Lessons Learned Building a $2 Billion Company from Scratch with Neo4j CEO & Co-Founder Emil Eifrem

So my name is Emil Ephraim and as Kaitlyn mentioned, I'm the CEO of Neo4j, which as you'll soon learn, is an infrastructure company. And in this presentation I'm going to discuss a few things that we've learned in building this company from scratch to where it is today. So before we get into the meat of the conversation, just a few words about Neo4j so you understand where I'm coming from for the rest of the presentation. So Neo4J is a graph database company. We in fact were the leader in the category of graph databases. Millions of downloads and hundreds of Global 2000 customers. And there's only just a handful of new database companies right now, what used to be called NoSQL for those of you who are around back when that term was coined, who have reached escape velocity, and Neo4j is one of them. And in fact we just raised a $325 million Series F, which actually is the largest investment in database history. And folks like Gartner are saying things like graph techniques will form the foundation of modern data and analytics. So we are truly going for it. We're headquartered in Silicon Valley with close to 600 people worldwide, including in Europe, in Latin America and in Asia. So just to set the stage here, I want to talk a little bit about kind of the backdrop of what's going on in our industry for the past decade. About 10 years ago, if we go back to the way things looked back then, you had this insane chaos happening in infrastructure, not just in databases. I of course have lived in a database world and maybe it was most obvious in databases. This is five years before Cloudera went public and the world was figuring out whether Larry Ellison would again just turn some of these companies into features or whether they could actually sustain independently. And most offerings they were deployed on prem for a combination of reasons. Security concerns probably buyer immaturity and lack of enabling technologies like Kubernetes. And so as a result you had this massive market fragmentation. Tons of distributions of the same open source code. Venture capital funding is was just pouring into this new wave of startups. And then traditional massive companies all the way from the IBM's to the Accentures of the world, trying to figure out their play, so to speak. And the market adoption and the value was heavily enterprise biased. Mom and pop shops wouldn't put in Hadoop, they couldn't. Right. So the people who could adopt, either they had the organizational sophistication to deal with them themselves internally, or they had the budgets to pay a vendor to do it for them. Or possibly both. Fast forward to today, a decade later and there's still plenty of innovation happening, but it's far less fragmented. The market appears to have coalesced around just a handful, like actually less than that. Three, four database models in the database world that more likely have a significant long term role to play. And there's new entrants that pop up fairly regularly. But there's no longer a whole market churning through this tidal wave of partially baked technologies that are looking for some semblance of competitive advantage. So new projects are shifting to the cloud more and more every passing day. Most workloads are born and raised in the cloud. There's even workloads that are shifting from the on prem world into the cloud. And thanks to this, you know the cloud, but also things like consumption based or pay as you go, pricing and just general technology evolution. Right now we're at this very, very interesting time where any company can get access to nearly any technology through the cloud. So this is hopefully well trod and known. You know that that's not news to anyone listening in to this presentation, but it's backdrop to explain where we are. So what we're seeing today then is this generation of infrastructure companies that are thriving with their customers and with their communities and for that matter with their investors. As a result of all of that, they started as open source infrastructure companies, but later on they got funding and became basically viable commercial open source startups. And then over the last couple of years, maybe two to three years, they become these public companies, these public companies. And if you add them all up together, the market value is just massive. It's between 50 to $100 billion just out of these generation of infrastructure startups. And so we're living our own journey like this here at Neo4j, informed by, by their success, of course. And there are some patterns emerging and I'd love to share them. I'm going to punctuate them as it were, with some specific examples both from them and from our own journey here at Neo4J which hopefully will make it, make it stick. So what we've seen emerging here is these four stages of a virtuous cycle. And throughout that evolution that I talked before, from open source project all the way through to a public independent company, you see this virtuous cycle of growth that just grows and changes as marketing companies emerge. It starts with awareness so that people can get excited about you and find you activation. Then is that moment when they decide to do something with your product, hopefully useful or interesting. And then you learn from your users and from the usage and you evolve your product and those services to fit their needs. And their success then drives awareness directly and indirectly and draws in the next wave of users. This is a generic cycle, but let's spend some time talking about