$0 to 2 Unicorns: Bootstrap vs $250M VC - Jason Cohen [WP Engine]

But in AI, actually, that's not true anymore. It's flipped in AI. It's easier to get distribution than it is to build the product because AI products almost never work. If you make the wrong decision and get lucky, that's still wrong. If you make the right decision and get unlucky, you still made the right decisions because once again with AI, you can build something no one wants. I don't care how efficient it is that you built it. If you're picking between only two options in that sort of huge life-changing situation, then you haven't explored the options enough.

Thanks for having me.

Yeah, I mean, first of all, SmartBear became a unicorn maybe 10 years after that, not the following year. But to your point, they were built very differently in different industries, but that tells you there is not a blueprint. So, that's easy. We can dispense with that idea.

So WP Engine was my fourth company. SmartBear was my third company. The more successful they are, the more you want to talk about them, I suppose. But that's not how life really is. There's lots of companies and outcomes and everything. And so, at SmartBear, I had a nice exit. So, I wasn't forced to do anything in particular and not in a particular timeline, which is nice. And it just so happened with WP Engine, about 2 years in, I saw, 'Wow, this one could actually be really big.' Like it was clear the market was big and growing. The company, it was easy to sell things. A lot of people wanted it. We had a great niche. So, I just, in other words, everything was kind of lined up with like, 'Oh, this could be actually venture sized.' Most things aren't, and that's not saying anything bad. Most of my companies weren't, and that's not saying anything bad about them. Just most products aren't going to make billions in revenue or be worth tens of billions. Like just most aren't like that. It's okay.

I think we get wrapped up around the whole VC thing, but how many companies do VCs invest in every year? A few thousand? How many of those survive? A few hundred? Okay, but there's a million companies, a million small businesses just in America. In the world, much more. Okay, so VCs don't fund very much. So, you know, very few successful companies are funded by VCs, so let's not get carried away. Anyway, but WP Engine was. They had these characteristics of like, 'Oh, this could be really big.' And so, I could have a different path this time, a different journey, a different set of challenges, different goals, and so on. And a combination of the fact that it was my fourth go-around with a bootstrapped company, and that I was a father already, and I was older, so just for personal reasons of wanting something different, and having done other things for at that point still more than 10 years, right? I still had done many things. It felt like I wanted to do it differently this time, and this one could be done differently this time. That was the confluence that made me say, 'Yeah, I'll take money this time.' And of course, it wasn't $250 million at first. It was 1.2 because I had some angels. And then the second round was, get ready, two million, right? So, like we started modestly.

Yeah, well, I never wanted to raise more money than the amount of ARR we had. That way we're not, we have enough money to do something that's appropriate to our size, but not so much that you know, sometimes you drown in that. And also, I never raised at a crazy valuation. Again, you don't want a low valuation where you get diluted for nothing, but you don't want a high valuation that's too high that you have to grow into before additional growth kicks in and finally you can get there because then the next round might be hard. So, we were very practical in terms of raising the right amount of money at the right valuation to where it was always possible. Of course, we also grew really fast. But so good. So, I guess that makes all things good. But if we had raised at too high a valuation, it might have still been hard to raise again. I think a lot of companies are finding that now where they're very healthy, they're very good, growth is okay, and it's hard to raise money because, 'Oh, wait, you don't have AI or something.' And so, all of a sudden, they're not interested. It's very difficult if you haven't kept those metrics together.

Well, I don't think SaaS companies are going away. I've yet to find someone who was paying for a substantial SaaS product and then vibe coded such that their company is now not using SaaS product. I haven't yet to see one. Of course, the stock market valuations are down now. You could argue they were too high. You could argue now they're too low. But when you have companies like HubSpot with more than two billion in revenue still growing 30% year over year and profitable and their stock goes down, you say, 'Okay, it's either temporary or it was valued too high or something.' But it's not like HubSpot's going away. That's not the conclusion.

