Airtable's Howie Liu spent 2.5 years building before launch because he saw it as a platform, not just a spreadsheet tool. He validated with only a dozen diverse customers, arguing a small set of representative users captures the essential market dimensions. Liu warns founders need genuine conviction about the long-term opportunity before committing to a deep, slow build.
Chapters
And I think what's been consistent but not necessarily understood is that we've always wanted to be more of an app platform than a collaboration tool. So when we first launched, we have to— had to obviously pick an MVP that was, you know, simple enough to build quickly, although quickly for us was like 2 and a half years.
You spent 2 and a half years before you shipped a product?
Correct. Yeah, actually weirdly parallel timeline to some other great PLG companies like Figma and others, which we can come back to and like why I think that was a really interesting moment for creating this new grade of like desktop-rich, uh, you know, kind of, uh, experiences that were web-based, collaborative, instantly shareable, and kind of disrupted like, you know, certain work patterns. Um, but we spent 2 and a half years building this product, and the reason for that was this was not meant to just be like a simple project management tool or even like a spreadsheet replacement, but really like an app platform that was way easier to use. Now the V1 looked like a spreadsheet. That was partly on purpose because We wanted to make it really intuitive. You could just start using it, didn't have to learn a new UI, but then also partly because we didn't have the time to build out the interface layer or like additional views.
You didn't have the time after 2 and a half years?
I mean, hey, you know, we were working, we were working, you know, methodically.
And it makes sense. And obviously it worked out for you, built this incredible tech infrastructure. But just on this meta point, if a founder— because you do some angel investing— if a founder came to you and said, Howie, here's my pitch, and I'm going to spend the first 3 years building it and then I'll start shipping to the real world.
Well, don't lead with that part. You know, talk about like the big, big—
So what has to be true? What kinds of products lend themselves to that type of formation versus the lean startup iterative?
Yeah, I think, I mean, my belief now in hindsight, and I had a very short-lived startup before Airtable called eTax, ended up getting acquihired by Salesforce. So I only have a data point of 2, maybe like 1.2, but My belief is that either you have to have very strong conviction in the endgame and ideally even like the midgame of the company, meaning like if you're building SpaceX, you have like enormous confidence that there's like massive market opportunity at the end of the road, right? If you can really make a cost-effective, safe, like scalable rocket, there is a market for rocket payloads that are cheap, right? And so, you know, there's a big payoff if you can do this. You also know you have like durable advantages because other people are not going to go and just enter into this, uh, you know, the space really quickly. So, you know, the mid and end game are strong, but then the, the, you know, kind of beginning of the game is really hard. Um, I think Airtable looks kind of like that where we thought really methodically about the space. I mean, the premise was really, you know, from working at Salesforce, realizing that there's a large class of business applications that really need a platform, not a solution. to be served well, as in you need to customize the data model per customer. Every customer has a different workflow, and maybe it's just a long-tail use case that doesn't warrant a really good best-of-breed vertical solution like a Workday for HRIS, right, or Salesforce for CRM. And even Salesforce itself did a really good job of serving customers, not because they built all the features for CRM, but because they built a platform that could be customized for every customer install. So that was the insight. And I felt pretty convicted based on both working within Salesforce, doing some early customer validation, talking to customers, understanding what are the different workflows you're doing in spreadsheets or underserved by current products to kind of validate the market. Obviously, you can never be 100%. But I felt pretty confident if we could nail the execution of a platform that could serve all these use cases with a real kind of application or relational database model, that we would be differentiated, we'd be competitively advantaged. So it wouldn't be easy. I mean, the fact that it took us so long to build it meant that it wouldn't be something that other companies could just really quickly fast follow, right?
And by the way, how many companies did you talk to when you were doing— over the 2 and a half years, how many prospective customers did you talk to?
Probably like dozens, not hundreds.
Wow.
And that was— do you think that's a lot or—
It seems like a small number.
Yeah. Yeah. Well, I mean, I mean, maybe you're just—
What do people think?
It's like, you didn't build multiple views. You only talked to 12 customers.
Yeah, exactly.
Kind of the lazy startup approach.
Yeah, yeah.
No, I mean, on that point, I think of it as actually— so in math, there's this concept of like eigenvectors, which is where you have a really large multidimensional search space. But to reduce the dimensionality of the space, you can pick like vectors or dimensions that best represent that space, right? So like, you know, to make it very simple, like imagine a 3D space. You have a bunch of like, you know, kind of dots throughout it. Well, find like the, the 2D axis, the plane that best captures, you know, the, the, um, spatial coordinates of all, all the dots, right? Um, if they actually kind of naturally tend towards, you know, uh, a certain plane. And so I think the idea here is, you know, you don't have to go and collect every single dot, first of all, across every single dimension to understand like, you know, the diversity of customer use cases you want to solve for. And I think like if you're picking 12, you know, frankly, even like 3, and I'll come back to like how we apply this concept even today to product development. But I think if you're picking like 12 customers across a wide range of industries, company sizes, or whatever, like the space that you want to, you know, kind of fit well into, it's pretty unlikely that you're going to overfit like exactly to just those 12 and that every other company that looks like them is going to be completely different or like your solution is going to be inapplicable, right? There's always kind of this, you know, healthy balance between, yeah, you want to collect more data, but I don't personally think you have to go and like, you know, get 100 customers to validate the premise if you have 12 or even like 5, you know, pretty representative ones that are not all like weird outliers that are super early adoptive and like nobody else will ever look like them.
