He raised $150M with $0 revenue. Then hit $1M ARR in 4 months. | Michel Tricot, Co-Founder of Airbyte

Michel Tricot (00:00:00) :
One of the traits that you should have when you have a level of PMF is, people are willing to go above and beyond to make your solution work for them. Because it's solving such a hard problem for them, and it's making their life so much easier. That they are willing to go through a little bit of a hurdle initially as you're starting because they know that the outcome is going to be insane for them. At the beginning of 2024, the product was very bare bones, and yet we were able to generate a pretty good amount of revenue in very little time. And the conversation is, I always go back to it, which is if you're solving a real pain, people will adopt your solution even if it's incomplete. And that to me was the signal.

Pablo Srugo (00:00:39) :
How fast did these products hit a million ARR?

Michel Tricot (00:00:42) :
Cloud, I think like within the first year it did. SME within the first four months of its life.

Previous Guests (00:00:51) :
That's product market fit. Product market fit. Product market fit. I called it the product market fit question. Product market fit. Product market fit. Product market fit. Product market fit. I mean, the name of the show is Product Market Fit.

Pablo Srugo (00:01:04) :
Do you think the product market fit show, has product market fit? Because if you do, then there's something you just have to do. You have to take out your phone. You have to leave the show five stars. It lets us reach more founders, and it lets us get better guests, thank you. Michel, welcome to the show.

Michel Tricot (00:01:19) :
Hey, thank you for having me, Pablo.

Pablo Srugo (00:01:21) :
This is interesting. I think a lot of it is because of the timeline. So you went through YC in the 2020 COVID batch. You raised $185 million since and maybe all of that was in that kind of 2020 to 2021 period. Where things were just way up and to the right. Some of the companies from there have failed, some have plateaued, you seem to still be growing but you're kind of coming out of that era and I'm sure there were a lot of difficulties that came as kind of a result of that. So we'll get to all of that because it was a very particular, very unique moment in time. In a sense, you could argue in AI it's similar, but I don't think today is the same as it was then and certainly there hadn't been a time like that. I mean, at that point, people would talk about comparing that to the dot-com era, right? So these are very unique moments in time. Maybe as a first question, because your product is not the simplest product, just tell us a little bit about what it is that you do.

Michel Tricot (00:02:15) :
Yeah, so basically, Airbyte is a data infrastructure product and the value we provide as a product is wherever you have data, you can bring it into a place where you will be able to act and make decisions out of it. So our mission as a company is to make data available and actionable. So a very simple example is for an analytics use case, you might have data about your customer across twenty different systems and you want to be able to monitor everything that your customers are doing, like the health of your customer, what type of touchpoint you've had with them. And in general, what you're going to do is you're going to be using a warehouse for that. So it can be Snowflake, BigQuery, Redshift, you name it and the data needs to be going from these operational systems like Zendesk, Salesforce, and others into these warehouses. And normally, your executives are looking at this dashboard on a daily basis. So you need that thing to be reliable and that is what Airbyte does, is we're basically the pipe between every single type of operational system into a warehouse for the case of analytics.

Pablo Srugo (00:03:20) :
Into say a Snowflake? So you're in between the two.

Michel Tricot (00:03:24) :
Exactly, and the very big specificity about Airbyte is that we're also an open source project. Because a lot of data projects always start very organically and open source is great at that point when people are tasked to build these types of pipelines. They're just going to go online, find a project, and do the movement of data. And also the number of data silos, and operational systems that exist is just infinite. It just grows every day and it's very hard for a single company to connect to all of those. And here we are working with our community to address all of those.

Pablo Srugo (00:03:59) :
And maybe just for a little bit more context, what are the alternatives? Who do you compete with? Is it some kind of closed source stuff or is there a bunch of open source? What is that world like?

Michel Tricot (00:04:07) :
The biggest competitor is internal. So it's basically teams building homegrown systems and having to maintain them and just allocating a lot of their teams to that. The other ones I would say are Fivetran, in a way Informatica as well. So every company that has this ELT, ETL type of motion with data. The thing is we're not on the transformation side though. We're really about the movement.

Pablo Srugo (00:04:32) :
Perfect. So now that we have, you know, that context. Maybe take us back to that time, like 2018, 2019. What were you doing? What leads you to start Airbyte?

Michel Tricot (00:04:41) :
Yeah, so I've always been in the data space since I finished college in 2007. I started in the medical space, then I went into the financial space, then the ad tech space. That was actually a very big moment in my professional story, being in ad tech between 2011 and 2017. That was just the explosion of data online and I was very fortunate to be there. And then what I realized, I started to work in another startup doing a lot of the mapping technology for autonomous vehicles. And I realized it's always the same problem, you have data feeds and you need to integrate them into a system. And at the end of 2019, that's really when we came up with an initial idea coming into YC about how do we help. Can we solve the problem of just accessing data and doing it reliably instead of always having to reinvent a new technology, always having to just build custom systems and that was really how we got into YC. We pivoted a few times, but we can talk about that. That was a fun moment.