You could say, 'But AI is the next wave.' Obviously, it is. But I guess one of the problems with these prognostications like 'AI will kill everything' and so forth is the timeline of these prognostications matters. For example, in 2010, there were people saying that the stock market's going to crash. Well, it didn't. Now, eventually in 2020, '20, '21, it crashed. Okay, so 10 years later, it crashed. So, you know, you can say, 'Well, I was right. I told you it was going to crash.' Well, you know, that's terrible investment advice, right? Rather, you want someone who says, 'You know, it's going to crash in the next 3 to 6 months, so that I'm getting out now.' And then it does crash in say 4 months or something. Like you don't have to be perfect, but if you're not specific about your timeline, the prediction doesn't mean anything. You know, Elon's been saying we're going to have self-driving cars in about 2 years, and he's been saying that for like 15 years, right? And if anyone's going to have it first, it's Waymo and not Elon. So, it's the fact that you made the prediction is irrelevant if you don't have a timeline behind it that's meaningful. So, will AI take over everything? Yeah, probably. But is HubSpot going to go out of business in a year or 10 years or never or something? Like I don't know. Okay, well, then it's not really a meaningful statement. It's not helping you and me and anybody build a company now and make decisions.

So, I think what is useful is to say things like the customers have the same problems they always had, meaning marketers are trying to get leads and be seen in a sea of stuff that's difficult to get attention in. And influencers are trying to get people by being crazy and doing stuff and other sorts of stuff that's not like marketers at companies. And you know, engineers still need to make code and the finance team needs to do their annual audit and do something about that and have systems. Like all these issues, those are still issues. And so, what AI does is say, 'Oh, but we may have new solutions that could never exist before or it wasn't cheap enough before or whatever, not effective enough.' And so, we can solve these same problems we've always had, but in a different and let's suppose better way.

So, that's pretty interesting. So, what is true is if you're a traditional company, if whatever AI is going to do to the functionality in your area, that is true that I don't know the timeline, but the clock is ticking for which whatever you're doing now is now the old way. And that's okay, but that market will become smaller over time. You'll fight with other legacy products. The only way to get a new customer will be to steal one from someone else, and that someone else is probably pretty good. And if they are going to switch products, are they really going to go to another legacy one or whatever the new stuff is? So, in other words, while I don't think it will kill it super fast, I do think some of these pressures might appear relatively soon. And so, in any way, that's your future. So, especially on a podcast called Billions, if you're trying to build a big company, that takes a long time. I know even in the age of AI, we always see, 'Oh, but this one got there in a month.' Right, but almost all of them don't. So, even in AI. So, settle down. Right, it's going to take many years. And so, if it's a legacy non-AI solution, and it's going to be 5, 10 years before you get to the kind of size that you're aiming for, okay, well, and you don't have AI. Like it's possible, but it doesn't seem likely that you'll be relevant.

Same problems in new ways.

Right, our customers have WordPress sites. They need to build WordPress sites. They want to add content. They want to get found on SEO. They want to send people there and track them with metrics. They want to have a content calendar and write stuff for that. They want to be found globally, not just locally, some of them anyway. They want to convert those folks that come to whatever. If it's a media site, they want to have them click a lot so they get ad revenue. If they're e-commerce sites, they want them to buy something. If they're a company, they want them to at least understand the product if not buy. Like these are still what they want. So can we deliver something that helps those things happen more because AI could help sometimes. So for example, in e-commerce, a lot of e-commerce people don't even know this, but when someone searches on an e-commerce site, they're instantly three times more likely to buy. Now, that's obvious in retrospect because of course they're showing some kind of intent. What are they searching for? A particular product by name or at least a category that they're researching, so it just stands to reason they'd be a lot more serious than random riffraff that comes to the site. Okay, so no surprise. It's interesting that the number is 3x. Okay, you know, we have tens of thousands of e-commerce stores that do significant revenue. We have even more that don't. You know, so okay, we're not Shopify, but we still have a lot of data. So we know this. And okay, so the search experience, like what happens next is very important because this is a golden opportunity right here. This is what you want to have a great thing that they want to click and buy because like this is it. So we built an AI-based e-commerce search bar. So it could be semantic and understands your product. Like okay, you can imagine what you would want it to be. And even be able to say stuff like, you know, I have, I pronate my foot and I like a wide toe box and I'm looking for a shoe. Like right? Like whatever they put in, you want to satisfy that because here they are. So that's for example a product that we have. So that's an interesting layer where it makes the customer more money. It makes sense for us to do it because we have access to their database of products. We're also physically there when the person's clicking because we're running the website. So like we're there, you might say, or we have access at least to all these things already. Therefore, it makes sense for us to do it.

