to do this all the time."
Grant: So who was really doing product at that point?
Grant: So he was doing-- John was doing both CTO, product, and then sales engineering?
Edith: And support, and the joke was a chief t-shirt officer.
Grant: OK. So he had some marketing responsibilities himself.
Edith: Everybody did everything.
Grant: You hired someone in marketing before you had sales, did that marketing person focus on sales enablement collateral stuff? What was their--?
Edith: A lot of content marketing. That was what was working for us, so we doubled down on it.
Grant: Sure. I think that's a great strategy early on. So, content marketing around just--?
Edith: Feature flags.
Edith: We realized that there wasn't a lot of good technical content out there around feature flags, which was really good for us, because we could put out a lot of content very quickly SEO for it.
Grant: Sure. Who was writing the content?
Edith: Like, I wrote a lot of early posts. I would track how they're doing, I could see that we were doing well on Google. I knew that we would start to have competitors. I said, "We need to double down and write more content." As part of our weekly sprints, because we're an engineering company, everybody had a blog post.
Grant: Wow, everybody had a blog post every week?
Edith: We figured out who was a better writer, and if you were not a good writer you--
Grant: Got different tasks.
Edith: Our best writer was our lead engineer Patrick Keating, some of his blog posts are still just gold.
Grant: That's great.
Edith: We lavish him with praise and say"Please write more."
Edith: For a while every time a new employee joined we made them write a blog post.
Edith: This was our way of sussing out content.
Edith: We said "It doesn't have to be fancy, just after your first week write a blog post," and we could suss out "Are you a decent writer? Should we encourage you to write more?"
Grant: Sure, OK.
Edith: It wasn't a choice in the beginning whether or not you wrote the first week post.
Grant: You needed content and you wanted to know who was good. OK. I love that.
Edith: Then as time went by, now we're 70 people, we don't make people write blog posts anymore. But in the early days you're all rowing the boat together.
Edith: In the early days everybody answered support tickets.
Edith: It's like, "This is what we do."
Grant: No other option.
Grant: OK. Then you had some someone doing marketing, helping facilitate a lot of this content marketing, but were they also doing lead gen? I'm sure you have all these functions today. If you think about the maturity level, what were some of the functions that were fairly mature early on? And that were actually really drivers?
Edith: We were experimenting with everything. My theory was "Let's
just try everything and see if it works."
We tried paid advertising, it didn't work at all for us. B ecause nobody wakes up in the morning and says "I'm going to search for a feature flag banner and system." It's just not something people do.
Grant: Turns out, when you're creating a category it's hard to get people to search for that thing.
Edith: Yeah. So my theory was, "Let's just systematically work through every technique I knew from prior jobs and see what sticks." We tried outbound email, didn't work. Now, three years later, we do outbound email and it's very successful. But back then, it did not work.
Grant: Interesting. It's also an important piece, which is like, it's not just marketing. It's everywhere. You have to try things again, you can't just because it failed once.
Edith: That's what I tell the team. Now we're almost five years in, and I was like "Please. Just because we tried this thing in the past and it failed, doesn't mean it can't not succeed now."
Grant: Right. I have the same thing with food, with the foods I didn't like when I was younger. I make sure I retry them every five years to see if they're better now. Some things are. I now love dim sum.
Edith: There's this theory-- This is a digression, that your taste buds change as you get older. When you're young you love sweets. I remember eating Starbursts and Pixie Sticks and Skittles. Now it just seems way too sweet.
Grant: Yeah. That's funny.
Edith: When you get older your taste buds react more to bitter.
Grant: I just thought that I was maturing and my palate was getting more refined.
Edith: Well, literally.
Edith: The thing about marketing is also what worked at one company can get saturated. At TripIt we tried to do content marketing and it was just an utter flop, because people have been writing about travel since Marco Polo.
Grant: Yeah, OK.
Edith: Every newspaper has a travel section. People love to write about their vacation. There is so much travel content out there.
Grant: That's really interesting. That's a great point.
Edith: Versus technical content, there's just not a ton out there.
Grant: Yeah, and it changes. There's new technologies every time that change the game, and it's like "How you did continuous deployment, or how you did deployments change when you move to containers, and you move to serverless," all these things change. So, that's a great point.
There's a real hunger for good technical content. But it's harder, there's so many content factories out there that will pump out listicles for you, but not good technical content.
Edith: They'll pump out a listicle about top 10 ways to save money on your credit card bill.
Edith: Everybody has credit card bills. But if somebody wants to write how to--
Grant: How to deploy your enterprise software in air-gapped environment. It's a pretty hard listicle to write.
Edith: You can't ship that off to some freelance writer.
Grant: Which puts a lot of responsibility on your team, because not only do they have to develop it and deliver it and support it and sell it, but they have to also describe it and write it. There's a lot-- it's a big challenge, I think.