Pablo Srugo (00:05:43) :
And this might be an unfair question because sometimes these are unanswerable, but why do you think it wasn't built yet? I'm just thinking out loud here. Obviously the data has been in many places for a while. There has been that SaaS app revolution, which would have been around 2019. By that point, companies had so many SaaS apps and all those sort of things but that was a trend that's been going on, right? And yet, it seems like startups or enterprises would continuously just build their own pipes. And yet the warehouses, obviously that's been news that has grown well before, let's say 2019. So what do you think was a missing link where there was no real product yet?

Michel Tricot (00:06:18) :
Yeah, doing data analytics at scale has always been very expensive. Before, you had to use Hadoop, you had to use older technology that was really just for enterprises and having access to real data systems if you're an SMB, or if you're mid-market, or if you're an early enterprise, et cetera, et cetera. This thing was a massive budget and I would say the first warehouse that really came and made data affordable for basically every single company in the world was Redshift. And that was in the 2000's, '13, '14. Then you had Snowflake, and Snowflake really boomed around, 2019, 2020, maybe 2018, but it was more that the technology of processing data at scale was not affordable for a very, very long time to basically every company in the world. And the moment you start offering something that is on demand, consumption based for data. You're opening up a new type of usage for data and it comes from the new families of connectors or systems you need to connect to. Because maybe if you're just building for enterprise, the only things you need are SAP, MS SQL, and other heavy systems, but maybe you don't care about a random SaaS product. But by opening this technology to new categories of companies, suddenly the breadth of connectors that you need to address just blows up and that, to me, was the problem that was unsolved. The world was waking up to, now I can actually do something with data, and it's not going to cost me my legs. New use cases, new audience, and new needs.

Pablo Srugo (00:08:02) :
So it was a combination of one trend is just the number of SaaS apps you might want to pull data from and then the other piece is these warehouses that let you, once you piped all the data in. Do you, like, interesting analytics with it. What was missing was the pipes. At that point it was just, you know, custom ad hoc.

Michel Tricot (00:08:16) :
Yeah, but if you go back before the cloud era, before AWS came out. The amount of investment that you had to make to start hosting servers was massive. Then you have cloud, and suddenly any company, any individual, can start putting up a website and a product online with very limited investment. And that's exactly the same thing that happened with data. It started with just servers, and mid-2010s, like 2015, 2016, we started to see, oh, actually now, let's also address the data use case. And so that became something like the level of investment plummeted, and it opened up a much bigger, and much larger pool of customers and use cases.

Pablo Srugo (00:08:56) :
And who are? You mentioned we went to YC, so I assume there's co-founders. Who are the co-founders?

Michel Tricot (00:09:01) :
The co-founder is Jean. We've known each other since 2012 and funny enough, along the years we had a lot of side projects together where we started to build a few things. But both of us are startup people and founder type people and basically, it was having two jobs at the same time. And we learned how to work together, but we never got to be a hundred percent on one company until 2020 or 2019, I would say, like stars finally aligned for us.

Pablo Srugo (00:09:31) :
And when you get to YC, what is it just pure idea stage or do you have something built? What stage are you at?

Michel Tricot (00:09:37) :
Yeah, I would say getting into YC, we were in an idea phase. Which is we knew what problem space we wanted to address, but we did not really have a solution yet. So it was very clear, we want to make access to data as easy and as reliable as possible. But how do we do it? That was the question and yeah, the first month, we actually. Jean and I were pretty good at just getting to talk to people, reaching out to people and that's always been part of the Airbyte story. And we spent the first month just talking to as many potential users or even batch people, batch members about what kind of data issues they were facing, trying to understand what could be a good solution. And funny enough, at the end of January, we started on a marketing type of use case, right? Marketing was very, very hot, completely transformed by data and that was the initial idea we started with. Providing more relevant data to marketing teams and then COVID happened.

Pablo Srugo (00:10:46) :
So does that mean, when you say you focus on marketing. That means there's a select number of kind of apps you need to integrate with, like they kind of rise to the top?

Michel Tricot (00:10:53) :
Yeah, and it was also the type of use case. Which is, it was more the value we were providing. We were also trying to build more of a layer on attribution, et cetera, et cetera, and just really get data across all the different acquisition channels that you might have. And also giving you visibility on top of every single person that is using the product, and integrating that into the marketing stack. So really funneling that data over there.