So that's the kind of way that we're thinking about how can we, you know, ultimately the best way for any company to think about it is how do you make your customers successful in the way that they measure it? Best is revenue. How can we make the e-commerce firm more revenue? That's the best answer. But okay, not everything is that. You know, if it's a finance tool, you could say, 'We want the next audit to have zero discoveries, right?' I don't know what that means exactly, but as a vision statement anyway, that's pretty good. And yes, that's exactly what they want. Yeah, right? They don't expect the finance tool to create revenues. That's okay. But like what is ultimately what would they see as undeniable huge success or value? You should be asking that. Now often you can't measure it. Often the measuring of that number is not in your purview as the vendor. But what you can do is say what it is anyway as your goal, as your vision statement anyway, because there's some things you could measure that are proxies of that, or at least just as you're saying like well, then what kinds of features, what kinds of functionality, what things would we want to tell a customer in a sales pitch, what we want to promise on our homepage, what would we want our existing customers to at least think. So what reports might we show them every week to show them something. In other words, like it imbues, this concept would imbue all the things that are in your control with a direction. So that's part of what a strategy is, right? So I do think it's useful to say what that thing is even objectively even if you can't measure it because it still can make everyone aligned around that. You can always, by the way, ask the customer. 'Hey, do you think this is making audits better? Do you think this is increasing revenue?' And whether or not they have data. You know, if they're saying yes, then maybe there's a scale and I don't know. But like if they're basically, if the trend is to the positive, you're onto something. You know, because okay, it's not hard data. I know that, but it's better than not asking and just saying, 'Well, they click a lot, so I guess they like it.' Yeah, that's not true. You know, sometimes when they spend a lot of time in the product, it's because the product's confusing and they can't figure out what to do. Sometimes when they don't spend time in the product, it's because it's just working. So you know, all these kinds of metrics like time spent in the product, they're okay, but you know, they're pretty far away from what you probably want to be optimizing for.

Well, I think at a certain scale, you can start thinking about how you allocate the capital of your company. You might say some of that is, well, it's all money, it's people, however you'd like to say it. There's a lot of sort of well-worn ways to do that. These are not that useful for a startup. But when you have, you know, we have more than a thousand people and we have the ability to try things, and so there's, you know, some people say that one of the roles of the CEO is to be a capital allocator. When you're starting up, you hear that and you're like, 'Well, I get it. Like I decide what to do and what to spend money on.' But you would never think of yourself that way. It's just not how you think about what to do. And you shouldn't, by the way. You shouldn't. But at a larger company, it does make sense to say, 'We have a budget. We have things we need to do. We have, you know, kind of running the business stuff and incremental stuff we need to do, but also if we never try anything new, that would be bad. And so there's some way to allocate the amount of attention, people, money to like you say, like big new, it still needs to be adjacent things. It doesn't make sense for us to try self-driving cars, right? Because there's no current expertise, capabilities, assets, brands, customers. None of that would apply, so it wouldn't make any sense. So it still needs to be adjacent, which means leveraging some of the things I just said, where it's like, 'Well, if we built this, our current customers would want it.' So the distribution channel would be in. Now maybe or maybe not they would want this new product. That's the big bet, but at least we have the distribution. So that would be a good bet, right? That's the typical bet because distribution's harder than building, right? So

but in AI actually, that's not true anymore. It's flipped in AI. It's easier to get distribution than it is to build the product because AI products almost never work. Like most of the things I use don't work. I spend a lot of time, I spend, Cloud Code is correcting it. Right? And then sometimes it gets it really good and it's amazing. I'm not sure if it's saving time overall. I think so. But like, you know, sometimes it saves a ton of time, but then sometimes like you're just wrestling with it and you're like, 'What am I doing? It's like three lines of code. I could just fix this. It's so confused, you know?' Or I just watched a video the other day of someone showing how to use Cloud Code better. I'm like, 'Good. Maybe I need to use it better.' And I would say like in 45 minutes, probably a dozen times, he was like, 'See, we'll just tell Cloud Code to do this. Oh, well, it didn't quite. Well, you know, in other words, like it doesn't work that well, right?' A lot of the support AI doesn't work that well. A lot of times when you talk to support now, it's still not AI. But I thought we, I thought it was already there. Well, they're not.