Edith: It's a big challenge, but also a big advantage. The early days of selling I remember we hired sales people and they were pretty green, and engineers can be pretty tough audiences.
Grant: Yeah, that's true.
Edith: They would get kind of beat up on sales calls, and I would say "You are the expert. They are looking to you for advice. They're asking you the tough questions because they are putting their jobs on the line to buy you."
Edith: "Literally. If they buy our solution and we go down, they get fired." Now I think that really helped our salespeople, because now they act like trusted advisors.
Grant: And they have to know it. You can't go in there and sound foolish and be trusted.
Grant: So you have to get the knowledge too.
Edith: Yeah, that's why I'm like "You are the advisor on how to use feature flags correctly."
Grant: Ultimately I'm guessing what happens is they just end up hearing things your team has been saying, or you say, or John says, someone-- They're just repeating the things that they've heard before. You can put together a lot of knowledge from just being around this conversation, and start to really understand these things, and then have a very intelligent conversation with someone by just delivering what experts-- Because this is all you've thought about for last six years, basically .
Grant: So your insights, the things that I'm sure you randomly say in the office or what happens in some slack channel, is the most advanced thought anyone's had about feature flagging. That permeates throughout the rest of the company and they can distribute that out to the rest of the world.
Edith: In the early days we were all about we all needed to be in the same place, so we went through Heavybit. When we hired our first salesperson, one of our interviews was you had to come eat lunch with us. Because we all ate lunch together every day. We're all filed down, and you've been to Heavybit.
Grant: Yeah, of course.
Edith: We'd get our lunches and we'd sit together. So part of the interview was you had to come in and get lunch with us. We were like "You don't have to eat lunch with us every day, we get it if you have a doctor appointment or just want to go to the gym, but--"
Grant: Sure. "We do this."
Edith: "We do this," and we had a daily stand up with everybody in the company. It wasn't just engineers, it was literally marketing, design, me and our salesperson would stand up and say "Here's our blockers and here's what's happening." It was really effective at indoctrinating the sales team, or the salesperson--
Edith: Into, "This is how an engineering company runs."
Grant: That way, even that, because that's the context through which your product is even being sold. Exposing them to this daily stand up, which is a ritual that many software companies you are selling to are doing, gives them an even deeper connection to that sale.
Edith: This is how our weekly sprint works.
Edith: Because then when a customer or prospect says "I need to wait for the right sprint to integrate this--"
Grant: You know what they're talking about.
Grant: You're like, "OK. Sure. How long are your sprints?"
Grant: You can ask the right questions.
Edith: The other thing I told the team was, we were really tight knit, and I said "I know a lot of engineering led companies hate their salespeople. They look at them as these idiots out in the field. We're not going to succeed unless we say, "This is our face in the field, closest to the customer, getting feedback and bringing it back to us."
Grant: That's great, and that's been true for your sales people? That's amazing.
Edith: Yeah. They're the people who are out in the field talking to prospects. They're the ones who are like, "That feature that we think is really cool, nobody cares about. This feature we didn't think anybody would want, everybody asks for."
So we have to treat our salespeople like they're part of the team, which seems dead obvious, but I've seen so many tech companies think that "We could just spin up the sales office over there and we'll build our product over here, and everything's gonna be fine."
Grant: That's great. So even though you didn't want to be sales heavy when you first started the company, once you realized it was important you really embraced it and made it a core part of your culture. Really, you probably softened the sales role in terms of not, you're not hiring the finger-gun, slicked back Lamborghini driver. These are people that understand technology and appreciate and love technology as well.
Edith: Jason Lemkin gave a great talk at Heavybit, I still refer people to it, and one of his roles about selling to technical people in the early days is "Would your team buy from this person?"
Edith: If not--
Grant: Don't hire them.
Edith: Yeah, because they're going to go sell to technical people like you.
Grant: That's very true. Still, I think you need a different persona in sales than the engineering persona. You need someone who's like, I always say "You have to be just slightly shameless."
Grant: You and I as founders are slightly shameless, because--
Edith: I saw you handing out Replicated hoodies when you shouldn't-- at places where you shouldn't have been handing out Replicated hoodies.
Grant: Yeah, exactly.
Edith: I'm not going to say where it is, but--
Grant: Everywhere, basically. I've got a bag of like, eighty of them.
Edith: You had your then-girlfriend handing them out too.
Grant: Yeah, in Hawaii somewhere. She's my wife now.
Grant: That's why she gets to enjoy the life of Replicated, but that's it. You have to be somewhat shameless in order to expect that someone doesn't mind you calling again, or doesn't mind you sending another email or saying-- Because for me the perspective is always "I know that the software that our team builds is so great, and it's going to be so helpful that I just know you're gonna love it when you start using it."
Edith: Yeah. I believe that. Our net promoter score is in the 40s and 50s.
