The counterintuitive ideas that make Entrepreneur First work

Last week I had the chance to talk to Matthew Clifford. Matt is the CEO and founder of Entrepreneur First, a talent investor that funds hundreds of founders every year, pre-idea and pre-team.

They run cohorts twice a year, in six different countries around the globe and during the programme, they help people find a co-founder, develop an idea, and start a company from scratch.

As of 2019 the company has had 1000+ people go through its programme, creating 200+ companies worth a combined $1.5 billion.

It was a fascinating conversation on investing, technology and (some) politics. It was so good, that after I wrote a 2,000 piece on it, I realized the transcript was an order of magnitude better.

If you want to learn more about Matt, I absolutely recommend his newsletter, Thoughts in Between.

The Interview

Thanks for jumping on the phone with me. As I said, I’m fascinated by Entrepreneur First. One of the few things that is a highlight for me is the idea of co-founder matching. What do you think is the counter-intuitive idea that makes the entire thing work?

Matthew Clifford: I think there are a few things. Some of it is very intuitive, and then there’s a couple of counterintuitive things. I guess that’s, in my experience, the key to most startups. It’s 90% obvious and 10% twist and that makes the whole thing work. The obvious start is, and this is sort of what we talked about on Twitter, you have to start with the right people, I don’t believe you could do it with a random sample of the population. They do have to be very smart, very driven, very ambitious. And you do need a really carefully curated set of skills, background, areas of focus. And you need a structure, you need a time-limit and you need some incentives that push people in the right direction. I think that’s all really obvious.

I think there are two things that are less obvious but that turn out to be really important. The first is: your job as the program is not to bring people together, that’s the first non-intuitive thing. People say ‘how do you build great teams?’ and we answer, “Oh, no, we don’t. We just build loads of teams, get rid of the bad ones and the good ones are the ones that are left.” That sounds facetious but it’s really true. Our main role during the program is to say “maybe you should break-up, maybe this isn’t the right team for you. Maybe you should go back into the pool and find someone else.” And what that means is that it kinda gives this survivorship bias in our favor. By the time you get to the end of twelve weeks, the stuff that’s left is really, really good.

The second counterintuitive thing, which just really comes to this point about starting before there is an idea, is that the most important thing by far is productivity, even if it’s in the wrong direction. So, I think, that others try to evaluate teams by like putting an almost VC lense over what they build during those first twelve weeks. People ask questions like: How big is the market? Or, what are the gross margins? And the answer is it doesn’t matter, it’s a twelve-week-old team. To try to evaluate whether we should invest in this time over stuff like that makes no sense. Intuitively, that’s what people do. What we do instead is, we say… startups, at least in the seed stage, are really learning machines. They’re about how quickly can you resolve uncertainty in the model that you’re developing, and therefore what matters most is the speed of learning. Even if in the wrong direction. And the best proxy to speed of learning is productivity.

What we now do is we set start-ups and, you know, what would a great week look like? What would you be able to achieve this week if everything worked? And then, basically, a week later, you say: What happened? Did you do it? And broadly the teams that consistently are productive, even if they’re building the wrong thing, do better than the people whose ideas sound great but can’t find the right thing.

Yeah, you had that quote: “You rather get a team with high productivity in the wrong direction than low productivity in the right direction.”

You also mentioned that Entrepreneur First has a bunch of criteria for selecting people. Problem solver. Smart. A track-record. Being determined. But how do you pre-select people for productivity?

MC: It’s very hard to, that’s the truth. There are a few things that are good indicators. For example, I think that it’s very unlikely that someone that’s never been productive will suddenly become productive because they decide to start a company. You can do basic competency-based interviewing, like what you would do in a normal job where you get people to give you examples of stuff they’ve done in the past… but the twist is that what we really are talking about is joint productivity.

I think that you can be a very productive individual but what matters most is, “can you be more than the sum of your parts with your co-founder?” One of the reasons we have the structural model that we have, both in terms of how we spend our time and how we deploy our capital, but also how much risk we ask founders to take, is that we say: over the course of these 12 weeks, neither side knows what we’re getting into.  You don’t know if this actually for you, you don’t know if you want to build a business, necessarily… Well, that’s not true.

