The Exchanges

Q increasing, like, uh, you know, which is, uh, uh, amazing to watch. Um, but going back to the beginning of that, so the, the, the, the vision was always to be a platform because, you know, there's, there's that, um, kind of cliche in, uh, in venture and, uh, in startup circles that you need to be a tool before you are a platform. So how do you navigate that?

A Well, I mean, initially not very well. So, I mean, uh, The, uh, the idea was, yes, we were going to be this platform that was, you know, why we started the company. Um, and we had this vision of multiple data sets, multiple, um, teams, multiple sets of use cases, all meeting into this platform. So we started a company, we were super excited, we applied to all sorts of incubators, we got to a Y Combinator interview, and, uh, and then we didn't get into Y Combinator. I have an email from Paul Graham that says, um, You know, a platform is only as good as its first product, and you don't have a first product, blah, blah, blah. So, you know, um, so it was not, like, you're right, like, it actually was, uh, was a tough sell. Um, and to be fair, I mean, when we started, um, raising our product, like, the first beta of our product, we did call it a data platform, as opposed to calling it a very specific use case. Um, and everybody loved the idea, like, the users loved the idea. Um, but, People were not coming back to the platform, and also, but nobody was paying for it. Um, so that was a bit difficult. At some point, we decided to name it monitoring, which was the, the world, the name of the category before us, really. Um, infrastructure monitoring. And without making too many changes to the product, we had a few small things to make sure it could, it would be called that way. Um, imme…

AI assessment note: “At some point, we decided to name it monitoring”

Q And, and you store all of this, right? Like, I, I, I think I read that there were some, um, improvements around, you know, storing log data and that kind of stuff, but, uh, that's, you, you just store massive amounts of historical data as well on behalf of your customers?

A Yeah, and, and we did some things there. I mean, look, the, the biggest challenge with observability is that, um, any application can generate any, any arbitrary large, uh, uh, amount of logs. Um, and so the data volumes grow much faster, um, Uh, then our customers revenue, which is a problem. Um, and so for, to solve that, there's a few things to do. One is you create the right feedback loop so people understand what they, what they need and what they don't need in terms of data produced by the application, and when they send too much, you can fix it. But the other one is also you just need to be more and more efficient in terms of how you can send more data and store more data, and, uh, and it costing less. One thing we've done over the past few years is we, uh, decoupled the, uh, storage from the compute. Um, so it allows us to, uh, to store storage, which is much cheaper. Um, the, uh, uh, different, different basically, uh, uh, differently from the, uh, the compute, which is a lot more expensive.

AI assessment note: “Yeah, and, and we did some things there.”

Q Okay. Yeah, same, same general thing. Very, uh, very comparable journeys, uh, obviously other than the fact that you build a thirty-nine billion dollar market cap company, and, uh, I'm a VC with a podcast, but, like, other than that, like, pretty, uh, pretty close. What did you stay in New York?

A Well, first, because it was fun. So I moved in, uh, in 1999, thought I would stay six months. Uh, I had an internship at IBM in research, which was a very interesting place at the time. And, uh, and then the, uh, it was the tail end of the dot-com boom. Um, so there was a lot going on at that time in New York. There was not much going on in France from a tech perspective. So I thought it was a great time to start to stay and, and, you know, join startups. It was interesting. Of course, you know, then there was the dotcom crash, um, and 11, all of that stuff. So that was less fun. Uh, but I was, by that time, I was very attached to the city. I liked the dynamism. I liked the, uh, cultural aspects of the city, and I decided to stay. And I stayed until, you know, it was until

AI assessment note: “Well, first, because it was fun.”

Q Okay, so to start digging into it a bit more, so like, how does it work? I read somewhere you have your own, um, data model, so is that, is that fundamentally like a large, um, you know, database, for lack of a better word, and you have, like, you ingest just massive amounts of data to, like, to start with the backend?

A Yeah, at a high level, it's, uh, so we deploy all sorts of collectors on our customer side, you know, so, Uh, and all of our characters are open source, so customers are going to deploy agents and libraries and APIs, and we're going to crawl their accounts on clouds and things like that, so we're going to collect data in all sorts of ways. Uh, we send all that to our, uh, centralized service, and our centralized service, you can think of it as a very large database that is going to manipulate different data types, and those data types, you know, under the hood are going to be stored in, in different data stores, you know, so we have data stores that are, Appropriate for super high volume log events, or for time series, or for other kinds of.

AI assessment note: “our centralized service, you can think of it as a very large database”

Q And what was the trigger? You felt like you had, uh, something that felt like a repeatable process?