what this looked like for us in the early day. We had a lot of boots on the ground. We met developers where they lived on third party forums. We didn't have enough gravity to draw them into where we lived on neo4j.com or neo4j.org but if you start getting enough traction, then journalists will start to pay attention and usually through a developer's lens. And then community programs, they end up being how you keep your community continuously engaged. So for example, we started a program that we call Grasp for Good, which was to encourage really world impacting but non commercial use that didn't make us any money directly but generated a lot of awareness. These are things like the Panama Papers which some of you might remember it's five years ago, but it was the biggest news story of all of 2016, where a lot of politicians like the Prime Minister of Iceland and the Prime Minister of Pakistan were found out to have bank accounts in offshore tax havens and several of them ended up losing their jobs because of that. That was all backed by our technology. We didn't make a single dime. But it generated a lot of awareness and buzz and goodwill. And this Grasp for Good program also span other things that weren't directly revenue generating but generated buzz. Like the mission to Mars where NASA is on the record saying that thanks to Neo4j humanity will get to Mars two years earlier. So those are the kind of things that were very successful for us in the early days. And eventually if you're really super successful, the major industry analysts, the gardeners and the foresters of the world, they will eventually take up what developers have known for years and those going to start talking about it. And obviously the business press will cover you if you have interesting enough investors. And then activation that's trying to make it as frictionless as possible for users to do something useful with your product. As we alluded to before. For us the threshold or the bar to success in the early days was really really high. And that's true for most community edition open source infrastructure products. The tooling is really immature, the documentation is light at best, but if there's enough there, then those users will activate and use especially if they have a deep enough pain and antibiotics level pain. At Neo4J we collaborated with Other projects to get embedded in things like Linux distros so that we could find developers even before they ended up finding us. And as our tooling then matured, we saw a chance to improve that activation experience and make a step function improvement to adoptability by doing things like building a desktop edition. And over time, of course not shockingly to a conference audience like at Saster, our community started shifting to the cloud. So then it obviously made sense to give them cloud ready images. And earlier this year we launched another step function improvement with a perpetual cloud repair, which we're going to talk more about a little bit later. Then on some level, being born as an open source project is I think the first bold act of user centricity. From there we wanted to make it a lot easier for developers to exploit our engine. We ended up building out this, this query language called Cypher, which on some level is the equivalent of SQL for graph databases. It's a declarative query language for graph databases. We after a while realized there's an opportunity to expand the entire space, not just our product, by making it easier not just for Neo4J developers, but for all graph developers to use graph databases. We actually took that language when we open sourced it under the open cipher umbrella. And this is a little bit of a counterintuitive move because it's taking one of the crown jewels of your product and giving it away to your competition, including the cloud platforms. And we'll talk a lot more about the cloud platform firms later. But over time, that openness, that more inclusive strategy ends up being the rising tide that lifts all boats. And it ended up creating this critical mass of both user and vendor adoption, which is important in category creation to not just have user adoption, but vendor adoption. And that means your competitors using things like your query language that, that is what ultimately broke the stranglehold that SQL has had on database query languages for three, four decades. So in databases there really only been one officially sanctioned blessed query language called SQL until a couple of years ago when the ISO ended up blessing GQL as the formal query language for graph databases. And those thanks to a lot of work on that openness strategy from folks like Neo4j, but all of our competitors as well, or at least many of them. In some ways I'm thinking of Open Source as the original product led growth, right? Or kind of the, the OG PLG as it. As it were. Apologies for the, for the dad joke, but that's kind of how I think about it. And we've always had this motion that makes it easy for enthusiastic users to share with others and create growth groups. And that started in the form of an open source distribution which ended up growing the developer community as people shared their success and talked about how and when and why they were successful with our product. But we also attracted entirely new user communities using open source. So for example, releasing a graph algo package that attracted data scientists. And then over time we've invested in the ecosystem to create more on ramps onto Neo4J from other platforms like Spark or Confluent, Patrick Kafka. And now most recently a cloud service which I