Pablo Srugo (00:11:18) :
And how does this compare? This is maybe a stupid question but, what are the sort of things that this enables you to do that an Amplitude or a Mixpanel, or whatever, or even just, you know. Stuff that you're, well, I guess, Cablo, you're probably piping the stuff to create great charts. But, yeah, maybe walk me through the other ways people could access this sort of data and what this enables that those don't.

Michel Tricot (00:11:38) :
Yeah, I would say like you look at Amplitude and Mixpanel, they were very focused on product usage. So to us, it's just one source of data, but what you care about also is Google Ads, it's Facebook Ads and today that's going to be TikTok or Snapchat. So those are different types of data sources that you might want to target. So for us, it was getting this data from Amplitude and Mixpanel of the world, plus also linking it to all the ad platforms.

Pablo Srugo (00:12:07) :
Got it. OK, cool and so once you pick that marketing space. What's the first, you go on and build something? What is the first version of an MVP look like or something like this?

Michel Tricot (00:12:15) :
Yeah, we built something. It was a very simple pixel that anyone could just add to their website and behind the scene, we were working with a few design partners. We really had no real issue getting these early design partners because, well, those were mostly startups. But there was a gap in terms of all these companies that are starting a business, they need to have that VDP to understand, are they investing their very, very limited resources in the right place? So we had a few design partners, and initially, honestly, it was you install a pixel or a tag on your website and we give you access. We basically push the data into your system. That was as simple as that. Now, we were getting some interesting, let's call it, top of the funnel metrics. But a lot of activity on the website, a lot of people looking at the different pages, et cetera, et cetera. But every founder wants to have an explosion when they release. But what really happened is that we launched the product maybe a month before COVID hit and so during this first month, we were able to get these three or four design partners. But the moment COVID hit, every single marketing team across the US, they just got no budget anymore, like frozen budget, and they were basically all laid off. Companies had no idea what was going to happen with COVID. So they went from growth at all costs to let's work on our fundamentals and that affected marketing teams quite a bit. And yes, we are still getting some interesting top of the funnel metrics, but at some point I just had a walk with my co-founder and said, I don't care about this type of funnel metric. What matters is just, are people paying us at the end of the day? And are we actually building a product that is so vital for them that even when COVID hit, our product is still necessary? And that was not the case. So that was a moment of truth for us. So I'm like, OK, first real pivot.

Pablo Srugo (00:14:16) :
And what do you pivot to?

Michel Tricot (00:14:18) :
So we had about two, three months of source searching. At the time I was working with, we also had three exceptional engineers that I had brought in from my previous company and man, we're just remembering what the world was in this first half of 2020. It's just no one knows what the world is going to become. What you think is going to be two weeks at home became, I don't know, how many years at home and we just went back to the drawing board, and we started to really involve the team as well on, OK, what do we do next? What other options should we be exploring?

Pablo Srugo (00:14:53) :
So there were five of you this time?

Michel Tricot (00:14:55) :
It was five of us.

Pablo Srugo (00:14:56) :
And were you funded already? Had you raised a seed or you were still in YC?

Michel Tricot (00:15:00) :
No, we had raised. Let's call it a mini seed.

Pablo Srugo (00:15:03) :
OK, a couple million?

Michel Tricot (00:15:05) :
No, less than that. Raising at the beginning of COVID was really hard. Because all the VCs were mostly focused on just there is massive fire in their portfolio. I don't have time to invest.

Pablo Srugo (00:15:15) :
The upside is that phase didn't last long. But yeah, there was one month of just, I don't know what's happening and then at least two or three more months of just portfolio intensity. And then, you know, it just rates dropped and Zerp, and this that, and things took off. But you didn't know at the time that it was going to happen.

Michel Tricot (00:15:31) :
No, we didn't know. Actually, we said no to an offer before COVID hit and post batch, I was just, man, what a bad decision that was to say no to that term sheet. 

Pablo Srugo (00:15:41) :
Right.

Michel Tricot (00:15:42) :
But at the end of the day, I don't know, we probably would have made bad decisions if we had closed this one, and maybe we would have continued on that terrible idea that we had. So we knew we had a lot of ability to just continue to explore. At the end of the day, when you're building a startup, you don't know the truth, you need to discover the truth, and for that, you need to have money to discover the truth. Basically, it's research and yeah, we explored a few ideas. One of the engineers was a massive fan of podcasts. He said, oh, we should do a podcast app.

Pablo Srugo (00:16:12) :
You were in a completely blank slate.