A Yeah, so we're starting to get quite a bit of inbound, um, and we wanted to have basically people following up on that inbound and, um, And, uh, without having the, you know, the whole team and the high touch, you know, approach to, to everything. Yeah. But by that time we, we had packaged the product. We knew what we're selling, who we're selling to, uh, how we, people got deployed with it, what the first, you know, six months of usage looked like from these customers. So we didn't have to, to make all of that up. I think when you hire a sales team, you, you need to have all that sort of in place. Otherwise the sales team is not going to figure it out for you. You still need to do that yourself.

AI assessment note: “we had packaged the product. We knew what we're selling, who we're selling to”

Q Some of it was for acquisitions. Uh, and then you've expanded into, um, into, uh, uh, other Like synthetics and like other things recently. Do you want to talk about like how, how you, how you think about this horizontal expansion and what, what you cover currently and what you plan on covering?

A Yeah. So, I mean, for the longest time we only had one product, which was doing infrastructure monitoring and which was doing metrics. Um, we didn't have the traces. We didn't have logs. Uh, so we started, you know, for the first few years of the company, we grew that product and we made sure it had the right feed, we made sure we knew how to sell it and everything else. We knew from the very beginning that we wanted to, um, bridge the gap between different teams and different silos, so we, we knew we wanted to add the rest. Uh, it was actually the, the biggest, um, wanted to be a step function on the trajectory of the company was when we actually managed to do that, and we noticed that we could actually also sell it, and customers are also going to adopt it in addition to, uh, uh, what we had initially. Um, the, the way we see it, there's a number of categories that are still separate, um, And we want to keep unifying them. You know, so you mentioned synthetics. I mean, synthetics for those of you who, uh, don't follow the, uh, the industry, uh, is basically simulating, uh, user behaviors, you know, so simulating users clicking on your applications, for example, or simulating, um, API calls, um, to continuously validate that everything works, um, even if the users are not on it yet. So it used to be an, I mean, there's an existing category for that. Um, it's a category that is…

AI assessment note: “there's a number of categories that are still separate, um, And we want to keep unifying them.”

Q Mm-hmm. Were there some terrible moments of self-doubt, uh, along the journey?

A Yes, quite a few, yes, yes. Uh, especially in the beginning, um, you know, so there's this, uh, you know, when you start a company, you have this romantic idea that, uh, you know, I'm my own boss now, you know, it's going to be awesome, we're going to, you know, kick butts, uh, we have all these great ideas. Um, and that lasted about three months, at least for us, it was about three months. Um, three months later, we had, we had no money, uh, we had no customers, we had no products. Um, we, it was also winter in New York, so it was pretty sad. Um, and I think these are the moments where you, you have to go back to the basics and figure out, okay, so what, what are we doing? Why are we doing it? Who are we doing it for? And make sure that you, you get that right, so you can actually carry it across to the other side of it, you know, when you, you have a product. Uh, when you have some money, when you have some customers, then you can actually have a, you know, a threat to pull, basically. Um, so, but these are difficult moments. Initially, I, I sucked at fundraising. Um, and I remember, you know, you fly to take all these meetings. I remember being, you know, hotel rooms, um, in, uh, you know, other side of the country, uh, having had, like, four noise in the day, like, in the hotel room. He's like, all right, so I need to call my co-founder and tell him, uh, you know, it's not …

AI assessment note: “Yes, quite a few, yes, yes. Uh, especially in the beginning”

Q But in layman's terms, is that like a, just like a gigantic cluster in the cloud that just like process, cranks this whole thing, and you have like connectors that feed data into it, and APIs to push data out of it? Is that, is that roughly the...

A I mean, well, I mean, there's a, there's a number of queues and data stores, basically. Um, and some of it is completely homegrown, um, after generations of, uh, of iterations. Um, And some of it is still, uh, like open source on the shelf, you know, that works very well for us. Like, we use a lot of Kafka, for example, that works well, and we still use a lot of Kafka. Uh, database-wise, you know, we mostly built on, I mean, we still use, in some areas, like some standard databases, like Postgres and things like that. Um, but we, uh, we obviously shell them and scale them a lot. Uh, but a lot of the, the core data stores, like the time series, uh, Um, the log data, the event data, all of that is on completely custom data stores that we've built over time. We, we published actually a, a few, a series of articles on our, um, our new event store, which we use for logs and traces, uh, which is called Husky, you know, not a dog name.

AI assessment note: “there's a number of queues and data stores, basically.”

Q I believe, uh, which, uh, is the, um, you know, AI product. I'll let you describe it better than I can. Um, but there was a little bit that you were doing, uh, and so, so generative AI has increased, um, Not just the scope of what you can do, but also the, the level of confidence you have in, in terms of, like, not working people up at night?