mentioned before, with a perpetual free peer which offers the most frictionless way for our users to share their success with their peers. So those four pillars is the broader framework through which I look at these modern infrastructure companies and I want to move on from that to a couple of more specific points. So the first one is absolutely central. Like distribution is just so, so, so important. And you want to make sure that you exist in all form factors that are relevant to your users. You want to make sure that you are everywhere. And that's on prem, you know, of course, right, that you know, we all at Saster believe that the, we live in this cloudy world and of course that's, that's both the present and the future. But there's still a huge amount of ADOP in on prem data centers, right? And it's embedded and distributed. You know, the cloud of course is all the rage. But you know, on some level I feel like enterprises run on multi cloud data centers, but developers still run on laptops. So the fact that as an infrastructure company you have this on prem offering is actually a strength, right? Not just for production deployments where you can offer hybrid form factors, but also for developers running on their local laptops. And then we'll spend a little bit more time on this later. The cloud marketplaces of course, always deserve a special call out in terms of reducing friction to accessing your product. So that's the first one. And then second of all, you know, a free tier is a lot, lot more valuable than a free trial, right? And there's a lot of debate about this and every business is special, but for me personally, I come out very much in favor of a free tier. And if you look at some of the successful, some of the patterns from successful modern infrastructure companies then end up gravitating much more towards free tier than a free trial. A free trial is valuable. It's going to give you leads, right? Which I mean, we all know how important that is, but a free tier will generate users. And if that free tier is good, not too underpowered, right. It's going to generate not just users, but advocates, people who will promote your business, who will evangelize your product. And most free trials don't deliver a ton of user value. And I mean built into the entire notion is that that value is transient by definition. That's why it's a free trial. It's time bound. But you can learn things about your product from that free trial, but really only about the trial experience, right? And that early usage and that's valuable, you can optimize your trial. So that's important. But it's not going to do much to make the product that your customers use every year, every day, every year inside out, right? It's not going to make that one materially better. And with a free tier, if you instrument it well, then you can learn a lot more about the user journey and the usage life cycle through expansion and even through deprecation, retirement or basically churn. And those product learnings are going to circle back and cycle back into your product to bring value to your entire commercial customer base, even though those are perpetual, free forever users. So I come out very much in favor of having a free tier. And then the third tip, this one is probably more broadly applicable but most stark in open source. Now open source is a very common pattern in modern infrastructure companies. I don't think applies to everyone, but I think for sure most successful modern infrastructure companies are open source by birth. And what's really interesting around cloud services and the platform shift to the cloud is that it provides just a fresh opportunity to re segment the free versus paid and redraw that monetization line. And that's a really consequential early decision. It's a capital B, capital D, like big decision in open source infrastructure. What features are you going to float into the free, let's call it the community edition, just for ease of wording, versus in the, let's call it the enterprise or the commercial edition. That's typically a decision that you make early on in the first few years, maybe somewhere between 0 to 10 million of ARR the first few years of your company. And sometimes you get that decision right and sometimes you get it wrong. And even if it was right for you at let's call it 0 to 10 million, it may not be right for you at 100 million. And what's really interesting, as you shift to the cloud, then all of a sudden you have this opportunity to redraw that monetization line and that's I think a very interesting opportunity. And then in addition to that, the cloud form factor also have a lot more knobs and levers available right in on prem open source it's basically just features and possibly choice of open source license. Those are really the two dimensions that you have to play with on the cloud. All of a sudden you have memory consumption, you have memory footprint on disk CPU consumption, you have higher level metrics that relate to the abstraction in your domain. So like Kafka streams or Elasticsearch documents, or in the Neo4Jsense like nodes and relationships, you can price and meter based on that. And so not only does it give you a chance to redraw, you also have more knobs and levers available. And if you match that up with the first two tips that be everywhere in the free tier, all of a sudden it's much more safe for you to drive users onto that free tier because there's multiple fences that can gate whether the users should flip over into a commercial edition. And you can start monetize the entire spectrum of your user base rather than just open