Yeah. So in AI, it's funny. Everybody wants to try it. Like you make a new AI product, it's very easy to get people to try it. And then they all cancel in 4 months because it doesn't really work. So AI has this weird opposite pattern to how products have been. The other thing though is again this notion of like, 'Wait, we already have a product, so is that different?' Yes, it is different. So I think there's three, you might say, categories of AI products. And you know, you could split it up in any number you want. I find this particular one useful. One is you have incumbents who are adding AI to the product. So Notion is adding AI and Google Sheets adds AI and so forth. It doesn't work very well, right? Have you tried using the stuff in SharePoint? It's okay, and it doesn't help that much, right? So okay, but they still should do that, right? Because if they can get it to be better and better, they do need to do that. So they're smart, but it's not that interesting.

The second kind is AI for experts. So you are a software developer, can AI make you a better software developer? You're a designer, can AI make you a better designer? You're a marketing writer, can it help you write content better, faster, more, whatever. And so there the answer to me is often yes. And the reason is that because to the extent the AI doesn't give you the right answer, the expert knows what to do anyway. If it writes something that isn't good, the writer fixes it. You know, if the design isn't good, the designer just doesn't use it and so forth. So the expert was already doing the job and therefore can still do the job and maybe AI can accelerate parts of it. Great. That's an area where I think if you're making a startup in that kind of a zone, the fact that AI doesn't work all the time is okay or could be made to be okay.

The third kind is AI for muggles. I'm not a software developer, I want to build an app anyway. I'm not a designer, I want to make a website anyway. Right? I'm not a writer, I want to write a book anyway. And the thing is AI's not good enough for that. So it is good enough to be a lot of fun. Right? If you're not a coder, you can make some demo, some weird thing that you have to maintain for the rest of your life and you don't know how to do it and then you can't bring it over the line. Like you can't get that last bit to make it good because you don't know how. And you can't cajole the AI enough, you don't even know what to ask necessarily. And so while that's really interesting and again, you could make a bet and you could guess that some of these things will work, but you better put a timeline on it, you know.

Because it surely, like in 50 years I suppose though, you know, I don't know. Anyway, that's not in sight right now. So you could do that. A lot of the things where people say they're going to change the whole world are in that category. So I'm not saying don't do it or I don't like it or anything like that. It's just it's not as obvious as it might seem that that's possible right now. So I know if you succeed it's probably great, you're probably right. Like in that sense it's an asymmetric bet in your favor. It's just I think the probability of success is much lower than people give it credit for because AI is not that good.

I've never heard of someone getting to that point who's not a coder. Not once ever. So yeah, they're struggling. But I mean that's part of why I don't write about AI, I don't talk about AI because whatever we say it probably doesn't matter in 6 or 12 months it'll be a different set of things. So I like to try to write about things that are evergreen. I guess I would say my, you know, if the, like I was saying a vision statement is something that you try to do. It may not succeed but that's your goal. So I want to write things where they're true, you know, 50 years from now. So some of them are, like some of the things I've written are almost 20 years old and they read like, so then people will read them and say, 'Holy crap, I didn't realize this was written in 2008. It feels like it was written now.' I'm like, 'Great, that's my goal.' Now I'm not claiming they all succeed or that I even believe myself from 20 years ago, right? I've grown and changed, haven't I? You know, so it's not like I'm saying that that actually happens all the time but that's my goal. So I don't talk about AI because that's going to be different and I'm going to be wrong. So I try to talk about things that I feel are at least worth thinking about and considering if not deeply true. And that maybe helps people think or helps people understand a certain mechanism or something like that that might still be true. That's what I do. So I'm not an AI expert, I'm not an AI strategist. I can explain what we're doing but you know.

Oh, I don't know. I mean there's hundreds of articles so.

But the topic, you know, things about founder psychology I don't think have changed at all. They're sort of like this weird combination of actually thinking your ideas are better than everyone else and also having imposter syndrome at the same time which is very funny. There's tricks like, you know, I've been at it for 6 months, it's not really working but I'm not sure whether what I should do is persevere and push through and it's coming or whether I should give up. That's very difficult and I don't think AI solves it by the way, you know.