A Yes, but I think most importantly, we don't have to, uh, to lead too hard with it. Like, the problem with AI in general is if you start solely with the, uh, the promise of automating with AI, you're sort of on the hook to doing it all the time. And the general case for what we do is not something that you can solve with AI. Uh, I think Right now we'd be lucky to solve, you know, two percent, three percent, five percent of the cases with AI. Maybe in a year it's going to be 20%, maybe, but it's going to be gradual. I think if you, if you try and insert yourself saying, I'm going to automate it for you, uh, the incentive will be to try and do too much, uh, break the confidence with the, with the user, uh, and then, you know, that working out in the end, which has been the story of AI in our business. I mean, in the, uh, in, um, systems management and, Um, for the past 20 years basically, like that cycle has repeated again and again and again. I think our strengths come from the fact that we can pick and choose. We can say, hey, we're here for observability, we're here for security, we're here, uh, to, uh, make sure you can solve your issues, and little by little, we'll do more of it for you, um, up to the point where, you know, maybe you have only one percent of the issues to solve yourself.

AI assessment note: “Yes, but I think most importantly, we don't have to, uh, to lead too hard”

Q Is that experimental? Is that, is that working?

A Well, we have, we actually, it's experimental and working. You know, it's not something we've rolled out widely. Um, but the, uh, the idea there is, um, say you, you had a page, um, because something broke on your application. And by the time you get on, on Slack, the body's there already and told, and he's telling you, um, This is what I looked into. I looked at, checked this piece of data, this piece of data. This is a, a notebook where I put my investigations. You can, you can follow what I did there. Uh, I think this is probably that, um, my recommendation is to, um, uh, restart the service. You want to click here and restart it.

AI assessment note: “it's experimental and working. You know, it's not something we've rolled out widely.”

Q But a vector database does not behave any differently than an all-app database, from your perspective?

A No, and also the AI applications, uh, typically the part that's a model on GPU is only a small part of that app, like the rest is, You know, it talks to a database, and it talks to a web server, and it talks to its source files, you know, all those, all that stuff. Um, so that part is, I would say, um, more of the same. More, I mean, it's not exactly the same. Like, there's many new technologies. We have many people working on things like GPU profiling that are very different from what you would do with a CPU, but I would say you can think of it as very similar, uh, just another iteration of what we've been doing the whole, the whole time. There's a second thing, which is a bit different, which is, uh, understanding the models themselves and how they behave. Um, and that one's quite nice, completely open-ended, uh, mostly because the models are changing very fast, but also the applications around those models are still fairly early in their life cycle. I think, you know, we, there's a few things, uh, that we understand well, so we understand what an image model looks like, you know, and we understand how to build a chatbot. Um, and those form factors are sort of there. But for the rest, we're still looking for the right form factors, and the models themselves, as I said, keep changing underneath that. So I think it's going to take maybe a few years for that category of, uh, obs…

AI assessment note: “No, and also the AI applications, uh, typically the part that's a model on GPU”

Q Okay. Interesting. And, uh, so how did you build your Salesforce to reflect that? Is it the combination of account executives and like, uh, SDRs, BDRs, like business development representatives?

A Yes, so it's fairly typical in that, so we started, uh, so when we started, the companies that were in the cloud were not the large enterprises, so there was not a ton of business to be had with the super large companies. Um, when we started, we sold to, you know, SMB mid-market, and for those, we had an inside Salesforce, uh, that started, you know, mostly inbound, and then that transitioned to being in, uh, majority outbound, you know, for what they were. Uh, doing. Um, that inside sales team is fairly typical. Um, it's, uh, you know, that's SDRs and, you know, and, uh, account reps. The account reps are doing the closing. You know, it's not, we don't have, um, an inside team that is generating leads for, for an enterprise team. That's not how we do it. Like, the inside team is closing their own deals. Um, and more recently, I would say about four years ago, Uh, as the cloud migration started to take foot in, uh, large enterprises, we started an enterprise Salesforce that's selling to companies over 5000 employees. Um, and there, that's a, uh, again, bit of a different model. Um, but the adoption, the underlying adoption is still the same. It's still bottom up, you know, so small, small company, large company, the product is adopted the same way. Um, and the users in the end are very similar, you know. So when you're a developer, uh, at a very large enterprise, You don't thin…

AI assessment note: “that's SDRs and, you know, and, uh, account reps.”

Q that's super interesting. Um, sort of almost like logistically, how long Before you jump into an industry, do you start planning that? Like, how does that start? You do, uh, you do a, um, like research efforts, uh, you, you have like an internal consulting team, or is that you as CEO who decides this? And then how long do you plan before you start building and, and then launching?