source, they pay zero. And then big enterprises on your commercial edition, they pay millions of dollars. So that's the third one. And then the fourth one is around the cloud platform. You know, keep your friends close in the cloud platforms closer. And what's very, very stark as you study these escape velocity infrastructure companies in the cloud, is that the cloud partnerships can be this exceptionally successful driver of business for these companies. And you know, I'm not going to go on the record as guessing what incremental growth it's giving them, but it's a substantial part of their business. So if there' any chance that you can get like as deep of a marketplace integration as possible, just, just go ahead and do it right away. But not only that, you know, obviously if you can get the cloud platform field engaged, and there's been a few of these cloud platform partnerships that have offered really, really beneficial terms to both the field for the cloud platforms as well as well as for the for the partner, that can really end up driving a lot of success for you in the cloud. So that's kind of the one hand of the equation. The other hand of the equation, of course, is that they can also be your fiercest competitors and especially if you're open source. And there's a big fork in the road for open source infrastructure companies, which I think is dissimilar from for example application vendors, which is when one of the cloud platforms enter into your category, can they do that with your own product? And that's that big fork in the road, especially if you're doing category creation. If a cloud platform enters into your space, that's good news, right? When you do category creation, your high order bit is getting the word out about the entire category, right? And so you know how all startup CEOs end up saying whenever a big company enters your space, oh, this is great validation. Well, in category creation that's very much true, right, because the high order bit again is getting awareness about the concept or the category out there. But if they enter into your space with your own product, you are living in a very painful world because they have the massive benefits of of scale. And then if they compete against you with your own product, that's very, very challenging. And there are really only two ways that you can protect against that. And they're not mutually exclusive. But unfortunately some of them end up being hard to retrofit onto your business kind of in after you've started the company. But the first one is of course licensing, right? There are certain open source licensing licenses where the cloud platforms can just adopt you wholesale and they don't have to strike a deal with you or anything like that. More permissive licenses and there are certain that protect you against that. So that's the first one. And then the second one is differentiation between your feature differentiation I should say between your free and open edition and your paid edition. Those things combined, if you play it right, can provide a good moat to against the cloud platforms and force them to enter into your category with an either in house product that they built themselves or another one of your competitors products. And that's a completely different universe to live in compared to if they compete against you with your own product. So that's kind of the fourth one. We're coming up to the full bottom of the the hour and I wanted to provide just not too much content since it's early on in the, in the morning for at least for those of you in California. But a couple of key takeaways from this. There's never been a better time to build infrastructure companies than now. There are certain product categories that are very, very sticky where the half life is very long, when there's really only one chance for you to build a company in that space and that's when you have a platform shift. And obviously we live in this really magical world where there's two concurrent platform shifts going on. One is what this conference is dedicated to, which is the shift to the cloud and then the second one of course is the shift of mobile. And both of those platform shifts leave white space in the market for infrastructure, which sometimes is harder to get to market than more vertical application centric companies. And then there's a, there's a clear blueprint that has emerged. If you look at successful infrastructure companies today, you should never just blindly follow it or imitate it, of course, but it's important to understand it and be inspired by it. And there's a couple of key ingredients in that. Open source is a big part of it initially monetizing the enterprise exclusively, both with an on prem form factor, but of course increasingly with a cloud cloud product that monetize the Global 2000 in a, in a very enterprise sales centric way. And then you can launch this self serve first or PLG cloud offering through which you can monetize the entire spectrum. When you've created awareness around your, your infrastructure product, that awareness is typically not limited to a specific size of company, but it's limited to whenever you have data problems wherever you are, developers or something generic as that. Right. And you will never ever in the early days be able to hire salespeople to target any vertical, any size of company, any geo. But with the self serve cloud product you can address that and so you add all of that together and it's a magical time to be building infrastructure companies. So that's all I had as a queue up. Would love to hear any type of questions from the audience.

