Datadog, Inc. (DDOG) Earnings Call Transcript & Summary

Okay. Good afternoon, everyone. Welcome to the conference. I am Mark Murphy, Software Analyst with JPMorgan. It is a great pleasure to be here with Olivier Pomel, CEO and Co-Founder of Datadog. First of all, Olivier, thank you so much for joining us today.

Thank you for having me.

So I'm sure that many investors in this audience are completely familiar with the Datadog story. There are going to be some who are not. Wondering if you can help them by describing Datadog? Describe some of the problems that you solve, how you've developed this platform? And why was it that you and your co-founder, Alexis Le-Quoc, decided to start a company around observability at this time when there were already plenty of competitors?

Sure thing, yes. So what we do is, Datadog is a platform, unified platform, for observability, monitoring and now also, security. And we built it specifically for cloud environments, be they public clouds or private clouds. And obviously, the good news here is that just about everybody is busy moving from legacy IT to the cloud, and so we do serve quite a wide customer base on that. The reason we started the company is that -- so my co-founder and I actually didn't want to start a monitoring company. We are not here to build a better mousetrap. We didn't start by looking at the Gartner Quadrant, [ Masina ] sales, it was part of the subcategory that we'd like to build. We started Datadog because we used to work in development and operations. In our previous company, I used to run the development team, Alexis, my co-founder, used to run the operations team. And even though we were in this situation where we were good friends already, we work well together, we hired everyone on our teams, we had a no-jerk policy for hiring as well. We still ended up with development and operations that hated each other and that were fingerpointing all the time. And so the starting point was really, how do we get people from these different backgrounds with these different job descriptions to actually work together, share one, common tooth, one common view of the world? And that was the birth of Datadog. So we started with infrastructure monitoring, then we expanded to what is now known as the all-observability space with application performance monitoring, with log management, a few other things around user experience monitoring. And today, we're also expanding beyond that towards security use cases and a few other use cases that always revolve around getting more people with different job descriptions into the same platform to actually work together and share the same reality.

So you mentioned one common truth. You've had one common platform as well. Could you speak to the advantages and disadvantages of this? Because there are plenty of other software companies out there. What they will do is they'll verticalize the solution, they'll diversify it out in other different ways. Would you ever consider doing that? Maybe why or why not?

So for us, it's actually very important to have one common platform. And it's one common platform not only between different job functions, so development, operations, product management, business analysts, security engineers, like a lot of different roles that basically get around our platform. But it's also the same platform that is being used by a very wide gamut of customers that range from individuals, students, tiny companies all the way up to the largest companies in the world, with everything in between. And the reason for that is really that it's -- today, when you're a developer, your experience is not very different whether you work for a Fortune 500 or whether you work on your own. You use the same tool set, you use the same building blocks from Amazon. It's not like it was 10, 15 years ago when if you were in a big company, you'd buy the Oracle tool suite. And then you build everything on Oracle. If you were in a small company, you do everything open source. It's not like that anymore. Everybody has the same building blocks. So we built it that way to sell a very broad customer base. And what this gives us is a product that has to still be usable by mere mortals that don't have very large organizations to support them, which keeps the product simple. If you direct your product solely at very large enterprise users, what happens is that even if you start with a very easy to add product, you're being pulled constantly towards more complexity, more functionality, things that make it harder to stand the product on its own. And you become what I like to call an enterprise abomination, and pretty much every single enterprise software company follows that path. And all way to structurally prevent ourselves on becoming that and to have a product that can stand on its own and be frictionlessly adopted in a bottom-out fashion is to serve a very wide customer base.

So the aperture is wide on the customers that you're serving, but the data scale of this problem is astonishing. And I heard you mention this a couple of years ago, Datadog is handling 10 trillion records per day. You've talked about how everything needs to be persisted. You have to keep all this data. You are monitoring millions of hosts and containers. So what was the engineering project like, first off, to tackle such a large project?