You definitely know you want to build a business, but you may not know you want to build this business. We don’t know anything about you, but we’re willing to deploy 10,000 dollars on a person because it seems like they could be great. What we get to assess over the next twelve weeks is whether that person is very productive, in combination with their co-founder. And that’s why we both stage our capital and their time risk such that we’re then only making an investment decision fourteen weeks later. So, by the time you’ve been observing someone for fourteen weeks, you know a lot about them and they know a lot about themselves. I really think one of the best reasons to do EF is just self-discovery. I don’t know if it’s possible to sit down and analyse yourself at a desk and decide whether you’re “ready” to start a company. I actually think one of the greatest things about EF is that it gives people a relatively risk-free way of evaluating whether this is something that they really care about.

It’s an asymmetrical bet, for both parties. It’s $10,000 on your end and a few months on theirs. And if the people are good, it should have zero impact on their career

MC: Exactly.

We were talking about joint productivity and being more than the sum of the parts. Have you found any connecting patterns in terms of co-founder matching? Something you weren’t expecting? I don’t know, it could be a younger founder matches better with an older founder because their levels of experience fit well?

MC: That’s a really good question. I don’t know if we could go as far as saying that they match better, but certainly that people are more likely to match with people near their own age. One thing we found that is quite striking is that early on when we had cohorts that were on average much younger than they’re now – I think that the average age in the first three cohorts would be 22, 23, some younger. The average age today is closer to 30. Today, younger people do less well in those cohorts than they did early on. I don’t think it’s because they’re less talented. I think it’s because it’s a higher bar in a way for a 29-year-old to look at a 23-year-old and say ‘I wanna work with you’, even if they’re actually the most talented person. I do think that you need a degree of homogeneity in a cohort. I don’t mean gender or ethnicity or anything like that but… It’s pretty hard to have a 19-year-old and a 60-year-old and expect that to work. Doesn’t mean it can’t work, but it’s just harder.

One of the main reasons that startups fail is that people give up and co-founders split. Other than just poaching holes through the team, is there any way that Entrepreneur First protects from that? How do you solve for that? Because a big reason that start-ups might stay together through the dark times is the guilt of separating from a friend or colleague. Those underlying forces that often bring startups together aren’t there when you don’t know your co-founder.

MC: Yeah, I mean…

Maybe I’m wrong.

MC: No, no, it’s a great question. One of my contrarian beliefs is actually the opposite. To me, it’s sort of weird that the default is: you know, VCs really want to seek founders that have known each other for ten years and worked together before… To me, this is a little bit like a medieval arranged marriage, like you have to marry someone in your village. Wouldn’t it be better to expand the pool and find some sort of optimal way to decide on the right person? To me, it almost feels like in the same way that, at the start of the online dating phenomena, if you met your partner through an online dating service you kinda pretended as you met them at a bar because it was less embarrassing… but today, everyone meets someone online. I met my wife in college so I don’t fit that model. But everyone meets someone online today.

In a way, it seems so crazy to trust it to chance to meet at a bar when you could do really great filtering on an app. I feel  we’ll eventually get there on co-founders. I would argue that all the forces that are meant to be positive that people have relied on for a long time, are also potential negatives.

Yeah, it’s true you feel guilty for breaking up with someone you’ve known for a long time. But you also probably can flip that entirely and say you've probably stayed together for too long. Of course, if you’re like Patrick and John Collison (editor’s note: co-founders of Stripe), that’s awesome. But if I had started a company with my brother, it would be a much higher bar for me to say you know what, this is not the opportunity that maximises our potential outcomes so let’s stop.

Whereas, obviously at EF you get to say that multiple times. Now, I think the interesting question is when does that stop being a benefit and become a cost? One of the reasons that we put a lot of, I don’t wanna say pressure, but we really try and test the relationships early, our goal is that if you get to Demo Day, you’re gonna be really robust. If you came through the baptism of fire that is EF, it is very, very unlikely that you’ll break up after this point.

I think that the data supports that that is true. It is quite hard to gather data on something like this. But I was the other day at an event hosted by a Limited Partner that funded many, many GPs from across Europe and I was speaking about talent investment. I asked people, informally, to raise their hand if your level of founder break-up in your portfolio is above various thresholds. I wanna say that the average was well north of 20%. So, well north of 20% of people see deals lose a co-founder before Series-A. And I think that number at EF for seed funded companies is way below 10%.