Yeah, thank you so much for kicking off Saster Annual so well, I do see a few questions coming in and feel free my audience to keep dropping them in as well as we go. But we will start our first question. You've been laying the groundwork for the mainstream of graphs for some time. What were the early signs that the nature of infrastructure companies were changing? Where did your foresight come from?

Yeah, it's interesting, right? Probably it's one of those that were typed up just before the last slide that I mentioned. I think really the early sign was the shift to the cloud. Right. It's back to the notion that infrastructure companies. Right. It's sometimes very hard to build when you don't have changes in adoption patterns amongst in particular big enterprises, but really probably developers everywhere. Right. And that happens when you have platform shifts. Right. When you have shifts from, you know, let's take no databases best, let's take the database space. Right. So Oracle was built on the shift from mainframe to client server. Right. That was a big platform Shift. And now we have this other big platform shift from client server, maybe to web, which is where MySQL was built, but now from that to the cloud and that opens up this white space where you can build infrastructure contents.

Yeah, thank you so much for that answer. Again, remember to add your questions to the chat. But moving on to our second question. What advice would you have for a founder looking to disrupt the market today and build a category?

Man, I guess I'm on the record of saying that ignore all advice. So I'm not sure if that should be my meta advice. The one thing though that whenever I talk to earlier entrepreneurs and founders is that you end up getting so much advice around where you should spend your time. You should do X, you should do Y, you should do Z. The only thing that I think that you can't do too much of, in other words, do as much as you possibly can, is spending time with customers. And when I say customers, I don't necessarily mean paying like the people, the economic buyers, the person writing the check. You should spend time with them to make sure you get that check, collect that check, but really with the users of your product. And you should know absolutely everything for your Persona, for your user, the product. Like when do they wake up in the morning? Would they attend a 7am you know, Saster session in their, in their time zone? Where do they go for information? What magazines, website, blogs, substack, what do they read for their information? Who do they hang out with? Like, you should just know everything about that, about that Persona. And so that's, I think, the only kind of advice that transcends all the tactical kind of day to day and the huge, the contextual pieces it is spend as much time with customers as you possibly can.

I love that customer obsession. You can't go wrong with that. Our next question, what should ISVs be thinking about now if they want to compete?

Oh, ISVs. Yes, that's a, that's a blast from the past. Independent software vendors, by the way, is what it is. Yeah, I think there's, there's a couple of things, most of which I touched on, you know, in, in my presentation. And since this is infrastructure frame, you know, a lot of it ends up coming out of, I think, open source. Right. Because if you're not open source today as an infrastructure company, it's very hard to be choosable. Right. And so for, and so that's great for users, it's typically great for the, for the vendor, for the startup as well, or for the isv. I guess is how the question was framed. But of course it opens up that flank against the cloud platforms where all of a sudden they could take your product and compete with yourself. And so that would be. I think the probably most important part is choosing a license and a differentiation between your product editions. And when you do that, be mindful not just of the users and how you go to market and how you monetize, but also the compete, you know, the competitive vector from the cloud platforms. And I'm of the, in the kind of, let's call it the Larry Ellison kind of school of thinking versus the Jeff Bezos school of thinking. Right. Jeff Bezos is famously customer centric and customer. You mentioned the Caitlyn customer obsessed. Whereas Larry Ellison wakes up every day wanting to kill the competition. Right. And I'm very much of the Jeff Bezos kind of school there. So I generally think you should ignore your competition and you should just focus on users and build amazing products. The one exception for infrastructure companies would be if you're open source, make sure you're mindful of the cloud platform so they can compete against you with your own product.

Yeah, Jeff Bezos, I love his quote. It's always day one.

Exactly.

Yeah. One of my favorites. So our next question is somebody saying that they loved your comment about not taking advice. Were you ever given any bad advice? Or it's, were you ever given any bad advice? Or how do you get the confidence to not take others advice?

Constantly given bad advice, honestly. And you know, it's, it's fascinating and on some level, like, how do you get the confidence? Well, I think some of this, my entire presentation was trying to draw out patterns of success from people who have gone boldly gone before me. Right. And I'll, I'll give you an example of bad advice. I think that was the question, right?

Yes.

Example of bad advice. Right. Yeah. And so, so bad advice for us as an infrastructure company was you have to sell into the line of business. Right. If you're not solving a business problem, then it's not worth solving. And if as an infrastructure company, if you just solve problems for developers or data scientists, that's somehow not as good as solving problems that are experienced by the line of business. In other words, people who own a P and L. And that is clearly very bad advice. When you're an infrastructure company, you always exist on the, the build path, right. In the enterprise IT buy versus build. If they don't go down the build path and want to build themselves with their own project, you don't have a play if it's on the by path. They want to buy an off the shelf solution, a turnkey solution and your only play There is through OEM with another SaaS vendor or another company become embedded in someone who solves that business problem. Right. But as an infrastructure your direct play is always on that build path. Right. And so trying to market to the line of business, probably 90% of the advice that I got was there and 100% from the people who have an MBA. Like no offense to the people on the call who have an mba, right. And so that was, that was end up just being very bad advice. And then I guess the second part, if I remember the, the confidence correctly was like the confidence, right. And I think some of that comes back to, I would say two things, right. Deeply understanding the user. It's coming back to that like understanding the customer, understanding the user. Right. And seeing what they actually need because that, that ends up being the source of truth. Right. But then also the patterns of success, right. You look at what a. I mean my three examples could have been other companies, but I happen to choose Confluent, Mongo Elastic. Right. They all look at much greater scale than you are as an early stage 0 to 10 million ARR startup, which is typically when you grapple with these things and you see that that's how they go to market, then that ends up being of course confidence inspiring as well.