Yes. Yes. So that's, I mean, pretty much at the core of what we do. And by the way, those numbers are much higher today than they were when we communicated them at the time. The one way we do that is that we don't actually rely on one single data store for that. We gather many different types of data. We gather high volume metric data. We gather event data, log data. We gather data -- a lot of metadata about infrastructure and how it ties together, metadata about end users, what they're doing with the platform and things like that. And what we do is we constantly rewrite all of the various data stores we have to adapt to the volumes and the evolutions of these different data sets. And we tie all of this into one unified platform that has a data role that actually is built for integrating data types that might have different velocities, different chips, different characteristics. So that's what we've built. Obviously, we think we -- for some of our data stores that the sixth or seventh generation already of those systems. Some others that correspond to more recent products and more on data types, I think, in second or third iteration. By the way, we just posted an engineering post on the architecture of one of our new data stores that we use for log data and a few other products that rely on that as well. So that's a constant concern for us.

How do you prevent bottlenecks, right? When you're dealing with that much data, how do you make sure it scales? How do you make sure a transaction isn't delayed? How do you keep that type of surface area secure?

So on the reliability side, I mean, obviously, we dog food quite a bit. We use our product. So we had to become experts at building and running very large distributed systems to process all that. Because to your point, we get all this data in real time. A lot of the data we get has to be acted upon in a matter of seconds. Otherwise, they have no use to our customers. So any little, miniature gap we might have in that or delay is actually going to close customer-facing problems, and that's something that we build the business around. In terms of securing it, same thing, we started building a lot of our tooling for that. And more recently, we've entered the security space with a product -- a cloud security monitoring suite that we -- I think is the one right now. We finally completed all the various components of it. And that was born into some of our internal tooling, and that was also developed by being heavily dog fooded internally by our teams who use that product to build. By the way, we believe that to fully solve the problems of securing large cloud environments and applications that are being homegrown builds by us and by our customers, you can't just rely on the security team to do that. You need to bring to bear the full operations team, need to bear the development team. The operations team is typically 10x as big as the security team. The development team is typically 10 to 20 times as big as the operations team. So really, you need to mobilize that much larger set of engineers to solve that problem. And that's what we're doing with our product, that's what we're doing internally. To be honest, when we started solving those problems, when we started with Datadog, we were facing the same issues that some of our customers face today, which is that we had issues getting security engineers aligned with developers and with operations folks. They were also fighting a little bit. It was very reminiscent to what we had seen before starting Datadog and we decided to start Datadog. That's also one of the reasons we thought the approach we had for observability would actually lend itself quite well to security as well.

So you've had this incredible vision. You've been executing on it very well, to say the least. There is some competition out there. I think usually, you will refer to this as a build versus buy type of a decision. There's going to be a part of that decision tree that would involve the hyperscalers products, open source products. Can you speak to us what is happening competitively? Are the hyperscaler products improving? And how do you think about out-innovating and staying ahead?

Yes. So the -- so first of all, on the hyperscaler side, like, those products have always been there. Ever since the service provider started providing services, there's been a built-in management console with it to help you understand what's going on there and get some -- keep tabs on the service. Those products, though, structurally, are very myopic to the cloud provider itself and to the software that is built in-house by the cloud provider. So it typically doesn't work when you want to go across provider, which, by the way, most of our customers do today. And also, it doesn't work very well with technology that is not homegrown by the cloud providers. Say for example, if you run a third-party, open-source database on your cloud provider, that's not something that the providers are going to be -- to understand very well. So structurally, we see these limitations there. Now in terms of the broader build versus buy, I think the problem is not so much that we have to out-innovate the competition or what everyone else is building. The problem is more of innovating to keep up with the growth of the problem. And the systems, software, infrastructure, applications, all of that is becoming more and more complex. It's becoming more complex very quickly. It's not stopping. And it's only going to get worse as more and more is moving into software, algorithm, et cetera, et cetera. And -- but there's not more and more engineers to actually power all that. So we're going to have more and more of these things resting on less engineers, less people to build it. Which means there's going to be [ terminally ] more complexity per person and more problems for us to solve there. And we see that, by the way, in the short term, there might be some macro headwinds that make companies reduce their workforce, which means they want to do more with less, so that probably has become even bigger. In the long term, it's going to happen anyway because there's not enough engineers that are being produced or taught or trained today to fully service all the needs that are occurring with digital a transformation and the way enterprises in general need to transform and do more in software. So we think that this is going to be permeated too and -- short term, mid-term or long term.