Michel Tricot (00:16:13) :
Yeah, but there was always. I mean, everyone in the team was an expert in data. But we were just, OK, maybe the world is changing, something has to change. So there was a lot of fintech. I love fintech but in the end, we're always going back to, man, we know there is a problem with data. We just need to figure out how to address it and after a lot of Zoom whiteboarding sessions, customer interviews. Every time we had an idea, what we did is just put up a campaign talking to people on LinkedIn. We explored tokenization for private data in healthcare, like that's data. We know, how do you make sure that you can share the data and still make it somehow anonymized so that people can use it and we're always trying to find people in the appropriate audience. And just asking them, what problem do you have with data? What would be your magic wand solution? Oh, and by the way, we've been thinking about this idea. What do you think? Where does it fall through?

Pablo Srugo (00:17:11) :
And what? I'm actually curious on those conversations, what were you looking for? What was the signal that no, this is not good enough or versus yeah, this is the thing.

Michel Tricot (00:17:19) :
Yeah, I feel like when you do this type of discovery, it's very easy to be stuck in very tactical problems that people are facing. Which are going to say, oh, yeah, I have a problem with this particular piece of the system. But what you want is to get to a stage where people are actually describing to you the magical solution and I call it the magic wand solution. And that's something we continue to use even internally. If you had all the power in the world and you could make a wish, what would it be? And in general, that's always what guided us in terms of was it a good idea? Was it not a good idea? I think for me it was, was the person in front of me able to rephrase what I said with their own words. Was it about, did I get a feeling that, it's a lot also about feeling and interpretation, but that is so important that people just want to continue to talk to you. Just to understand, is this solution going to just solve my deep problem? And very often they will say, yeah, guys, you're just completely wrong, no one cares about that. There are already massive players and they do an amazing job. And so at that moment, OK, fine, let's go to the next idea. Yeah, I think at that time we did about, what? Two hundred calls across all these different ideas. That was always something we would do quite a bit. The thing we also discovered, we went back to a marketing use case for some reason, and that's actually how I found my current VP of marketing. I met with him during that time. Talking is not enough. At some point you need to show, and that's when. Because I can be talking about A, and you understand A prime. The thing that really works is when you can get people to project into how the product works and how to use it. So you don't have to go and code everything. Mockups generally work, but even for an infrastructure product where UI is not that important. Having a UI, being able to show the different concepts on a screen and experience on the screen. This is what gets you the best feedback and then it's also much easier to iterate.

Pablo Srugo (00:19:18) :
And so out of all these ideas, what's the one that you picked to do Markoff's for?

Michel Tricot (00:19:23) :
Yeah, so that was. Yet again, a marketing attribution product.

Pablo Srugo (00:19:28) :
OK, back to square one.

Michel Tricot (00:19:31) :
Because we said there's still something, it's a real problem, anyone who can solve attribution. Two problems in the world, is attribution and recruiting. If you can solve one of these two problems, you're the king of the world. Knowing how to best spend your money is the best thing, and knowing how to recruit the best people, that's the most important thing. So we went back and tried to have a different approach. But yes, that was still, like, we killed it after basically three more weeks of investigation. And at some point we said, OK, the data integration space, it keeps going back to it. When we talk to marketing, when we talk to data teams, when we talk to healthcare, it's always coming back to, man, I'm really struggling to get the data out of these systems. I really struggle and that's when we, I think it was end of June, we said, OK, we're going to do a sprint. So we're going to give an idea a month and we're going to just go as fast, and as hard as possible on this idea. And we're going to try again next month if this one doesn't work, but we will also keep ourselves more active in building something. Because talking is fine, but at some point we're engineers, we want to be building stuff and we built L by Two open source. And that's really how we started. We said, yes, let's figure out a way to just move data between systems in a very non opinionated way. So not targeted on a specific vertical. It's not specific to marketing, not specific to sales, not specific to product.

Pablo Srugo (00:20:52) :
I'm going to ask you for a small favor, tiny little favor. In fact, it's not even now that I think about it. It's not even really a favor for me. I'm actually trying to help you do a favor for you. Just hit the follow button. You won't miss out on the next episode. You'll see everything that we release. If you don't want to listen to an episode, you just skip it, but at least you don't miss out. When you had that marketing product, this was a necessary piece of delivering the attribution, right? And what you realized was, actually all the other stuff is not actually as important as just this pipe that then people can use for whatever. They want to use it for marketing, fine, but they can use it for mirroring different other data types.

Michel Tricot (00:21:29) :
Yeah, what we realized with attribution is there are a lot of business use cases that are so specific to how you run your business, to the intricacies of how you think about doing marketing, et cetera. Building a solution that works for everyone just is not enough and a lot of the time, you realize that what people do is they fall back to just, give me the data into my warehouse and I will figure out what to do with it.

Pablo Srugo (00:21:51) :
Yes, right.

