Confluent, Inc. (CFLT) Earnings Call Transcript & Summary

All right. Let's get started. Hi, everyone. This is Pinjalim. I work with Mark in the software research team at JPMorgan, and I'm delighted to have here Jay Kreps. He's the CEO and Co-Founder of Confluent. Jay, welcome to the conference.

Thanks for having me. Happy to be here.

Great. I don't know -- I was just telling you I was actually a little bit of a nerd, so I played around with Kafka for some time and created some topics, and it's fascinating. But let's start with maybe give us a little bit of a background on yourself and maybe on Confluent, a few lines, just for people who may not know about the company.

Yes, happy to do it. So the open source technology is Kafka. The company is Confluent. The technology came out of LinkedIn some time ago as an open-source project. And probably the easiest way to understand the space is in comparison to the world of databases and data storage. So classically, the problem around data has been how do we take a pile of data and store it and keep it safe and look up the little bits at the right time and process it at the end of the day? That's kind of the world of databases. If you think about how software is used today, there's not just isolated bits of software here and there. It's really much more intertwined. And so data has to flow between systems. Systems have to react and respond and be up to date across a big company. And so that data in motion versus data in rest, that's kind of how you can understand Confluent is. It's not just about the storage of data, it's about that kind of real-time flow or streaming and the processing of it. So how would that play out in like a real use case? Well, you can imagine like in a retailer, you have sales that are continuously happening maybe in stores all over. You have the movement of inventory with distribution centers. You maybe have an online part of the business that needs intertwine with that. You need to be able to take that kind of stream of sales as it were, that are happening all over and have a real-time picture of what you've got, be able to drive kind of inventory and logistics, be able to drive promotions for customers. That's kind of gone from something that was operated on a monthly schedule as a kind of batch processing job to something that's much more real time. And that's common across virtually all types of businesses. Retail is easy to understand because we kind of have some idea of goods and services in stores. But whether it's large banks, tech companies, insurance, automotive, really across the board from big manufacturing problems to customer interactions to payment systems, that's increasingly driven around this technology for data in motion. So Confluent has really helped to pioneer this area and build a complete cloud offering around it that we take out to customers. So hopefully, that's quick enough to count as your 30-second intro, but that's a very fast introduction to the space.

That's great. Yes, the underlying change in the data architectures that you're just talking about, the need for real time. Maybe take a step back and talk about why do we need real time right now, right? Why does the previous architecture does not work?

Yes. Well, I think a lot of this happens as software is directly participating in the operation of the business. So if it's just around the edges, the Office apps and whatever, maybe you don't need to be that up to date. But once the production of goods and services, the interaction with customers, once there's a big role for software in that, and once the software has to take on more of the kind of intelligence and decision-making and integration across, that's where the view of the world has to be up to date with what's happening in the world. And that drives this need for data to move around in real time and the need to be able to make decisions faster than once a day or once a week or once a month. And that's kind of what pushes this architectural shift. And if you think about it, it makes sense. The only reason you had these kind of end-of-the-day batch schedules was mostly that's kind of how mainframes worked. You submit the job, it goes into some -- it runs. Obviously, we had some history of building systems that way, but it is much more logical in a digital system that the data would flow and you would react to it as occurred. And so it's been logical for a while. It wasn't practical with the technology to do it. And it's become practical. And then now it's become necessary. I think there's significant pressure on customer experiences, what people expect there. There's significant pressure in the operations of the business and how that works that has really driven the adoption of this. And I think that's what's kind of caused the change overall.

Great. Now obviously, even base architectures have been there for quite some time. I remember TIBCO talking about the 2-second advantage in 2010, 2012, right? So these message buses, the messaging queues, the ActiveMQ, the RabbitMQ. Why don't they -- why aren't they able to capitalize on this? Why is Kafka really good at capitalizing this change?