I think the reason is that you would have already done it when we were like ‘you should break up, you should break up.’ Actually, some times starting with a friend doesn’t give you the opportunity to say: this is actually the wrong thing.

It’s an interesting take, absolutely. I think you might be onto something. You mentioned arranged marriages. I know you are a Medieval History fan.

MC: My first degree is in Medieval History.

I believe that there are specific events that change the course of history. For example, modern Western civilisation. A big part of that is modelled after Roman law, language and culture. But if Athens had won the battle of Syracuse, all those influences on today’s society would have been Greek. How big of a part do you think that the acquisition of Magic Pony, that one big event, played in the course of EF?

MC: It’s a great question. So, I would say, financially it was very important, frankly. Because venture is a very, very long game and I didn’t start EF wealthy, and neither did Alice. Neither of us had money from our families, we’re not from wealthy backgrounds.

You do this for like five years and you start to think “Am I an idiot? Should I be a management consultant?” There is a moment, being really frank, that psychologically having a really big win which your personal exposure to it is quite large, is like “huh, okay. This is validating that we’re not throwing our lives away.” That was really valuable.

One thing I’ve always been surprised by, that has really become a very important lesson about how I try and help founders is: it is very tempting to believe that there are these silver bullets that will happen in your business that, when they happen, everything just gets easier. We’ve had probably a dozen of these over the last 8 years. And when we started, everyone thought we were absolute idiots. We were incredibly young, we had no track record, no experience, no money, no brand. Everyone said, this is stupid, don’t do it. But we kept saying to each other “as soon as we have the first cohort, it will get easy.” And guess what? It didn’t get easy. So we kept saying things like this. “When we first raise from a fund it will get easy.”

It wasn’t…

Magic Pony was definitely one of those. When we announced this company was being sold for 150 million dollars, fifteen months after doing EF, we said “Now it will be easy!” It certainly didn’t hurt, but it didn’t double applications or anything. I think is like one of those interesting things where, as a marketing problem in general, the people that you want to reach often are the people that are not paying attention. It was great, but it didn’t overnight move the needle. Similarly, when we raised the round of funding from Reid Hoffman, that was not quite a dream come true, but definitely beyond our wildest expectations. Reid Hoffman had been a hero of mine since before EF and, he has been incredible to work with. It’s a massive privilege to have him on our board. But I also thought, wow, from now it’ll just be easy. And it wasn’t easy. It didn’t just double applications. These things have bumps, but they didn’t go exponential. So I kinda feel like, yeah, Magic Pony was great, but I feel like the biggest reason it potentially was a turning point is that if we hadn’t had that, potentially we would have started thinking “are we doing the right thing? Is the opportunity cost too high?” I think it came to that, but it didn’t transform, sadly, the business. As great as it was…

Great answer. And I think, like in everything in life, it doesn’t get easier. You just get problems that are increasingly hard. That goes for work, strength training, relationships… Things just keep pushing the level of difficulty.

MC: Yeah, you get higher quality problems. I have a toddler and, yeah, it’s more and more higher quality problems.

It seems that opportunity cost as a mental model is something that you think about pretty often. There’s no denying that so far the EF is, by any standard, successful. That said, do you think it’s the most successful model it can be, in generating impact? Why for instance the EF model vs a YC model… What’s the opportunity cost of being a talent investor versus other models?

MC: One of my vices is that I’m very ambitious. And I wouldn’t be doing this if I didn’t think it wasn’t the biggest impact thing I could be doing with my life.

I think YC has just been the most phenomenal success because it captured just the perfect opportunity at the perfect time and place. It had this insight which, which in retrospect is obvious but it wasn’t at the time, that Silicon Valley is a network-intensive game and if you could just find a way to be “the gatekeeper”, or to be on the inside rather than the outside… You could charge well above market price or often times invest below market price for that. And that there would be compounding advantages to that. Because you’re basically defining the inside, it gets better and better. It’s a beautiful model. What’s striking is that it’s been very difficult to copy.