So Olivier, you used this term -- you're basically trying to keep up with the pace of innovation, right, and the pace of change in the industry. But when we look at this, the growth has been astonishing, right? You've surpassed $1 billion in revenue. You talk about how it is still early days for you as a company, and you just mentioned these medium-term and longer-term opportunities. Can you expand on that? How much of the industry is -- today is transitioning toward cloud? And also, if you could touch on some of the next-gen technologies like containers?

Yes. So we are still early, both technologically in terms of the vast technological migration that is taking place right now. But also in terms of our market and our penetration of that market. So if you think of where we are today, we're still not in every single -- we're not at full scale in every single cloud provider, in every single geography for every single customer segment. So for us, I would say there's maybe an order of magnitude, more scale we could get there just by being fully penetrated in the workloads that are in the cloud today. And in addition to that, there's an order of magnitude, more workloads that are not in the cloud today that will be as this migration continues to take place. In addition to that, we have to factor in the fact that we're still at the beginning of digital transformation, and that we're not replacing workloads 1 to 1 between the legacy world and the -- where the world is heading like 2, 5, 10 years from now, so there's going to be a bigger footprint there. And then in addition to that, we are also covering more and more of the problem space that our customers are facing. That's one of the key characteristics of our platform and our business, is the ability to layer on top of the surface of contact we have with our customers to serve a bigger and bigger problem for them and at new categories. So that also opens us, I would say, an opportunity that is maybe a few times the size of what we have today. So when you layer all of those, the penetration and being at full scale everywhere for the existing workloads, the workloads that are still moving to the cloud, the expansion of the domain of what's covered by application and software in a digital transformation world and the fact that we're entering new categories, I think we're still barely scratching the surface what we can do basically.

Okay, barely scratching the surface. So again, it's over $1 billion in revenue in the last 12 months. The growth rate has been in the high 70s. You've had a free cash flow margin close to 30%, right, so over rule of 100. We don't see much of that. We have companies here that are growing half that rate, and they're burning hundreds of millions of dollars annually, right? So why this difference? To what do you attribute this level of efficiency you've been able to maintain?

A few things. First one is, we're selling into a transition. We're not trying to rip and replace what is already out there. We're selling into this transition to the cloud and to digital transformation. We are not the ones triggering this transition, so we don't have to do years and years of hard selling to convince customers to move to the cloud. That's not us doing that. It's happening anyway. It's happening with or without us, and we have to be there at the right time to accommodate these customers when they go through that. So that's number one, setting into this transition. It allows us to start very small with customers as they start running in cloud environment and then to grow with them as they keep migrating. The second part of it is that we have a platform that allows customers to adopt our products in a frictionless way. So they don't have to call us, they don't have to place an additional order. They can just use the building blocks and consume more of our products that way and grow with us. It means any product of ours they start using, they can start at a very, very, very small level. And then as they find more value, they can use more and more and more of it and grow over time. So that really helps with the overall efficiency of the business. The third factor, I would say, is more cultural. It's the fact that when we built -- when we started Datadog, to the point you earlier made, there were a lot of different companies doing monitoring and analytics and all that sort of stuff. It was very difficult for us to raise money initially because any time we said, well, we're a monitoring company, the first word that came back from investors was a crowded market, and it was perceived as being difficult. So we were always afraid of not challenging the company to the next step. And so we built a very efficient culture in the way we are running the business, in the way we operate, we look at return on investment. We look at productivity. We look at every single aspect of our business.