Yes. So there was a bunch of previous technologies that went after part of the problem, and they kind of only solved a bit of it. And why did they only solve a bit of it? I think people didn't have a good idea of how to do something more general. And so in there, I would put these kind of message middleware, the TIBCOs of the world, as you said, some of the ETL products that were just about moving data from data store to data store. And these different products, maybe around application integration, they would focus on kind of 1 type of pipeline, 1 type of data. The reason that, that doesn't work is companies are increasingly interconnected. If you're going to build 1 kind of custom pipeline for every pair of applications or systems, you're going to have a lot of pipelines, right? The pairs of things go up faster than the things. And so that was, combined with, I think, an understanding of how to solve this correctly, which is really more inspired by the database world than it is by the kind of messaging middleware world, which is around real-time processing capabilities and how you would take some of the ideas in databases and apply it to real-time data movement. And to do that, you have to have kind of cloud-like systems that can grow and scale with all the usage across the company. There's a set of things that have to be there, and you have to have the pressure to need to do it. And all of those, I think, have enabled this next-generation technology to dominate some of the older technologies across a bunch of verticals both in the capabilities, but also just the raw adoption. Like you would see much more Kafka adoption than all the things you kind of listed put together.

Yes. On that topic, I think in the last earnings call, you kind of showed us a chart which was mind boggling. Kafka adoption, something like 10 or 15x [indiscernible] I believe, right?

Yes. Yes.

So obviously, Kafka has done really well. Now how are you thinking about converting those free users, right, to monetization, right? What is the trigger that you typically see that people follow?

Yes. So this is an area that's kind of in flux, I would say. Companies that have an element of open source -- on-premise, it was really the way you described where it's kind of converting over. In the cloud, I think increasingly, people have a preference to start with a managed service. If you talk to kind of modern software engineers, a lot of them just -- that's their first step, is go sign up for these services. That's a mindset shift, so it doesn't happen overnight. But especially as companies have a little bit more pressure on these kind of big expensive internal platform teams with lots of engineers, they're looking for managed services that just kind of do it. And so I would say that's an area that's shifting. The reasons people come to us beyond just, "Hey, I just want a managed service that does it," are our platform is cloud native, meaning it takes this capability, but it lets you just kind of expand elastically and use what you need on demand without all the operational burden running it. It's complete. It's really a full stack for using this kind of data in motion. How do you process it? How you connect it up to different systems you have? How do you govern and secure it? And then available everywhere, so we offer this across the 3 major cloud providers as well as a licensed version you can run in your data center to hook into older legacy systems. And we connect all of those up. So those are kind of the 3 areas of differentiation that we offer.

Okay. Got it. I want to shift gears a bit and talk about the demand environment. There is so much volatility in the stock market, obviously. There's changes in the macro side with inflation rate hikes and whatnot. Eastern Europe and the war in Ukraine is a different thing, geopolitically. So describe kind of the demand conditions that you see going into June? And what do you expect kind of the second half to play out?

Yes. We've been watching this closely for probably all the same reasons everybody else has. We haven't seen any pullback in our customers. Nothing related to the Ukraine war that's kind of significant as a trend. So we haven't seen any slowdown yet. Obviously, we're kind of watching it closely, and we would adjust if we did.

Got it. Let's talk about Confluent Cloud. That's a big driver of the growth, and it seems like you're more than -- I think more than 50% of new ACV is now coming from Confluent Cloud. I wanted to ask you, what is driving that? Is it a concerted effort from a go-to-market perspective? Is there an element of tailwind from less people available to manage Kafka clusters? Is it just an organic growth? Talk about that a little bit.

Yes, I would say it's a combination of things. The most fundamental thing is like these managed cloud services are such a dramatically better way of consuming any data system. It's like what customers want. And so as your service gets good enough to really meet the demands in different segments, it becomes very natural that, that's what customers want. And so we've seen that shift in our business. Now for customers that are building things on-premise, they can't really get the managed version there. And so it wouldn't be applicable for them. For those who are building on-premise and in the cloud, they would often consume both, and actually hook these together and have data flow between environments. But yes, I would say the #1 thing that has driven the growth there is just actually the fact that people would prefer to get a managed offering and then our managed offering has gotten extremely good. And that's been a journey for us. We started this not long after we had our software offering in place. The company is overall relatively young, will turn 8 this year. But that's come to scale pretty quickly. And I think it's just a better way of serving customers because you can really take the problem off their plate and have them just focus on how they get value out of their technology rather than having to take on a new competency in running some new data system.

Just to double click on that. Is there a go-to-market component of it where you're positioning cloud first today?

