BackStartups

The shape of returns

The power law

Why almost all of an angel portfolio's return comes from one or two investments, and what that single fact implies about everything else you do.

Venture returns are not distributed normally. In a normal distribution most outcomes cluster around an average and extremes are rare, which is how most of the world behaves and how most people instinctively model risk. Startup outcomes do not work that way: most investments return nothing at all, a few return the capital, and a very small number return enough to pay for everything else several times over.

This is not a claim about probability so much as a claim about arithmetic. A company that fails can only take your cheque; a company that succeeds can return a hundred times it. The downside is bounded at 1× and the upside is not, so any portfolio of enough investments will have its total result determined by its largest single outcome.

Everything else on this site follows from that. It is why diversification matters more than selection quality, why the entry price matters more than it feels like it should, why follow-on capital is worth reserving, and why selling a winner early is the most expensive mistake available to an angel.

What the distribution does to a portfolio

Consider a portfolio of twenty investments where nine fail completely, six return roughly what went in, four return a few times the cheque, and one returns fifty times. The nine failures cost you nine units. The six break-evens contribute nothing. The four modest winners contribute perhaps ten units of profit between them. The single large winner contributes forty-nine.

The portfolio works. It works entirely because of one investment, and that investment was not identifiable in advance — if it had been, you would have put everything into it. This is the structure that angel investing has to be designed around.

The uncomfortable implication is that a portfolio without a large winner does not return capital, no matter how sensible each individual decision was. Twenty carefully chosen companies that each return two or three times is a worse outcome than nineteen failures and one fifty-times return, and it is a much more likely one for an investor who avoids risk within the portfolio.

An illustrative twenty-investment portfolio at £10,000 per cheque
OutcomeInvestmentsInvestedReturned
Total loss9£90,000£0
Roughly break-even (1×)6£60,000£60,000
Modest win (3×)4£40,000£120,000
One large winner (50×)1£10,000£500,000
Portfolio total20£200,000£680,000

This is a constructed illustration to show the arithmetic, not observed data from any study or fund.

Why selection cannot save you

The instinctive response to a high failure rate is to select harder — more diligence, higher standards, fewer investments. This helps less than it feels like it should, for two reasons.

The first is that the failure rate is driven mostly by things that are not knowable in advance. Markets do not appear, co-founders fall out, a larger competitor arrives, funding conditions change three years after you invest. Diligence catches the visible problems, and the visible problems are not the main cause of failure.

The second is that raising your bar shrinks your portfolio, and a smaller portfolio is less likely to contain the outlier that pays for it. An investor who makes five very careful investments has a meaningfully high chance of holding no large winner at all. The same investor making twenty investments of the same average quality has a much better chance, because they have bought more tickets in a lottery where one ticket pays for the book.

This is not an argument against diligence, which prevents specific and avoidable losses. It is an argument against using diligence as a substitute for portfolio size.

What follows from it

Four practical consequences, each of which has its own chapter on this site.

Portfolio size must be large enough to contain an outlier — which is the arithmetic behind the recommendation that angels plan for twenty or more investments rather than five. Cheque size must therefore be small enough that twenty of them is affordable, which is usually a smaller number than a new angel first has in mind.

Entry price matters disproportionately, because the outlier's return is divided by what you paid to enter it. And winners must not be sold early: an investment sold at 5× to lock in a gain removes the possibility that it becomes the 50× that pays for the portfolio.

Try it yourself

Stop reading, start calculating

In short

What to take away

  • Design for one outlier, not for a good average. A portfolio of uniformly decent outcomes does not work.
  • Portfolio size is a bigger lever than selection quality, because it determines whether you hold an outlier at all.
  • Never sell a compounding winner to lock in a gain — that is the position paying for everything else.
  • The entry price on the winner determines the size of the whole portfolio's return.

The power law: common questions

Does the power law mean diligence is pointless?
No. Diligence prevents the specific, visible failures — unassigned IP, a broken cap table, a preference stack that makes your shares worthless. What it cannot do is raise the base rate enough to compensate for a small portfolio, which is why it is a complement to portfolio size rather than a substitute for it.
How large does the largest winner need to be?
Large enough to return the entire portfolio and more, which with twenty cheques means somewhere upward of twenty times your entry — and in practice, well upward, since dilution reduces what a given exit returns to you. This is why entry price and follow-on rights matter so much.
Is the power law specific to venture?
It appears wherever outcomes are unbounded on the upside and bounded on the downside — book sales, film revenue, city sizes. Venture is a particularly clean example because the downside is capped at the cheque and the upside genuinely is not.