I wish I had been there when you had -- found it difficult to raise money as an early-stage private company. So we run a large-scale CIO survey. It is showing that 20% of IT budgets, if you look at it today, 20% are being spent on public cloud. They tell us it's heading toward 45% in the next 5 years, right? So there's this major re-architecting going on across IT. Datadog is -- you're critical, you're involved in all this cloud activity. Do you have a view on where this is going to settle out, that -- this mix between cloud and on-premise?

Yes. So we actually don't think of it clearly only as cloud versus on-premise. We think of it as public cloud, private cloud and legacy on-premise. So to us, legacy on-prem is mostly going away. I mean, it's never really fully going away, I mean, the same way the mainframes are still around and not away yet. They might never be. But it's going to compress to a very small sliver of what enterprise needs are. And then the rest is going to be public or private cloud. And we're actually working equally well in both, in public and private clouds. That's fine for us. We've seen different gyrations in the market traction there. Like when we -- in the earlier days of the company, we saw public and private cloud grow neck and neck pretty much. I remember running the analysis a year before IPO when we were pretty much 50-50 in terms of the balance of workloads we had in public cloud versus private cloud. I think more recently, you've seen public cloud grow a lot faster, and that's because it's been a much less risky approach. If you want to be certain you're going to be successful, you just go with AWS or Azure or Google, and you're going to be fine. Whereas if you try to roll your own in a private cloud, chances are you shoot yourself in the foot in some way, and so -- but that might change again as the technology maturity increases, as new technologies like Kubernetes, for example, allow you to have some form of a common control plane for both public cloud and private cloud. We might see some of that go back and forth between the hyperscalers and the private infrastructure [ ability ].

So you have such a direct window into that world. What type of industry structure do you think is going to evolve if we said -- just said, okay, just look at the hyperscalers you just mentioned. AWS, Azure, Google. Do you think this is going to end up being kind of a Coke and Pepsi structure? Do you think it's going to be sort of more of a typical kind of 3 company industry formation? Do you think it's going to be something more diverse than that?

So right now, it's really 3 plus. So there are 3 global players, and then there's a number of local players. And the local players are not going away if only for regulatory reasons. And the 3 global players are all -- I mean, they're not equivalent in scale. Like we all know Amazon is bigger, Azure is second and Google is a little bit behind. But still, like, roughly speaking, they are all very, very large scale and they're not going away. We see that as a very, very healthy environment, by the way, so there's a lot of competition there. There's a lot of innovation, and it gives customers a lot of choice. And one of the worries, I would say, 5 years ago was that the cloud was so dominated by AWS that it would lead to unhealthy practices down the road. And obviously, ecosystems are worried about that, the customers were worried about that. Everybody remembered the Oracle days and things like that. So I think that future is -- doesn't exist anymore. I think now, we're looking at multiple vendors. The biggest question for me is, when it comes to data residency and locally -- data needs to remain local to certain geographies. Is that something that is -- is that an opportunity that's going to be seized by local players? Or is it going to be seized by the hyperscalers with some form of a local structure? And I don't know that yet.

When you say local players, you're talking about Alibaba and others?

Yes, Alibaba. Europe is busy building their own, right? There's a few smaller players there. You have similar initiatives that are popping pretty much everywhere.

Okay. So if we look at the solutions you have today, Olivier, you're running across infrastructure. You have APM, you have logging, you have security, you have other products. I would love to get a sense of how much of the opportunity you think you've actually penetrated right now today. And so if we -- for instance, if we just look at infrastructure, what percentage of cloud workloads do you think are currently being monitored or managed with Datadog or some similar solution?

So I point you back to my first answer, which was that we have about an order of magnitude more to get observability in general with the existing cloud workloads. One way we look at it is if you think of us as being a fraction of what our customers pay for the infrastructure, we price the product so that we would be somewhere in the single-digit percent of their cloud deal. Meaning, anywhere between 1% and 10%, depending on the shape of their business and how much of us they need. As we keep adding to the product, we think we can take that from maybe between 10%, 20%, something like that with the product expansions we're doing right now. And if you look at where we are today, we're about 1% of the combined revenues of the 3 hyperscalers, so that gives you an idea of the penetration we have today, what we can do incrementally today with respect to the workloads that are there and additional tools we can have in the future on the product expansion.