Michel Tricot (00:21:52) :
And at that moment you say, OK, fine. That's great because now we can really focus on more of an infrastructure play. Which is you want data in a warehouse, your own spaces where you have data. Let us build the pipes and that's really how we started. And at the time, we were also talking a lot with all the existing players, customers. So we did about seventy calls, I think, in the first month and I think it increased to about one hundred fifty over the next few weeks. But we talked to the users of Fivetran, Stitch, all the other modern ETL and ELT platforms. And that was, we always had the same feedback, which is we're paying for that. We're also rebuilding exactly the same system on the side because one, they don't have all the connectors that we need, or two, they have the connector but it doesn't work the way we want. And so we said, OK, fine, let's give you the simplicity of never having to build a platform. And, let's make the creation and the maintenance of these little pieces of connectors as easy as possible. And everyone we talked to, the moment we had the repo, we just invited them on the repo, said, look at what we're doing, what you think, et cetera, et cetera.

Pablo Srugo (00:23:00) :
Is that why you went open source, by the way? To let others kind of build on top of it?

Michel Tricot (00:23:03) :
Yes, that's a good point. That's actually something that took us a little bit of time to understand, is when people think about their data strategy. They always think about, what dashboard do I want? What warehouse will I be using? And they always leave the piece around, and they all generally have one or two use cases of what data they want to bring into warehouses. And in a way, it's an afterthought. How do I get the data actually in the warehouse? It's always an afterthought, and it means that this system always grows organically, and that's why I was saying our biggest competitor is internal competition. When people are building these systems themselves and at that point, you say, OK, the psychology of people who are building this system is that no one wants to be building pipes. We do, we love it, but no one wants to do it. It's a burden and then we said, OK, look, what do people do when they need to build a system? Well, they go on Google, and they say, oh, how do I get my Salesforce data into Snowflake? Because they want to find a script, or they want to find something that is, I would say, cheap to start with to deliver value as fast as possible and at that point. So it's basically to not do the buy, to go for the build. Open source is very good when it comes to the build versus buy because, in a way, it removes the build, but it also removes the buy. So as an engineer, you just take open source, you don't have to build, you don't have to buy, but you get the value out of it and for us, that was a great way for people to just get the interest of people. And to be added to their systems the moment the project starts without having to talk to anyone.

Pablo Srugo (00:24:36) :
When do you launch this version of Airbyte?

Michel Tricot (00:24:39) :
November 2020, and initially, we're very focused on vanity metrics on GitHub. Those, especially for open source, are very important because these vanity metrics, I don't care about how many stars I have, but I know that whenever I look at a project. Whenever I look at the library or anything like that. I look at these numbers, not because I want to be the popular kid in the blog, but because I want to see that this is a real project. So how many contributors there are, how many stars, give me some confidence. If there is one star, yeah, whatever, it's just some random guy that is building that thing. It's never going to be maintained. So we're very focused on the vanity metric to make sure that basically starts the snowball effect of trust.

Pablo Srugo (00:25:20) :
How do you, you know, let me ask you a question on that. The go to market, the zero to one go to market for something like this. I think about it today in AI, there's a lot of, maybe OpenRaver is an example, but there's a lot of building blocks. We've got a company called Backport, which is on the memory side and I always wonder, you know, on the one hand, I'm like, yeah, this makes sense. Take your pipeline, right? I'm like, yes, this makes sense. Just about every single company at some point is going to want to take data from disparate places, stick it in a Snowflake or something like it and, why build something from scratch when you can effectively get it for free and bring from there, right? Especially with open source. OK, fine but how do you go and get these people? You can't run a normal outbound, the marketing, it doesn't click to me. You know what I mean? How do you actually get one of these things spinning?

Michel Tricot (00:25:59) :
Yeah, so you need to understand where you're. So when you do something like that, you're not targeting execs, or you're not targeting directors. You're really targeting the people that are on the field having to do that, individual contributors and so what you need to do is to understand what community they are on. In the case of data, there is a lot of very big community on Reddit, for example. Hacker News, same thing. Who is reading Hacker News? It is people that are building and that are interested in new technology. There is a lot of other data communities. So after that, it's just doing some demos to a community. In general, this community, to be lively, they do, even during COVID, they would do invites one guest a month and here you have an opportunity to make a demo, et cetera, et cetera. And then it's also about, you go on LinkedIn and you search for everyone that has data engineering, analytics engineering, all these titles that do matter, and you try to engage with them. And so that's really how we bootstrap it.

Pablo Srugo (00:27:01) :
And is this stuff, the community-building stuff? So the direct makes sense, that that's something you could do every day, right? But, let's say posting on Hacker News, posting on Reddit, how much can you do? I mean, you certainly can post the launch, you can probably find a few forums where it's relevant, you post once, you know, but you do that and in a week, are you kind of done? Or is it something you can somehow keep pushing on?