A Well, mostly we, we hear about it from customers. Um, so, I mean, we have this strong, uh, bias in the company that our window into the world is our customers. Um, so for example, you know, we, when we think about competitions, uh, or competition in general, we don't spend a lot of time reading our competitors' press releases or websites. Uh, instead, we talk to our customers and we hear what they have to say about it, whether it's registered with them or not, what seems to be valuable to them or not. Basically, the, the world around us exists with the prism of our customers. Um, so from that, we get a sense of what's actually a problem for them, what seems to be real. And then we decide what we, what we might try to go and work on. But the way we build it is, so we're not, we don't do it, uh, like Apple, like we don't, um, disappear in a basement for three years, uh, and then, you know, ship a fully formed product, uh, that takes the world by storm. Uh, instead, from the earliest days, like we work with design partners, we work with customers, and we try to ship something to them as quickly as possible. Um, so that's really the way we, we develop.

AI assessment note: “mostly we, we hear about it from customers.”

Q Who does this when you launch a new product? Do you have, uh, an internal, I don't know, Sherpa team? Like, I know you're active on the M&A front, so like I assume in some cases it's whoever you buy. Um, but do you have, do you take people from other products to reassign them? How does that work?

A Yes, and that's, uh, in part why it's difficult to, uh, build multiple products. Because what happens when you're, when you, you start thinking about product number two typically is you have product number one that is very successful. Um, but if it is very successful, chances are, like, everybody is super busy just keeping up. Um, and everybody that is super good and you would want to trust with starting a new product is, uh, you know, load-bearing on, on your core product. So that's difficult. Like, you need to pull people away, and that's painful. That's hard. The second part is that, uh, scaling a successful product and starting a new product Are very different motions. And, um, I would say, Um, people feel very differently about it. Like, in one situation, you mostly walk into situations where customers love you, um, and you know exactly what, incrementally what you need to do next, and you can work with them. Uh, in the other situation, like, you don't have traction yet, you're trying to understand why, um, and you have to understand, you have to, to read the, uh, or decipher the cryptic feedback you're getting from customers. Because again, these customers in general are, like, everybody are, People are good people. Like, they don't want to hurt your feelings. So when you talk to them, um, they'll say, oh, yeah, yeah, no, this product, uh, this product is great. It doesn'…

AI assessment note: “you need to pull people away, and that's painful.”

Q machines that will do things for us that we may or may not understand is, is a good thing. Do you think there is a world just to, um, play a little bit like AGI, Doomer, where, uh, AI actually becomes a problem for Datadog in, I don't know, automating or being able to do all the things in a way that where humans do not need to be involved?

A Oh, but I think you, at some point, someone needs to control the AI in some form, right? So, uh, or maybe not. Maybe, you know, we just, you know, we're all pets in the end, you know, but, uh, I don't subscribe to that, to that vision of the future, but, um, I, I do think that the, uh, at the end of the day, um, I, I do see that as one more evolution in the history of innovation, which is we just do more. And because we can do more, more easily, we do even more. Um, and then we need to manage it and understand it. I think that's the, uh, the overall arc of things, and that's, uh, where we have, uh, potentially an even bigger role, an even bigger role to play in the future as it happens. Um, and, you know, when you think of the impact on our, on our business, you know, there's a, there's a few ways to think about it. I mean, the, the most straightforward is, um, and the, the ones happening right now is, um, it just, uh, the emergence of AI, uh, just, uh, Pushes more digitization and move to the cloud, you know, because, I mean, to capitalize on AI, you need data. Um, so it needs to be digital. Um, and, um, you probably are not going to build your data center for AI yourself. Uh, if you are a handful, maybe 10, 20 companies, yes, you are, because, um, you are at such large scale, and you're in the business of providing your services for others, but otherwise, you're not, because …

AI assessment note: “we have, uh, potentially an even bigger role, an even bigger role to play”

Q And you offered for free initially for those, uh, design partners?

A Well, there's a whole process. Um, and actually, a key part of building a new product is understanding what's valuable and what's not. Uh, and it's difficult because when you, when you start, and your design partners are typically your existing customers because you have strong relationships with them. And typically, they love you, they love your product, you know, they are, so they're, they're happy to spend time on working on new things. Uh, but they haven't necessarily thought, uh, through the whole value of what it is they're asking you to build. Um, so the way we do it is we start by building with them. So we ask for what the problems are. We get a sense of what else they might use for it. What's their next best alternative or what they were using before. That's not good enough. So it helps ground everything. But then as soon as we have enough product, uh, we, we basically say, okay, it's going to cost you this much to use this product. And what typically happens at this stage is that when the product is ready, when it's great, about half of the design partners just disappear. Uh, you know, the people who are showing up in the meeting every, every week or twice a week and were very happy to work with us, they, you know, they, they start ghosting the product teams because they, they realize actually I'm not going, I'm not, I'm not able to pay for it, I don't want to pay for…

AI assessment note: “as soon as we have enough product, uh, we, we basically say, okay, it's going to cost you”