Okay. So -- and of course, the cloud market in aggregate is a market heading towards many hundreds of billions. And if we stretch out far enough, perhaps a trillion. When -- if we kind of look at this math with the hyperscalers. So they're still growing, they're generally growing 30% to 40% plus. As you've said, you're nowhere close to fully penetrated, you're minimally penetrated. Is it logical to think that there -- from here, there would be a pretty long runway for Datadog to continue to outgrow those cloud platforms themselves?

Yes. So obviously, we think we're very, very early. We have a long runway, whether that's on market penetration, which itself is gated on us building our go-to-market teams and scaling the company. We also have a long runway in terms of the product expansion to cover all of the needs our customers may have that led to data application, the infrastructure and everything else about that revolves around it.

Okay. So Olivier, there's a notion of market convergence. I cover Elastic, I cover Splunk and Sumo Logic. I think investors have been in tune with this idea for a while that you're going to have several distinct capabilities. If you think about logging, monitoring, security, other areas, they're going to kind of converge into a single platform. You're going to -- basically, you're going to be able to measure the internal state of the system, right, by kind of looking at its outputs, and that's sort of the notion here. But from a single pane of glass, I assume you're subscribing to that view. Maybe you can walk us through that. What do you think is going to be the winning strategy to kind of emerge here and have people look back and say, this is the company that won that convergence?

Well, I mean, the single platform was kind of always our thing. So we -- we were the first to actually offer at the same time APM logs and infrastructure monitoring. And our belief is that we should not stop there. There's many more functions and many more -- which used to be different categories that we want to bring under that roof as well. So we've talked about bringing security under the same roof earlier. We think there's a number of other developer specific-use cases we can bring under that roof, and we've started doing that. We think there's a number of ITSM use cases we can bring under that roof. We think there's also a number of real-time [ BI use ] cases such as the premiums there being that if the application is running your business, pretty much all of it. If you understand your application in real time, you understand your business in real time, which is what we see our cloud native customers do with our platform today. So I think there's no shortage of unification that we can build on top of what we have, so we're very busy doing that.

How do you think about transitioning to kind of a second act? Because we -- you've done very well riding this wave with a forward vision. Across the software landscape, there are so many companies that they don't make it. They can't make the second leap, right, so there will be a tech transition and something will go wrong. Do you think Datadog can avoid that? And are there -- can you give us any window into what are the tech transitions coming up that you're monitoring for?

Well, the tech transitions have been the name of the game for us. So when you look at cloud transformation, when cloud transformation started, people were moving from running software on VMs to running software on cloud instances. Since then, the migration has happened from cloud instances mostly to containers, and then to orchestrated containers with Kubernetes. Now you see some part of the workloads moving to serverless, some part of the serverless being run on the edge or some slightly different platforms. So all of that has been happening, it has been happening in quick succession. So not being part of all that was not an option. Like, the players that couldn't do that just dropped out at the game. So I think it's become table stakes. It's fairly different from what might have happened in the space 20 years ago where, if you latch upon a tech trend, you were good for the next decade. That hasn't been the case. Now where we think the market is going and what the next one carries, it's really bring more people into the mix and bring more category and their categories under the same roof. In terms of finding our second act, actually, I think our second act for us was when we proved that we could actually expand the platform on top of what we had [ engineered ], which was in faster monitoring, and have customers really use a unified platform and adopt these new products frictionlessly. That started happening, I would say, in 2019 for real, and that's what decided us to take the company public. That's when we thought, actually, we have all the long-term opportunities, growth opportunities we think we can have, and we're confident we can actually deliver that vision for a very long time. So I think now the third act might be when we actually can bring all of these other different functions that were not on Datadog initially. All of the security folks, a lot of the business folks, and have them use the platform all day long, I think, will be the next step.

Why don't we -- we have about 6 minutes left. Why don't we check for the audience? If you have a question, go ahead and raise your hand, and we will run a microphone to you. We have one here.