Techstars is a great business, but it’s a great business in a very different way. They have a different inside which is about how you really make corporate innovation work on an actual scale and, again, I don’t know it as well but I think it’s a great business… but no-one has been able to take the YC model and really make YC work anywhere else or in a different way. The YC model is for a place and time, and the time is on-going but the place is still Silicon Valley.

It strikes us that the opportunity in talent investing is, in theory, orders of magnitude bigger than the opportunity in acceleration. The scope of the market is limited by the number of teams ready to be accelerated. In talent investing the scope of the market is limited only by the population of people that could be great founders. Right now, if you sit in Silicon Valley, it looks like that’s not true. You see many more great teams than you do great people that can’t find a co-founder. But that’s because you’re in Silicon Valley! Everywhere else, there isn’t the density of network around founding that means that the very best people know other people that want to start companies. Our view is that, as EF scales, being able to capture and amplify the efforts of and nurture really exceptional talents that are outside startup ecosystems starts to feel like the most valuable thing you can do. If you think really that talent and capital are the two resources that make the world go round, our job is to try and be the best possible home to the best possible talent. I think that if we do that well, we’re not competing with YC because it’s a different model, but I think we can have the same impact as YC.

Well, you guys are not competing.

MC: Yeah, yeah, I know. I’m a huge fan of what they’ve done, it’s pretty extraordinary.

It is, and hopefully, we’ll chat again in a couple of years and we’ll say the same thing about you guys, about EF. Speaking of that – scale and impact. How do you see EF in the next 5, 10 years?

MC: If you go back and see how we think about our mission… you know, we want to transform the lives of the most impactful people in the world. Full stop. And if you think about what are the levers you need to pull to achieve that, it’s pretty basic. It’s kinda like a funnel. You need to be accessible to as many as those people as possible, whether that be geographically or legally or whatever. You need to be attractive to as many as those people as possible. There’s no reason to be around if they don’t want to do it. And crucially, you need to be effective. If you provide that service and don’t get real outcomes, it doesn’t work.

The way we think about everything we do with scale is that it’s gotta pull on at least one of those three levers and, ideally, two or all of them. Obviously, over the last two years, the big focus has been accessibility which we’ve seen as being primarily a geographic issue. As you know, we just pulled out of Hong Kong in January, at the end of this cohort. Largely because of Market Size. If you go back to this accessibility thing, you just see it relative to the amount of capital you have to deploy, you’re not reaching the number of people as you could in other locations. But when I think about what the next five years look like, I think the real question comes down to something a little bit different. Which is, right now, if you’re to think about what are the real bottlenecks when it comes to achieving our vision is that every site where we wanna operate needs to tick both the talent box and the capital box. Why is it working in London so well? We’ve got great people, it’s a magnet for talent. But we’ve also got a very vibrant VC ecosystem. So if you put out 30 companies every 6 months, the availability of capital is not a constraint on your success. If we were to go to, I don’t know, Uganda, my guess is that the talent is great. I don’t know anything about it but I can’t think why it wouldn’t be. But, would I open EF in Uganda now? No. because I think that we could put out 10, 15, 20 companies a year and just like, there wouldn’t be a seed ecosystem. So the big question for EF is, how do you help talent transcend the ecosystem that it is in? I think there are a few things you can do. I think we’re very lucky to have such a great network in Silicon Valley and having Reid (editor’s note: Hoffman, co-founder  of Linkedin) on board, and founder’s fund and things like that, it gives us the opportunity to really create a more systematic, more robust pathway to Silicon Valley… and I think, again, we’re not gonna open up in Uganda in the next 18 months, but you can imagine a world where you could run EF in Uganda, or Mongolia or Ecuador and as long as you had a sufficiently systematic way of helping founders access capital and expertise from the world’s major ecosystems, you could make that work.

I see our role, strategically, over the 18 months is; how do we get to the stage where geography is less and less of a factor to determine who starts a start-up. Today, I would say it’s probably the highest order bit. Everyone wants to think it’s talent and ability, but honestly, if you’re born with easy access to Silicon Valley or in Silicon Valley, or end up in Stanford, that’s the highest order bit by far.