Yes. Yes. There's a significant go-to-market component to -- I actually think there's 2 reasons that companies that take on both cloud and on-premise often don't do well at it. One is it's actually quite different technologically to deliver a cloud product. And so your engineering team has to be set up to do that well and you have to figure out how you get synergies out of it. People focus on that because that's the more natural aspect. But I think the go-to-market challenge is actually as big. And I think these SaaS go-to-market models that are really kind of full life cycle from how customers land to how they expand to how they get to really large scale. I think they're quite intricate, and that's been a huge focus for us is to try and build that full journey. I think it's very important in this domain because there's a natural expansion to your use of Confluent, where you would start with 1 use case, but the data that would flow in from that use case would go out to other use cases. So in that retail example I gave, maybe you have the stream of real-time sales, maybe that's for some kind of personalization or promotion use case. It's going to drive many other things over time. And we want to set up not just to land the first use case, but to be able to grow this into a significant platform over time. And so how you hook together the self-service adoption with a more transactional single-use case sale to the spread to other sales to kind of a large-scale platform and for customers that are really at scale with this, they would think of this as kind of the central nervous system that plugs together all the different applications and exchanges large amounts of data. And there's a really sizable number of customers that are kind of doing this now. And so to orchestrate that, you do have to have kind of, I would call it, full life cycle capabilities and be able to tie them together in a way that they all naturally fit into that journey. And so that evolution has been a big focus for us in making sure that we have all that.

Yes. Okay. Now I remember during the IPO, I think you were saying that the cloud is kind of a complementary motion, right? You -- it -- well, it was not primarily focused on Fortune 500 and your large customers at that time, but it seems like we're actually getting some big cloud deals, including the 8-figure deal that you talked about last time, right? Talk about -- has your -- some of your largest customers been playing and adopting cloud as fast as some of the mid-market customers?

Yes. Yes. That's true across the board. So cloud dominates our total customer count, and now we have very large cloud customers across virtually every industry. And so has good representation in our 100,000-plus and 1 million-plus customer ranges, and we think that's really healthy, right? Again, we see this as kind of a journey for the customers. You also alluded to another thing that's very important to understand in this domain. So typically, when a company has a software product and a cloud product, it's a conversion story, where a customer will have A or B, right? Like you had Jira on-premise, then you got cloud Jira, you got rid of the old Jira, right? So Atlassian, it's very much a conversion. In this domain, that's actually not the case. You end up having to have our offering everywhere you have applications that it needs to plug into, and so when we look at the -- what's growing, actually, that portion of customers that are hybrid that have Confluent on-premise and in the cloud, that's growing extremely quickly. And the NRR in that customer base is the strongest of all of our customers. So we've given out that our overall NRR has been above 130% the last -- some quarters. And last quarter, we said that cloud was above 150%, but that hybrid customer base it's across both is actually the strongest. And it makes sense because there's now significant data that has to be exchanged between these older on-premise systems that are running key parts of the business, some of the newer investments in the cloud. That's a really important problem for our customers to solve.

Just to round out that cloud discussion, how should investors think about just the trend. More than 50% coming from cloud now in new ACV. How should people think about the trend? Because there is -- we'd get pushback from investors who are too much focused on this good sequential growth in cloud and try to kind of annualize the sequential growth, although it is based on consumption, so maybe that's not a fair way to think about it. But there seems like a little bit of variability from quarter-to-quarter. So how do people think about that trend?

Yes. It's challenging admittedly because the company is only 7 years old. So it's not like there's some decades of history. And of course, through fast growth and like a rapid shift to cloud, all those patterns change pretty rapidly. But yes, that's one of the reasons we've given out that the last 2 quarters, more than 50% of the net new ACV coming in was for cloud. And the reason for that was there's been a, I would say, good percentage growth in RPO and cRPO and people obviously want some idea of like, hey, well, what is that, is this big on-premise deal, people would ask us some of the answers as well. We obviously have some big on-premise deals, but there's also a really sizable progress in cloud. There will be some -- if you look at the amount of revenue added in a quarter, it does tend to change quarter-to-quarter in all these consumption companies. It's true for MongoDB. It's true for Snowflake. It's true for us. So we also gave out that through this year, we will expect kind of sequential growth in that raw quarter-over-quarter, which is, I think, another positive sign that comes out of having a little bit of that visibility into where it's going. So on the whole, we felt extremely good about the progress in cloud. It's been ahead of our plans both through last year as well as what we said for this year.