Michel Tricot (00:27:18) :
Yeah, so what you want to do is, I agree with, that's a very good point. If you just post and it dies, and you have no way of just reaching back out. Besides just going on the same community to talk to the people, then in a way, you're wasting some effort. So very quickly, we actually set up a Slack community for ourselves. So everyone was just going on to Airbyte. I don't know how many we were converting into community members, but we had a good chunk of people that were just joining our Slack. So it means that suddenly now we have direct access to them and they've also demonstrated some interest in what we're building. Even though we were very early and they became the first users of Airbyte and so for me, it's just. You need to have a way where a recipient, where everyone who is coming. Even if you don't have a cloud product where you can get people's emails or things like that. A place where you can just grow and get this user at home.

Pablo Srugo (00:28:09) :
Part of the challenge too is you really got to catch them. If you don't have a community like this, to actually get them to use Airbyte right now. I mean, you got to catch them at that perfect moment where it's not too late. They've already built it, it's not too early, they don't need it yet, you know, and it's just not like they don't have an existing product where you can just be like, who are they using? I'll just get them to churn. So there isn't a very clear, so you need to have this place that you can put them, and they can know that you exist and somehow engage. And when they're really ready to build, OK, then you're there.

Michel Tricot (00:28:37) :
That's correct and at the end of the day, it's such a painful problem that people have. That they will look at every single solution that exists because this is really, really hard and that, to us, was also a big thing. When we first raised the product, it was very immature. There were a lot of connectors that were half working, et cetera, et cetera, and yet people were super, super engaged. They were fixing issues with us, we were fixing issues with them and to me. That's one of the traits that you should have when you have a level of PMF is, people are willing to go above and beyond to make your solution work for the. Because it's solving such a hard problem for them, and it's making their life so much easier that they are willing to go through a little bit of hurdle initially as you're starting. Because they know that the outcome is going to be insane for them and that, to me, was the first. That was really the big signal.

Pablo Srugo (00:29:28) :
2021 is an insane year. I mean, I'm just looking at Crunchbase here is, $5 million seed from Excel, $26 million Series A, a few months later from Benchmark. Which is epic in its own right and then $150 million at the end of the year from Coatue and Altimeter Capital. So that was a crazy banner year. My question is, was it in terms of growth, in terms of usage. Was there some crazy inflection point? Is there something that happened, or was it just kind of this consistent growth, you know, throughout the year?

Michel Tricot (00:29:55) :
No, no, we actually hit a good problem to have but we hit a massive wall over the summer of 2021. Where the team was very small. We had, what? Maybe ten people in the team and suddenly we were unable to build any type of product. The community just blew up overnight and I mean, overnight. No, it took a few weeks but you get the idea.

Pablo Srugo (00:30:17) :
It just got really large, you're saying. The number of people that join your community.

Michel Tricot (00:30:20) :
It just got super large, and everyone was using it. Everyone was asking us questions. There was a lot of feature asks. There was a lot of contribution on the connector side. So we had to review all these PRs. We had to review all these fixes and we just couldn't continue to work on the product anymore.

Pablo Srugo (00:30:37) :
How did you get to that problem? Because every founder wants to have that problem. Was any of these things that you mentioned you did, did one of them work exceptionally well? Probably there's a lot of word of mouth. How did that? I mean, just maybe go deeper on how that happened.

Michel Tricot (00:30:48) :
Yes, if I were trying to reverse engineer why it happened. I think I can put three things that we've done. So the first one is, we did a post on Hacker News that worked really, really well. We published the slides that we use for our fundraising, and we pushed it everywhere. I think at the end of the day, people also want to feel, especially with an open source company. They want to feel a connection with what the company is doing, et cetera, and I think giving that window into what Airbyte was doing, and how we did the fundraise. We're just a young company of barely a year old. I think it created a lot of interest from the community, which is, oh yeah, that's a very cool project. Let me see if I can be a part of it. The second one is when we actually created a golden path for contribution. Because before that, you know, you don't want to invite contribution too soon in a project. Because that's going to disrupt building the foundation of the project. But I think it was in April 2021, we released what we call the first version of the CDK. Which is a very simple way to build a connector and contribution just blew up at that time. And the last one, raising the Series A gave us a lot of credibility as well. Suddenly, it's just, yeah, OK. I know I can bet on this product because this team is funded, and they're going to be here for longer than just a year or a year and a half. So to me, those were the three main events.

Pablo Srugo (00:32:14) :
Tell me a little bit about monetization. Through 2021, you know, usage is obviously growing crazy as your community is growing. Is it all free or do you start to charge something somewhere for it?