Do you think you’ll ever reach a point in which EF could create those geographies? If you bring enough talented people together in one place, and EF could be the driver, I see no reason why it couldn’t happen. Talented people attract more talented people, which ends up attracting all the other stuff from the ecosystem like capital. Do you think you’ll ever reach a point in which you can say: ‘Hey, I’m gonna go to this small town in Spain and the sheer density of amazing people will end up attracting amazing people?’

MC: I think and hope so. The question becomes then… I think what EF has is very strong local network effects. So, I think that so at the cohort level, a cohort in London gets better the more great people join it, so I think that’s true even across cohorts in London. If you are in, say, the twelfth cohort in London, you benefit if amazing people join the thirteenth cohort.

What I don’t think is true yet, that there true global network effects that we’ve managed to activate. I mean, at some very marginal level maybe the brand gets stronger if amazing people join in Bangalore that makes it better for someone in Berlin, but I can see a world in which those effects become very strong. I think you would require a slightly different model but I love the idea to be able to, you know, sometimes we say: “How do we make the infrastructure of EF so robust so it’s almost like the Amazon Web Services of running EF. ” As if you could spin off a site with a six-week notice, in any place and choose to either double down or spin it down without it being too expensive. My guess is that a good test would be, can you show up at a town and through the strength of the network, the brand, the software, the reputation, build a cohort very very quickly. Probably. But I think we’ll have to see.

What do you think are the most interesting problems to work on right now?

MC: Let me answer that with a European view. The answer is there are many but I think that with a European view, there are a few very clear things that I feel very strongly about.

I think the most annoying question in the world is: “Why hasn’t there been a google of Europe?” I mean, not the most annoying, but it’s amazing to me how popular that question is among policymakers and politicians. And the answer is really simple. One, because the people that would start those companies in Europe historically have not started them. You know, the Larry and Sergeis of Europe are probably working on a hedge fund, or for Goldman Sachs, because that’s the culture, or has been the culture. Second, there are structural reasons why scaling consumer internet companies in Europe is harder than in the US or China. I think the mistake is then to go to the next level and say: “That means there shouldn’t be Venture Capital in Europe, or we shouldn’t be building startups in Europe.”

To me, the question is: what kind of companies in Europe have a structural advantage? And I think there are some really obvious answers. We’ve made a multi-year, multi-national bet that machine learning is going to reinvent almost every possible part of the Enterprise stack, and I don’t just mean the software stack but the whole way of doing business. And, I think that enterprise machine learning is just getting started. If you think about the kind of tools that have some penetration right now, they’re kinda horizontal. People building frameworks and tools that allow companies to implement machine learning or whatever. What’s really exciting to me is what happens to the life sciences, to health care, to energy, to materials, to manufacturing, to engineering, when you take a machine-learning-first approach, to these huge, hundred billion dollar plus industries. I think the answer is you build hundred-billion-dollar companies. And the reason that is really cool is that the scale effect that holds you back with Consumer Internet in Europe, which is the exact scale consumer internet companies, say, ten million dollars in revenue to hundred million, to a billion, those guys are so concentrated in Silicon Valley. But the scale execs for the five or six industries I mentioned, they are in Europe. There are people who absolutely know how to scale those kinds of companies in Europe. And when those two worlds collide, I think Europe has as good a chance as anywhere else in building those companies.

So, at the end of the day, you think the edge across all those things is talent?

MC: I think it’s talent at the founder level, and I think it’s talent at the scaling level. I think the scaling talent bit is often ignored, or under-appreciated. We’ve just hired two really senior experienced people who have been through huge scaling before and you just realize, wow, you can’t intuit a lot of it.

A lot of scale is counter-intuitive and you just need people who have done it before. And, you know, it’s hard in software. In Europe, what are the companies who have done a billion dollars in Software sales? There’s some, but not many. If that’s your aspiration, you need to find a way to access those people in the Valley, or elsewhere. I think that’s gonna be hard for a while. I don’t think it’s hard in the same way in some of those ‘real world’, that’s not meant to be disparaging to software but, you know, world of atoms companies.

You’ve spoken about the US and China. How do you see the dynamics of the US and China playing over time in terms of talent, companies, venture… It’s an interesting topic, at least to me.

MC: I think it’s the important ‘Politics meets Tech’ topic. It’s really… even though there’s more and more talk about it, I still think it’s underrated.

It is, it is. That’s because people have never been to China.