Yeah, no, that was a great question. We have another. What's the trade off between competition and cooperating with public cloud providers? How do manage this?

Oh wow, that's such a good question. This is clearly from a practitioner, someone who's, who's living in that because it's. And unfortunately I don't think there's a. I, I've certainly not seen a, like a silver bullet here, like a magical way to do it. It is part of why running a company there's an art and that there's a science. Right. The science is there's going to be a lot of talks at SASTER about LTV to CAC and all those metrics and there's been a ton of that. This question really is in the arc side of things. How do you balance that? And fundamentally as with so many of these things, it comes down to people. You need to make sure that you are as competitive as is needed to win in the deals when you come up against the cloud platforms, but not so much take such an aggressive stance that you can't at the same time partner with them. Now the really good news is, again, coming back to people is that typically this ends up being multiple groups inside of the cloud platforms. Right. You might have a great relationship with a partnering team or with the marketing team, but then a vicious competitive relationship with some of the product teams. And that's fine because these are massive, massive organizations. But unfortunately, I can't provide any more specific direction than that. But that, that's at least some, some color on it.

No, that was a great answer. I. Oh, here another question. Has the pandemic hurt or helped you when it comes?

Yeah, that's a good question. I want to say this carefully because of course, the pandemic has been a tragedy on so many levels, right. From a, from a human perspective, also for us internally in terms of our ability to meet. I have a blurry background here because I'm actually at a hotel at an off site. One of the first. I think it's actually the first off site we've had since before the pandemic. Right. With a small team here, which is just fantastic. And not doing that for 18 months has, of course hurt us as much as it is hurt, has hurt everyone else. Right. With that as caveat, the pandemic has been amazing for our business. It's really accelerated it in so many ways. So our. I mean, I talked a whole lot about kind of the core of what, what we do, but as a graph database, our job is to figure out the cost and effect and trace how information flows from one point to the other. And so you start thinking through, like, what Covid tracing is, you know, that's all about. That's a graph, that's a network graph problem. Right. Like figuring that out. So a ton of people have, ton of countries, as well as companies, as well as states in the US and things like that have used Neo4J to do contact tracing of COVID Covid also ushered in an entire new focus on supply chain. And supply chain, of course, is a massive graph problem, because that's all about this. One vendor produces something that is shipped to someone else, that is shipped to someone else, and all of a sudden it has to go through the Suez Canal, but the Suez Canal is blocked for a week. If you recall, this is, I guess, six months ago or whenever that was. Right. A while ago. And how does that cascade across your business? Well, in order to figure that out, you have to graph out your entire supply chain, which is, which is this deeply connected data structure that drove a huge amount of usage of Graph databases and Neo4J. So we're surfing on this broadly secular shift to the cloud, which was accelerated by the pandemic, but also specifically graph technology was accelerated by it. It's been a very positive thing for us outside of the human tragedy, of course.

Yes, of course. But great point. There is always some positive that comes from everything. Our next question, has it been hard to hire cloud developers?

Oh, man, yeah. This one is just super tough. Like for those I think we have mostly, I don't know, like Saster went back when I went face to face disaster, it was mostly kind of founders and go to market professionals. But maybe today there's also a lot of engineers. If so know your value because there's, there's way fewer of you than, than as vendors as we, as we want there to be because it's a very unique skill set and maybe even more so the people who've successfully done a hybrid play. So going from on premises to the cloud while having the on premise and existing important product surface, that's an even rarer breed. So if anyone on the call has that skill set, please drop me a line after the presentation because I'd love to talk.

I love that connecting people in all different types of ways today. All right, well, we are just about at time, so thank you so much, Emil. I really appreciate you coming on and for the great session.