The power law

A seed fund invests in 25 companies. Sixteen of them close and return nothing. Is the fund a failure? Many readers say yes. In venture capital, the answer depends on one company. If one of the other nine becomes very large, the fund can be a great success. This lesson explains why, and what it means for you as a founder.

Most startups do not return the money

The data on venture outcomes is clear. Most investments return less than they cost:

• Correlation Ventures studied more than 21,000 financings from 2004 to 2013. Of these, 65% returned less than 1x, 10% returned 5x or more, and only 4% returned 10x or more.
• Horsley Bridge, an investor in many funds, studied approximately 7,000 investments from 1985 to 2014. Approximately half returned less than their cost.
• A 2023 update from Correlation covers companies that exited in the 10 years before 2023. Almost half of the financings lost money, and less than 4% of the invested dollars returned 10x or more.

These multiples are gross: they come before the fund’s fees and carry. The share of investments that return less than their cost is the loss ratio. In venture capital, it is usually more than half.

A few winners pay for everything

When most results are small and a few are huge, the results follow a power law. In Horsley Bridge data, approximately 6% of the deals made approximately 60% of all the returns. Those 6% were the deals that returned 10x or more, and they used only 4.5% of the dollars.

Predict first

A fund puts $1M into each of 25 companies. Sixteen return nothing, six return 2x, two return 5x, and one returns 50x. What multiple does the fund return on its $25M?

Type a guess before you continue. A guess first, even a wrong one, makes the answer easier to remember.

Your guessSave my guess

The fund in this example has a fund returner: one company that returns more than the whole fund by itself. To give back $25M alone, a company with a $1M check must return 25x. The 50x company returned the fund two times.

Fund multiple with all 25 companies    = $72M ÷ $25M = 2.88x
Fund multiple without the best company = $22M ÷ $24M = 0.92x
Best company's share of the value      = $50M ÷ $72M = 69%

The average hides the typical result

In a power law, the average and the median tell very different stories. The median is the middle value when you put all values in order. The average adds all the values and divides by the count, so one very large value can move it a lot.

Five seed rounds:  $2M, $2.5M, $3M, $3.5M, $100M
Median           = $3M
Average          = $111M ÷ 5 = $22.2M

Four of the five rounds are near $3M, but the average is $22.2M. The same thing happens to funds. A few funds find a huge winner, and those funds pull the average return up. The typical fund, the median fund, does much worse than the average. When an investor says that “venture returns 3x on average,” ask for the median too.

Run the fund yourself

The widget runs a seed fund with $1M in each company. Each company gets an outcome at random, with the bucket shares from the Correlation data. Press “Run again” several times. Then change the number of companies and compare the 1,000 funds at the bottom.

Simulator

Run a seed fund

The fund puts $1M into each company. Run it again and again. Watch which companies pay for the fund.

Companies in the fund

102550

Run again

Fund 1: the multiple of each company, from best to worst

Returns the fund: 25x

Fund multiple

1.83x

$45.81M back on $25M

Best company

27.2x

59% of all returns

Without the best

0.78x

the other 24 companies

Lost money

20 of 25

returned less than 1x

In fund 1, 20 of 25 companies returned less than their cost. The best company returned 27.2x and brought in 59% of all the money. Without it, the fund returns 0.78x, a loss.

One company returned the whole fund alone, because it returned 25x or more.

1,000 funds of 25 companies each

Less than 1x

6.9%

1x to 2x

44.5%

2x to 3x

23.9%

3x to 5x

14.1%

5x or more

10.6%

Median fund

1.96x

Average fund

3.03x

Funds that lost money

6.9%

With 25 companies, the median fund returns 1.96x, but the average fund returns 3.03x. A few very large winners pull the average up. 6.9% of the funds lose money.

Select 10 or 50 companies. More companies make a loss less likely, but they do not raise the average.

Several patterns appear after a few runs:

• In most funds, more than half of the companies lose money.
• The best company frequently brings in a third or more of all the returns. In the first fund, it brings in 59%.
• Many funds have no fund returner. Those funds usually end with a low multiple.
• The average fund does much better than the median fund, because a few funds find a huge winner.

The last pattern is the most important one. Most funds do worse than the average fund, because the average includes the few funds with a huge winner.

More companies, fewer funds that lose money

Select 10 companies, then 50. With 10 companies, approximately one fund in five loses money, because many small funds never find a large winner. With 50 companies, almost no fund loses money. The average stays approximately the same, but the spread becomes much smaller.

This is the argument for a broad portfolio at seed. In 2019, AngelList studied seed investments that did not lose value and found a very extreme power law. The authors said that a seed investor must invest broadly, in all the credible deals that it can get. They also said that this result is not true at later stages.

Do later-stage investors face the same power law?

Less so. The AngelList study found the very extreme power law in seed investments, and its authors said that the result is not true at later stages. By Series C, investors want large and predictable revenue before they invest, so they buy into companies with a known business.

Work it out

A fund puts $1M into each of 25 companies and gets back $60M. Its best company returned $40M of that. What multiple does the fund return without its best company?

Your answerCheck

The best funds lose more frequently

Many people think that the best venture funds pick fewer losers. The data shows the opposite. In the Horsley Bridge data, the best funds had more deals with a loss than the good funds. The best funds won because their largest wins were larger. In 2015, Andreessen Horowitz called this pattern the “Babe Ruth effect.”

For a VC, a loss costs at most 1x the money invested. A winner can return 100x or more. Thus a VC who stays away from risky companies also stays away from the few companies that make a fund.

What the power law means for you

The power law explains much of what VCs ask from founders:

• VCs ask about the size of the market, because a company in a small market cannot become a fund returner. In 2016, the Y Combinator seed guide told founders to show a market of more than $1B if possible.
• VCs push for fast growth and large plans. A plan that makes a good $50M company is not a plan that can return a fund.
• A sale that is very good for the founders can be too small for the fund. The previous lesson showed that a $150M sale gives a $500M fund only 3% of its money back, at 10% ownership.

That logic is good for a fund with many companies. A founder has only one company, and the founder’s choices are different. A founder can prefer a likely $100M sale to a small chance at $1B. The fund, with 25 other companies, frequently prefers the chance at $1B.

Work it out

A seed fund puts the same check into each of 40 companies. What multiple must one company return to give back the whole fund by itself?

Your answerCheck

Quiz · question 1 of 6

Test what you learned

Correlation Ventures studied more than 21,000 financings from 2004 to 2013. Approximately what share returned less than 1x?

• AApproximately 95%
• BApproximately 10%
• CApproximately 33%, the "one third fail" rule of thumb
• DApproximately 65%