Why Fund Economics Need Overconfident Founders
A typical early-stage venture fund makes about twenty-five investments. Its returns to limited partners depend almost entirely on whether one or two turned out to be extreme outliers. What that math forces, in three steps, is the part this piece is about.
A typical early-stage venture fund makes about twenty-five investments. Most of them will return less than the capital invested. A few will return between one and three times capital. A small number will return ten times or more. The fund returns to its limited partners depend, almost entirely, on whether one or two of those investments turned out to be extreme outliers.
This is not a bug or a sign of poor judgment. It is what venture-class returns look like when the underlying distribution is power-law rather than normal. The math is well-attested across multiple fund-level studies (Kauffman Foundation 2012, Cambridge Associates pooled data, Horsley Bridge sample analysis) [STRONG]. What follows from the math, in three steps, is the part this short piece is about.
Step one: funds must reject the merely-good
A company that will plausibly return three times capital is good for its founders, good for its early employees, and good for the LP capital allocated to it on its own terms. But it cannot be the outlier the fund needs. From the fund's perspective, partner attention spent on a reliably-three-times company is partner attention not spent looking for the rare hundred-times company. So the fund systematically rejects the merely-good in favour of the speculatively-large. This is not a mistake. It is the fund responding to its own economics correctly.
Step two: founders must persuade the fund they are the outlier
A founder pitching a fund cannot succeed by acknowledging, accurately, that their company is most likely to be one of the twenty-four-out-of-twenty-five returns less than capital. They have to credibly claim they are in the upper tail. The selection mechanism rewards confidence in one's own outlier-ness. Founders who present themselves as such get funded; founders who present themselves as merely-good do not.
This is where the fund-level math reaches the individual founder. The math at the top of the funnel has, by the time it reaches the founder, turned into a requirement that the founder be the kind of person who believes they are the outlier. Not the kind of person who privately suspects, but the kind of person who has internalised the belief well enough to perform it under sustained scrutiny.
Step three: the population must be large enough to find them
For funds to find their outliers, the population of founders pitching them must be large. For the population to stay large, the broader recruitment environment around the venture system — the visible success stories, the founder hero narrative, the cultural form in which entrepreneurship is rendered — must persuade many people that they could be the outlier.
The system needs many people to try in order to find the few who succeed. Most of the trying will not succeed. The math at the top of the funnel cannot work without the population at the bottom believing it might. The recruitment environment is the part of fund economics that operates on the population of prospective founders, before any fund partner ever meets any of them.
What falls out of this
None of the three steps requires bad faith from any individual VC. Each is locally rational. The aggregate of the three is a system that systematically rewards founders for over-estimating their personal probability of success and that maintains a recruitment environment which produces the over-estimation it then rewards.
What the funding step selects for is not conviction itself but performable conviction — the ability to hold a story in front of a partner under sustained scrutiny and have it survive. The publication's sister book theheld.ai opens with the alternative: a founder, walking in a London park after two failed companies, saying the single word so in answer to the observation that other people are already doing the thing he wants to do. That word, in that voice, was enough to move a cheque. It is not the kind of signal a fund partner is selecting for, and it is not the kind of signal the recruitment environment amplifies. The point is not that one is correct and the other wrong. The point is that the math at the top of the funnel selects for the first kind of signal and not the second, and the public record of what a founder looks like is shaped accordingly.
Founders who do not have the over-estimation cannot get funded. Founders who do have it tend to be wrong about themselves, in expectation, because the base rates apply to them as much as anyone. The welfare cost of being wrong falls on them, individually. The aggregate output — the technologies, the employment, the diffuse gains the longer treatment of this question defends — depends on a population of individuals being wrong, in expectation, about themselves.
That is what fund economics force. Not a moral judgment about anyone in the chain. A description of what the chain produces when each link is responding to its own incentives correctly.
The full mechanism, including what would change the pattern if the math changed, is in The Power Law and What It Forces. The companion piece The 33% looks at what this mechanism produces in the population of prospective founders. The deep treatment of the question whether the system is, on balance, good for society is in VC: Most Fail, Most Suffer, Some Win Lots.