Edith: I know that people get a lot of value out of it. That's why I'm like, "I know this is going to help you."
Grant: You just have to have that little bit of, "Yeah. I'm a little shameless," because you're not this purist around how people should operate, or what you want the world to be like. You're like, "Yeah. I want the world to be better and use the best software possible."
Edith: That was a brutal lesson for me. I had been an engineer, I had been a product manager and I'd managed sales people. But to go from managing salespeople to being the salesperson, I had to get much better at rejection. A 50% win rate is high.
Grant: That's really high.
Edith: Versus if you're an engineer, if 50% of your code doesn't ship you're like, "Something has gone dramatically wrong."
Grant: I think about it as, "I've actually never-- We've never lost a deal, they're just delayed."
Edith: That's why I'm like, "Nothing is ever a closed loss. It's just closed now".
Grant: It's just not now. "We'll get them some day."
Edith: I feel the same way, and it's true. We do feature flag management, sometimes people are like "Not right now," and then we'll come back a year later like, "Now?"
Grant: Then they want it even more, because they realize they tried to build their own version, and they've got all this stuff. It's a big pain, and no one knows what to do with it anymore. It's like, "Just give us the thing."
Edith: Something we did along those lines is, a lot of what we're doing is trying to convince people that feature flagging is a better way to build software. So we explicitly published how to build your own system.
Edith: With the theory that if people try to go build their own system, they would get sick of it and eventually come to us.
Edith: Because it's like, "You're engineers. We know that you could build this if you wanted to.
Grant: It's so true. I refer to this YouTube video that I watched, I was going to build a huge wood fence outside of the Replicated office, like literally like a 40 foot fence. It's like six feet high, and I was like, "I'll just watch a YouTube video and figure out how to do it." I watched a 20 minute YouTube video of this guy using this huge thing to dig up the fence post holes, and all this concrete, and how he's measuring everything. At the end it was like, "LA fence building," and I just called that guy and I was like, "I need you to build me that fence I saw on YouTube."
Grant: It's perfect. "Turns out I can't do any of these things that you just did. Thank you for showing me how the sausage is made because it was really complex."
Edith: Our thing is always-- Again, we sell to developers. We sell to engineers. It's like, "I know you can build this. Here's how to build this. Do you want to build this?"
Grant: Right. "Is this what you wanna spend your time doing?"
Edith: Yeah. It's like, "There's better ways to spend your time." And that gets back to
If you try to tell engineers they can't do something or that they're not smart enough to do it, they get pissed off and they'll go try to do it just to prove them wrong.
Edith: If you're like, "This is too complicated. You'll never figure it out." They're like, "Screw you, vendor. I'm gonna go build this."
Grant: It's funny, I always say "There's no--" especially our customers, who are these great amazing developer tools and DevOps companies. Every one of them could build the things that we've built. When I talk to VCs, like "Could somebody else build this?" I'm like "Somebody else could build every piece of software. If you've invested in a company because you thought that's the only team that could build the software, you're an idiot."
Edith: It's like the famous quote about Dropbox. When Dropbox launched on Hacker News, somebody was like, "This is just a file share."
Grant: Yeah, "You could do this with these seven steps in Vim," or something. But that's not really the point. Especially today, because so much software is based on open source and you're really pulling off of all these other components in the world, rarely is there a truly net new crazy thing. Even all the AI stuff is like--
Edith: It's just algorithms.
Grant: Tensor flow underneath the hood, right? It's hard, in my opinion, to really have something that's so differentiated.
Edith: What we sell at LaunchDarkly is reliability.
Edith: I go talk to people out of their homegrown systems, like I heard this horror story from somebody in New York where it's like, "We have this homegrown system to do our feature flagging, but the caches don't clear correctly." So he still gets calls at 3AM because somebody turned a feature flag off and the cache didn't clear, it's all messed up.
Edith: We sell you a system that you know will work.
Grant: Turns out, cache expiration is one of the hardest problems in software.
Edith: There's only three hard problems.
Grant: Right. That's one of them.
Edith: There's two. Caching, naming, and off by one.
Edith: Grant, it was really nice to talk with you.
Grant: Edith, this was so great. I will have you on again sometime. We're gonna go even deeper into some enterprise software stuff. This has been amazing.
Edith: I've been in software my entire life. My joke is that I made a company to fix all the problems that I had encountered in the 20 years that I have been building software.
Grant: That's the best. You get to build the things that have always annoyed you a little bit, and make other people's--
Edith: No. More than a little bit. The things that made me so frustrated when I was an engineer or a product manager.
Grant: That's what makes really great product people, is that there's just a lot of-- You notice all the little things that could be better, and you're like, "Why don't we just make it better?"
Edith: That's the best part. I said earlier, NPS is really high. I read the NPS surveys and people say "This changed my life for the better." That feels really good.