Got it. Changing gears a little bit about the coopetition with hyperscalers. You have -- all the hyperscalers you compete with, they have their native solutions. But you actually are forging really strategic partnerships with each one of them. AWS, I think we heard in Q4. You talked about Microsoft as well. Talk about that coopetition. Talk about the strategic relationship. How important are those, the AWS one and the Microsoft one? Do you have go-to-market motions associated with it? Is it fully technical? Just talk about that a little bit.

Yes. It's very important. So the dynamic is complex. I would say everybody who sells infrastructure software, if you're successful, then you will have coopetition with the cloud providers in some way because they're really across many different product categories and lines of business. Some of those, they're good in. Some of them, they just have some entrant. I wouldn't say that they're really like blazed massive new paths in data in motion, but each cloud provider has on the order of 3 to 5 technologies in that space that we would compete with. And so we cooperate with the cloud provider as a whole. We would compete with a few of their products. Those products would, of course, compete with each other as well. And so early on, I would say we didn't get much out of the partnerships. I think now we're getting much more. So we sell through their marketplaces, which means you can kind of draw down against your committed spend. It means that their sales force is incentivized to sell our product. And I think we have the thing that customers ultimately want and want to build around and that has the ecosystem. And so that's helped us progress really well. So the kind of strengthening of the partnerships has been across the board. It's been on product integration, which is really important so that we plug into their environments. It's been on go-to-market, which is an area that I think is contributing, but still building, and we're doing a lot of activity now. Demand gen, kind of joint selling stuff. I think that's been successful. And then also some of the kind of internal technology stuff, we're obviously a big user and builder around some of their internal layers and have needs, and we want to make sure those are met. So on all those dimensions, it's been very helpful for us as those partnerships have grown.

Is multi-cloud adoption a part of that conversation, right? As a customer, they would probably require or want to have support for AWS as well as Jira.

Yes. Yes. So multi-cloud and hybrid cloud is a huge thing for our customer base. And I would say that as we were getting started, it was more of a theoretical concern. Well, that you're going to have multiple clouds. It's like on paper, maybe yes. I think on paper, people felt that, that would be like the same application running in multiple clouds. In practice, what's changed now is it's not theoretical anymore. A lot of companies now have significant investments across multiple clouds. And it's typically not the same application into clouds. It's typically different workloads, different parts of the company, the company they acquired. The reality is they're just spread across different environments. So the data flow across that becomes incredibly important. And that's kind of unique to this area because it's about connectivity, like the central nervous system effectively has to span kind of all the parts of the body, and that's been a pretty significant differentiator for us versus any of the technologies that are just kind of unique to 1 cloud provider. Like ultimately, you could solve the problem for 1 application, but you can't solve it more broadly across the organization because you just don't span across the organization.

I think that tailors perfectly to my next question, which is the central nervous system, right? The idea of Confluent sitting at the center of all the data flows within an organization. How should investors think about the progress there? And where -- what other metrics should we look at? Is it the 1 million customers that are more mature in that kind of motion at this point?

Yes, I would look at this -- first of all, I think this is one of the things that's most exciting about the space and about the company. So the possibility of what can happen in this area. I think there's an opportunity to build something that's as big as databases have been. And I think we're in a pull position to get that. So I think that's fundamentally one of the exciting things. If I was trying to validate progress on that, I would pay attention to 2 things. One, how do customers talk about their architectures? Where does this sit? Does this rank as something significant? Do I see it showing up again and again in the pictures? That's kind of the qualitative of just like, "Hey, is it there?" Quantitative, yes. Probably the best proxy metric. There's no gap central nervous system that we -- stat that we put out, but yes, that $1 million plus, when a customer is spending more than $1 million, it probably means they have a significant number of use cases and value it. And so watching that customer count tick up, we watch it as kind of a proxy metric for, "Hey, these are the people who have really gone big in this area."

Understood. Now talk a little bit about the expansion dynamics. You had a really stable net dollar retention of 130% for -- or over 130%, I should say. For some time, you had disclosed cloud NRR, which is 150%, I think, and you had said the hybrid ones actually are even bigger than that. Help us understand what's driving that. Is that largely cloud growth driving node growth? Is it the high-risk functionality SKUs? Just talk about that.