MC: Right, yeah. I think there are a few things. I think that one, I also write a newsletter (although it’s not really about startups, occasionally, but it’s hugely about tech meets politics) and one of the things I’ve been writing a lot this year is just that there’s a real asymmetry in the ability of US firms to project influences into China and the inverse. In general, they have been completely or nearly completely unable to project influence into China. And the Chinese firms are getting better and better at doing the opposite.

Well, you can look at the NBA case.

MC: It’s so striking. I think it’s a watershed moment in history. Full Stop. It is the moment where it becomes a proper clash of values where the rising power starts to intrude in the dominant power’s home sphere. That’s a big, big thing. I think it also means that Europe is set for a fascinating role, for a couple of reasons. And Brexit actually becomes really relevant. One, I think, it’s pretty clear… well, we’ll see who’s elected president next year but, you know, Europe is currently the global (as in Ex-China) regulator of first resort for tech. There is currently a sense that the impulse of that is anti-Americanism. Maybe true, actually, to an extent… not anti-Americanism, but anti-American big business.

I think there’s also a lot of good reasons to think we need to regulate Tech the way we do, but I think that feeling is creating some structural barriers to the regulation of tech in the US. As in the sense that their “assets” are being attacked. If Elizabeth Warren becomes president we might see that unwind but, you know, hold that thought.

I think Europe has the scale to be relevant to Google, Facebook, Amazon, etc. So, GDPR is the most obvious example, but it’s actually gonna be more and more where you do see compliance and non-compliance turns very expensive. I think there is a question about: To what extent do we feel that that combined with the asymmetry of Chinese firms and big American firms… At what point does it become a geopolitical strategic question for Europe about… does it pick a side? Does it continue to say it’s gonna produce local champions? Which as I said, I do not think it’s credible in the consumer internet space. Or does it somehow try and play a broker role? Which I think would be a strategic mistake. My view is that plays very much to the way China views the world. Europe does share American values in a deep way…

It’s the West.

MC: It’s the West, yes. And I think that’s become a slightly dirty term but, I very passionately believe that it’s the right idea, the right framing. And I think that  Europe does need to take a… it needs to balance these geopolitical considerations against consumer rights issues which I also think are important. I think that’s the first thing I would say. The second thing is that I think there is a… I mean, how do we avoid a race to the bottom in a lot of these things? Particularly around AI ethics? Some of the stuff that is already happening in China is pretty worrying. It’s in the news a lot.

So, there’s a question of, is it possible to avoid a race to the bottom if you fragment the rest of the world’s markets? It’s the least sexy topic imaginable, but standard-setting, creating global standards with voluntary compliance across the west could be one of the highest impact things anyone can do right now. Yes, China is going to be huge and it shouldn’t be ignored, and it’s not going to be ignored, but I think there’s a clear sense there’s a standard outside China and then China has to decide whether to play with those or not. That’s a very different world than where there’s essentially a bipolar race and Europe decides to remain neutral. I think very passionately that that’s the wrong call.

How do you make sure those standards even matter? One of the things the west got right is the value of the individual. And that’s the one thing that China doesn’t really care that much about. It’s that race to the bottom, as you said, that it’s worrying. How do you think those things would play? Europe could set a standard, absolutely, but China will never even care about that standard.

MC: I would agree that… I think China will care if there is a question of… China cares, I think, much more than it did in the past. I’m not a China expert but, you know, I try to read about it. And my impression is that China cares much more than it did in the past about projecting influence and power outside China. Given that, there is a question of to what extent the rest of the world can coordinate what they want the standard to be. To me, the most damaging thing by far about the Trump phenomenon is the destruction of multi-lateral action by the west at a time when it’s most important.

If the US isolates itself, then I think the race to the bottom is inevitable because there’s no way of coordinating. But China doesn’t want to be in a position where it can’t access either the US or the European market. I actually think there are a lot of core values of the US and Europe that make the western system very attractive, at least in theory, for Latin America, Central Asia, Africa… And now is the time to be building those relationships and I fear we’re doing the opposite.

Well, it’s 6 pm already so thank you so much for your time. I could keep talking about politics and China and the US and big tech but I know you’re a busy guy. It’s been a pleasure and a fascinating conversation.