Michel Tricot (00:32:24) :
No, we started to charge but it was more early support packages for specific users. So we're not trying to monetize at the time, but we knew that there were people that were building critical pipelines on us, and having a financial relationship with them was just making them feel better about adopting Airbyte. So we did that, and we were just supporting them. We didn't have that many at the time, but the community, though, was growing and the usage was just growing massively. We were just adding thousands of new users of Airbyte every single week, every single month and so that led to the Series B. And over 2022, we started to really have that new internal effort at Airbyte on, OK, we have great brand awareness. We have great usage of Airbyte. Let's start thinking about the next steps for monetization. Because project market fit and product market fit are two different things and that's when we started to build the cloud product that we released at the beginning of 2023. So basically, it was around mid-2022 that we started to work on the cloud product, and we released cloud in 2023.

Pablo Srugo (00:33:33) :
And the cloud product is what? The non-open source version of Airbyte?

Michel Tricot (00:33:36) :
Yeah, it's like basically we fully host Airbyte for you, we manage different users, et cetera, et cetera. So we've always had that philosophy with open source that everything that is core to just moving data from point A to point B should always be open source. Everything that is above is going to be how we monetize. Because at the end of the day, we want to make sure that the project is successful and for that, we also need to get money.

Pablo Srugo (00:34:00) :
And I assume also there's, I mean, there's a huge change in. Let's say investor priorities between 2021, which was just pure growth, and I'm sure the usage was huge. And then, late 2022, and 2023, it became a lot more about revenue and profitability. Like, revenue actually, that good unit economics and all this stuff. So how does that monetization, cause you're right. Product market fit, you had product market fit with the original product clearly, but, you know, not in a way that generated revenue. So now you're trying to generate revenue. How does this cloud offering launch go?

Michel Tricot (00:34:32) :
Yeah, first of all, we're always very clear that even when we raised the seed, we were going to monetize Airbyte. We want to build a successful company, and we know we need to get revenue for that. So the launch in 2023, went well but we didn't see the same explosion we had with open source and, you know, we reached out to a lot of people from open source. And asked them, but why are you not using cloud? And some people said, oh, because we already have Airbyte open source installed internally. We don't want to do any kind of migration. OK, fair but to be clear, it was growing. We really started the revenue generation engine on that front, but we could not understand why some of the big users, big logos, would not go on cloud and at some point, they told us, I don't want you to see my data. And that's why I picked open source, because now my pipes, I have full control over them. And for us, that just clicked in our head, which is, yes, the value of open source isn't just in the fact that it is free and that you have a community, et cetera. It's just that it gives you full control over one of your most precious assets and that's actually when we started. I mean, we knew that was important, that was on the roadmap, but we didn't think it would come that early and yeah, at the end of 2023, we decided to just expand the product offering with what we used to call self-managed enterprise at the time. Which is a product that is fully self-managed with all the bells and whistles around what we had on cloud, but that you can deploy on-prem.

Pablo Srugo (00:36:08) :
And is that the one where you found kind of real product market fit?

Michel Tricot (00:36:12) :
Yeah, it went really, really fast and to be clear. Cloud also blew up.

Pablo Srugo (00:36:16) :
How fast did these products hit a million ARR?

Michel Tricot (00:36:19) :
Cloud, I think, within the first year it did. SME within the first four months of its life. So, yeah, and they just had instant growth. Now, the good, the interesting thing is we actually killed SME recently. Because we have a better product that is much simpler for people to use, and that still gives them the sovereignty or flex. And that, to me, was, in a way, the initial vision we had for what is an amazing data movement product, which is all your pipes. You have full control over them, but you have so little to worry about when it comes to maintaining connectors, monitoring connectors, et cetera, et cetera. You don't have to do a lot of that work.

Pablo Srugo (00:36:57) :
Perfect, let me stop there. I'm going to ask you a question that is. I'm sure people are thinking, I'm really curious on your answer. I mean, a lot of the narrative these days, maybe I would say anything between now and even going back all the way to 2023, was what's going to happen to these paper unicorns that raised a bunch of money in 2021, and are potentially overvalued. Even Brex just sold for $5 billion, which is a sick outcome and there's a lot of people saying, oh, you know, they were worth $12 billion, all this stuff. Which is nonsense but I'm curious, you've lived that in a sense. Because you raised $185 million in 2021, you were valued well over a billion dollars and I'm sure for a while, when all the multiples compressed. You were effectively trying to get yourself into that valuation. Maybe just tell us your thoughts on it as somebody who's lived it and living it. Let's say, in terms of the, on the one hand, the downsides of having raised that. Do you regret it? Is it like, it is what it is? It's not even that big of a deal. I mean, where do you stand on all that?