Yes. I would say the #1 driver is the spread of use cases. And so of course, an individual use case could get bigger. Maybe the business grows. But the #1 thing is what I described before, which is in that retailer, once you have the stream of what's selling that's available in real time, it's probably the only place in the company you can get the real-time feed of what's selling. And there's probably 500 other use cases for that data. And so where are they going to go? They're going to come and get that feed out of Confluent, and as they do, they bring in their own streams of data that they need to process with that sales data, and that creates something that attracts other applications. And so that spread of use cases is really the primary metric by which we would grow within a company. And the set of things we can grow into is actually quite vast. If you look at the large customers that we have, if you look at even some of the large open source adopters in kind of Silicon Valley tech, this is something that can have thousands of use cases within an organization. And so our goal with our customers is get them all to that. But that does happen over time. Those new applications are created or changed. It's not something where you just kind of roll it out all at once. Hence, the importance of that kind of customer growth-oriented go-to-market that I talked about, that ability to like land efficiently and grow over time. Building that, we think, is the key to efficiency in sales and market to continue to grow, to keeping and even accelerating the NRR. That's why it's so important for us.

Maybe talk a little bit about the streaming analytics part The ksqlDB, we don't hear enough of it. How broadly adopted is it? What are you seeing there?

Yes. We've seen really good traction with that. So you can imagine as you start to have capabilities around real-time streams of data, what are the next 2 obvious things that you need? Well, one, you need connectors to actually get streams out of other systems, so we provide that. Beyond that then, you want to make it easier and easier to use and build on these streams. And so ksql comes into play there. So rather than having to build custom code, you can now take kind of the database skills that your team might have around SQL and you can apply it to these streams as they occur. That can be much faster and more dynamic to work with. And so that's kind of 1 step up the stack in terms of ease of use, ability to create value. I don't think we're done there, but that's an important one for us. If you think about what we're doing with this idea of real-time and streaming data, it's kind of logically very obvious that you would want that. So what -- the project for Confluent is to make it easier and easier and easier to consume and get the value out of. And as you do that, it becomes the default way that data is used and flows. And so that's kind of 1 step up the stack from kind of something that is more equivalent to a low-level programming tool, like a database, to something it's more like a relational database that's easier to use.

Okay. Now shifting gear about investments. Talk a little bit about the core areas of investment in 2022. And is there a bifurcation that between cloud and self-managed at this point?

Yes. So it's not purely 2 separate pools. Like what we ship in cloud, we take those core components and offer a software release around a subset of them a couple of times a year, and that's our on-premise environment. So we think of it as a kind of cloud-first development model, where we're building this massive at-scale cloud offering and then we're offering some of those components on-premise. That gives us really deep IP that is unique to us across both that differentiates us both from competitors in the cloud as well as competitors on-premise. So most of those investments pay off across both, but we do build in a cloud-first way. And for some layers, they really are available only in our cloud product. So you can kind of think of that as being something of a superset of the capabilities. And so that's an area of investment. We're also invested in that kind of customer growth go-to-market, that kind of full life cycle go-to-market. I think that's kind of the key, we think, to really driving continued expansion and efficiency on the go-to-market side.

Understood. Let me see if there are any questions in the room. If you have any, maybe think about it, I'll come back. But on that profitability versus growth point, I think you were alluding last quarter that investors should see a higher improvement in margin in going into 2023 and maybe exiting 2024 with positive operating margin on a pro forma basis. Help us understand what are the levers that get you there, one? And I guess the second part to that, as you move towards that model, how should investors think about the growth, right? Is that is the Rule of 40 or Rule of 60, what should investors think about?