So it was refreshing to hear you talk about the private cloud. Can you talk a little bit about how your consumers or customers are using the private cloud, and what applications you're seeing kind of migrate there?

Also today, the -- so the [indiscernible] of running product cloud is really Kubernetes. That's what brought in back from the date, I would say. Before that private cloud initiatives were usually too complex. They were typically under delivering, and with Kubernetes, we see more customers opting for it. In terms of what they can run on private cloud, it really -- like, really, we can run on anything that could run on public clouds instead. I think they'll go private when they don't feel like putting the data on public clouds or they feel data is riskier or would jeopardize maybe some more of their regulatory accreditations and things like that. But we see pretty much the same portfolio of applications in public and private.

Other questions from the audience? We have one here.

I was wondering, you went for a consumption-based pricing model. Some of your competitors have decided to -- not to go for that. So at scale, is that still the model that will be prevalent? Do you think at some point, you would switch back to more like a non-consumption based pricing model?

So we're not actually a consumption model, we're subscription. But the unit for the subscription actually differ based on the products we sell. They tend to be mostly workload based as opposed to being user based. And in great part, that's because we actually wanted to not have barriers, to having more users on the platform, and that was the starting point behind the company. And it was a more natural way to scale and track the value we bring to customers. So if you go back to what I said earlier on who would price the product, we thought it would be a certain percentage of the customer's cloud bill. And the best proxy for that is to actually charge depending on the amount of infrastructure they have. So the way our revenue works is that the growth is driven by the number of resources our customers are going to inspect or monitor with us. It's recurring in nature because the resources are there. They're coming back. When you're starting this test containers, they are there and they keep coming back. But the growth in and of itself is going to happen whenever our customers start those things and migrate those things into the cloud, which is what gives it the usage-based aspect.

So Olivier, one of the biggest questions that's been floating around for several months, and certainly at this conference, is the notion of a kind of a digital pull forward or a pandemic pull forward, which would have been happening, say, last year that, as a result of lockdowns, right, there was a drive of activity. It accelerated movement to public cloud, it accelerated digital plans. And that, perhaps, you would get into a lull, right? And your results have not shown that at all. How do you answer that question or that notion of pull forward? And I think more importantly, any further thoughts on just the sustainability of growth coming out of the pandemic?

Yes. So that's a great question. The -- what we saw is -- so first of all, customer base is very diversified. So we have cloud natives, we have a lot of traditional enterprises. We have companies from every single part of the stack, from the industrials to transportation to services, to everything you can find online. And what we saw early in the pandemic was that we had a massive acceleration of everything that was online, obviously, but that was counterbalanced by a pretty brutal deceleration of everything that was not online. So if you look at our customer base, our customers that were doing e-commerce and online media were booming. But our customers in transportation and travel, on the other hand, were fighting for their life and shutting down everything they could shut down. If you look at it to navigate, we actually didn't see an acceleration at the time, we saw a slowdown because more of the industry was slowing down than was accelerating at a time. And then after that, as we go deeper into the pandemic, we saw a reversion to the growth rates that we have seen before the pandemic which are more indicative of the rate of cloud migration that we've seen at pretty much since the creation of the company. So now if you see what's happening now, like as the pendulum swings back, you see pretty much the opposite happen. Like, the companies that expanded very quickly in e-commerce and online media are reverting back to the main. They're not growing as fast. And you see that, and you've seen that in the news recently. Whereas the companies that were severely hampered on reaccelerating, for example, transport and travel right now are doing great, and they're expanding very quickly. So from where we stand, the picture doesn't change all that much. And by the way, we've seen that before, too. Like, even before massive events like the pandemic or what's happening right now. We saw, at various points in time, various parts of the customer base are going to do way better than others. Some parts of the industry end up always being depressed at any point in time.

So Olivier, it's a perfect image to end on here, where you mentioned the pendulum actually swinging back and coming back to us at this point in time. I can't thank you enough for taking the time out of your busy schedule to be with us here today. So thank you.

Thank you.