Michel Tricot (00:37:53) :
Yeah, on that side, I'm very pragmatic. Which is, yes, it's great. You have a big valuation, but I never celebrated having a big valuation. I never called myself a unicorn company because I do not care about that. What I know, though, is the problem we're solving is real, and it's just going to continue to grow. And I didn't know that AI was coming, but AI made it even bigger of a problem. So at that point, the only thing that matters to me is, am I building a product that really solves someone's problem? Is it a problem that has a massive scale? And if you think about it, every single company in the world needs to do something with data. It doesn't matter what industry they're in, they all have the same problem and the moment you're building an infrastructure product, that you have a great community and an open source play. And that you're almost seen as a standard in the industry. It will take the time it needs, but we'll get there and we have enough of the capital, and we're spending efficiently. So..

Pablo Srugo (00:38:53) :
That was what I was going to ask. Pragmatically, how did you handle it? Did you just minimize, you know, even though you have $180K, you just kept your spend very, you know, so you could. You could use that for a while and need to go out, and raise, and figure out valuation, and all this stuff.

Michel Tricot (00:39:04) :
Correct, correct and also, the investors we're working with are also very familiar with open source. That was, I think, a luxury that we had. Which is we had a lot of interest in Airbyte in 2021, and we were able to really work closely with specific partners. And make sure also that they understood what it means to build open source. And whether it's Accel, whether it's Thrive, whether it's Benchmark, they have all invested in very successful companies, or Coatue. They have all invested in very successful companies, and even, yeah, Altimeter as well is invested in DBT. They've all seen what it means to build an open source business. So they know that there is a bit of a lag when you have revenue, but then you have the hockey stick.

Pablo Srugo (00:39:48) :
And then out of all the kind of products you launched, and inflection moments you had. When would you say was the moment that you felt you found true product market fit?

Michel Tricot (00:39:55) :
I think it's when we launched SME. Because this one grew, at a time to be ten at the beginning of 2024. the product was very bare bones and yet we were able to generate a pretty good amount of revenue in very little time. And the conversation is, I always go back to it, which is, if you're solving a real pain, people will adopt your solution even if it's incomplete. And that, to me, was the signal.

Pablo Srugo (00:40:19) :
And was there ever a time in your five year journey where you doubted whether things would work out or maybe even thought that things might fail?

Michel Tricot (00:40:26) :
I mean, who doesn't? At that point, it becomes the thing of every founder. All the highs are very high, all the lows are very low. But what matters is the overall trend. But yes, of course, we always had that. Even, you know, back in April 2020. I was in my bed unable to sleep, and I was just, oh my God, Michel. You're starting your startup, and you have the worst possible thing happening to you that the world has never seen. Which is a global pandemic, man, what are you doing? The world is falling apart. So it happens all the time but then internally, we joke about it. All these types of things, we call them tier two fun. Which is when you live through them, it's not fun. But when you look back and say, OK, yeah, that was a bleep.

Pablo Srugo (00:41:11) :
That's right, that's good, I like that and then last question. What would be some of your, maybe top advice for early stage founders that are in that kind of zero to one phase?

Michel Tricot (00:41:20) :
Yeah, the first one is just, don't assume you know the truth because you don't. You might have a sense of what problem you're trying to solve, but you don't know what the solution is going to be and so it's good to have hypotheses but you need to really listen to what your early users are telling you. You need to discover the truth. You don't know the truth. The other one is, if you're doing it, you need to do it one hundred percent. I mean, let's say one hundred fifty percent. You can, you know, I've talked to a lot of people and I've gone through it myself. Which is, I was trying to have my day job plus building startup ideas, but it doesn't work. You need full, at least your full attention. If you need to call people to talk about your product, you're not going to do that during the day. So suddenly, you have to do it during the night. It just makes no sense and that's the most important thing you need to do at the beginning. And I would say the third one is, don't look for perfection. Just lower your ego and be ready to take some harsh feedback. Just be looking for those and get your product out as fast as you can. Because you don't know the truth. So just make an assumption, test it, and be humble enough to accept that this is not the right assumption.

Pablo Srugo (00:42:30) :
Perfect, well, Michel. Thanks for jumping on the show, man. It's been great having you.

Michel Tricot (00:42:33) :
Yeah, thank you Pablo, for having me.

Pablo Srugo (00:42:35) :
Wow, what an episode. You're probably in awe. You're in absolute shock. You're like, that helped me so much. So guess what? Now it's your turn to help someone else. Share the episode in the WhatsApp group you have with founders. Share it on that Slack channel. Send it to your founder friends and help them out. Trust me, they will love you for it.