Yes. So we gave a midterm and long-term operating model as part of the materials around our IPO. And one of the things that we wanted to give was a little more visibility in what's the time frame for getting to the midterm model. And so we said, "Hey, on operating margin, that will become positive coming out of Q4 of 2024." To your point, yes, what drives that change? What is it that we have to do? Is that some new thing? Yes, it's not some new thing. It's just the longer range plan that we had. The key for us is, I think, on the sales and marketing front, both bringing up the percentage of fully ramped reps, which has come out of the hiring buildup that we've done, and then I would say, connecting these go-to-market motions, the self-service, the more transactional sale, the enterprise sale, better integration across those. I think both of those are going to drive efficiency. We've seen really good success as we've built out kind of the self-service and as we've built out our kind of commercial sales team, that's a much kind of higher velocity. Those have done really, really well. But if you think about Confluent today, we effectively have kind of 3 go-to-market motions that are connected. We think we can get more leverage out of more deeply connecting those to kind of drive growth. And we've seen the kind of improvement of NRR, which is kind of the key driver of that. But if you think about hey, how do we drive go-to-market efficiency over the next couple of years? I think it's just doubling down on that, which allows us to continue growing quickly and do it with increasing efficiency. And then on the R&D side, we'll see efficiencies just as we get to scale, some efficiencies on the G&A side as well.

Just to double click, do you want to talk about the growth glide path on as you reach towards profitability, maybe beyond that? What should investors pivot towards? Do you have a framework in mind that...

Yes. Yes. So we've given this kind of growth and profitability framework that gives kind of, hey, long-term what do we think we're at; and midterm in this kind of higher growth model, where are we at? That -- I think that answers a lot of the questions people might have around Rule of 40, some other aspects that will give you a lot of those stats. I would kind of refer to that. That's something we took pretty seriously as we were building the company, as we were going public, that remains kind of our commitment on like, hey, this is what we want to be operating at. And I think we've shown like really good progress towards that. I think for us, it's just continuing to do that.

I want to see if -- okay.

Just one quick question. How many salespeople did you have at the end of the year last year? How many are you targeting this year and, say, '23 and '24?

Yes. I don't think we've given out the kind of raw count. I think when we -- what we've given people is 100,000-plus customers, which is probably the better way of modeling it. So we haven't -- we don't do kind of sales headcount-based stats that we release.

But is that the key factor for your accelerating your growth?

Yes, yes, yes, absolutely. So the kind of ramped sales headcount is definitely a key determinant and sales productivity per head is a key determinant...

Or just how about your sales cycle? If you think about like a -- not an average, but a, say, $1 million deal. Tell us a little bit about that cycle. I know every deal is different, but give a little color on that.

Sure. So there's really 2 dynamics at play. So because of the open source, we have some customers that are starting with us and starting small and growing. Those tend to be very quick, but they start smaller. And the large size comes with the growth over time. We have some customers which started with just the open source and then we're converting them over after they have some applications, and those might be bigger. And so typically, the conversions take a little longer, whereas the kind of organic starts are faster, but the conversions come in at a larger size. And so that's the breakdown. The -- for these conversion deals, you would expect a few quarters if it's a larger conversion. If you were signing up to start something today, you could go sign up in the course of this conversation and be up and going. As it sounds like, maybe you tried out. So that can be much faster. You would still have some life cycle for your project as it went to production and ramped up usage, but that ramp is not tied to the commercial transaction.

Any other questions? Lastly, I guess, I will be amiss if I don't ask you this, but you're obviously a visionary running kind of a very successful open source company. As you look out into the open source projects at large, nothing to do with Confluent, maybe in general, is there any open source projects that you think are interesting that investors should pay attention to?

Yes, I think there's a lot of interesting stuff that's happening out there. Probably the ecosystem that's maybe the most interesting is the set of tools. To me, it's a set of tools around machine learning. So the kind of PyTorch and TensorFlows and so on. It's very complicated. And I would say that the line between open source and kind of cloud services has gotten blurrier. In part, I think the success of open source projects was around creating kind of an open standard people could use. And in part, I think it was a way of driving a bottoms-up adoption model. The bottoms-up adoption you can now get through kind of SaaS-type sales models with cloud. So I think you're seeing more things that are kind of some hybrid between the 2, which is, I think, probably a healthy thing. And so there is really a whole interesting set of products coming up in a bunch of different ecosystems, but I think that machine learning ecosystem is very interesting. It has a lot of associated data flow that comes into it, which is one of the reasons I pay less attention to it. I don't have any particular insight on to what the winners and losers will be. I just think it's an interesting space.

Okay. With that, let's wrap up. Thank you so much for your time.

Thanks, everyone.