Fundamentals: why a portfolio approach matters
Most venture deals lose money. So why does anyone invest in them?
Because the few that succeed can make up for all the others. A startup that does well can pay back 10, 20 or even 50 times what was put in. Nobody, including experienced investors, can reliably tell in advance which company that will be.
That's a problem if you only make one or two investments, because you'll probably miss the big winner. In the historical data behind this page, about 6 in every 100 startup investments returned 10 times the money or more:
Put everything into 1 company: about a 6% chance of owning one of those big winners.
Split the same amount across 20 companies: about a 7 in 10 chance of owning one.
Portfolio Simulation Tool
Drag the slider from one to forty and watch how much less your result depends on luck.
What 10,000 portfolio simulations show - based on 50,000+ VC financings exiting 2004-2022
Each bar is how often the 10,000 simulations landed on that result. The height scale is set by the tallest bar each time, so it changes as you drag: compare bars within one view, not across a drag. Drag the slider above and watch the results crowd together.
The five figures above the chart are about the money only. They leave out fees, taxes and the manager's share, and they ignore how long the money is tied up: doubling your money is only about 8% a year if it takes nine years to get there.
There is a guess inside this tool too. The data powering it covers more than 50,000 venture capital financings that exited between 2004 and 2022, the most recent complete exit year, but it only shows us that investments ended up somewhere in a band (Below 1x, 1–5x, 5–10x and so on), not exactly where in that band. So we had to choose, and we chose conservatively: we assume results bunch up near the bottom of each range. For more information on all of the assumptions that went into the simulator, please scroll further down the page.
Where the Numbers Come From
Correlation Ventures, a venture firm with one of the largest databases of deal outcomes, has published how U.S. venture financings actually turned out, for companies that exited in each of two overlapping decades. Each row below is the share of those financings that ended in each bucket, before fees.
| Years | Below 1x | 1–5x | 5–10x | 10–20x | Above 20x | Status |
|---|---|---|---|---|---|---|
| All years | 54.7% | 32.1% | 7.3% | 3.5% | 2.5% | Blend of the two published decades |
| 2004–2013 | 65% | 25% | 6% | 2.5% | 1.5% | Published (Correlation Ventures via Seth Levine, 2014) |
| 2013–2022 | 48% | 36% | 8% | 4% | 3% | Published (Correlation Ventures, 2023) |
"All years" is Red Bear's blend of the two published decades, weighted by sample size: by how many financings each covered (about 21,000 and 35,000). The two decades are not one continuous run. They overlap in 2013, and a financing that exited that year can sit in both, so the blend is a weighted average of two published datasets rather than a single 19-year dataset. Each decade is scaled to 100% before blending, because the published figures are rounded and the 2013 to 2022 row sums to 99; blending the rows exactly as printed lands about a third of a point low on every bucket. The blend is then rounded to one decimal, which is why that row reads 100.1. Weighting the two decades equally instead moves the buckets by up to about two points. 2022 is the most recent complete exit year, so exits after it are not counted. A financing whose company has not ended yet, however long ago it was funded, is in neither decade. How the simulator turns these five buckets into individual deal outcomes is set out under the assumptions below.
Check Our Math
Nothing here is hidden. The table recomputes live from the same engine that draws the chart.
| Check (All years, 1 investment) | Expected | Simulator | Result | |
|---|---|---|---|---|
| Average per investment, formula | 2.49x | Average per investment, simulated | 2.54x | PASS |
| Average per investment | 2.49x | Average of 1-investment portfolios | 2.54x | PASS |
| Share of investments below 1x, source | 55% | Share simulated | 54% | PASS |
| Share of investments 10x or more, source | 6% | Share simulated | 6% | PASS |
| Share of investments below 0.5x, formula | 44% | Share simulated | 43% | PASS |
| Share of investments 2x or more, formula | 25% | Share simulated | 26% | PASS |
| At least one 10x in 1 investment, by hand | 6% | Simulated | 6% | PASS |
"At least one 10x" is the chance a portfolio contains an investment that returned ten times its cost or more. You can check it by hand: 1 minus (1 minus the 10x rate) to the power of the number of investments. Every figure in the left-hand column is a formula or a published source figure, never a second reading of the simulation.
Change the Years of History
Behind the tool
What spreading your money out does, and what it doesn't
It doesn't raise your average result. The average is what you would end up with if you could make the same investments over and over, and in real life you only get one go. With one investment, that one go will probably land well below the average, with a small chance of landing far above it. With forty, it usually lands close to the average. Watch the average as you drag the slider: it barely moves. What shrinks is the range of results you could get.
It doesn't make any single investment safer. Each company is just as likely to fail as before.
Owning a winner is not the same as the portfolio returning that much. Split 20 ways, a company that returns 10 times its money hands you back half your total, no more. The other nineteen still decide where you end up.
It does make your result depend much less on luck. You're less likely to lose everything, and also less likely to land a once-in-a-lifetime jackpot.
Think of flipping a coin. Flip it once and you get either all heads or no heads. Flip it 100 times and you'll almost certainly land near half heads. More flips don't change the odds of any single flip. They just make the overall result more predictable.
One caveat: the odds on this page come from real startup investments made from 2004 to 2022. That period includes the 2008 financial crisis and the 2020–21 boom, so the odds reflect good times and bad. But the simulator picks each company separately, as if it were its own coin flip. In real life, if you invest in several startups around the same time, they face the same economy: a recession or a slow funding market tends to hit them all at once. So spreading your money out helps somewhat less in practice than the simulator shows.
To see how much timing matters, use Change the Years of History above to compare the earlier decade (2004–2013) with the recent one (2013–2022). In the earlier decade, about 65% of investments lost money. In the recent decade, it was about 48%.
What it does not show
The future. Every number here is history: U.S. venture deals that already finished, one way or another. Later deals, different stages and different sectors will have a different shape.
Your deals. The outcomes come from industry-wide data, not from any Red Bear Ventures investment.
Fees. Everything is before fees, taxes, and carry, which is the share of the profit the manager keeps.
Timing. Multiples only. How long the money is tied up is a separate question the return calculator handles.
Marked-up values. Recent portfolios that have not finished yet often look much better on paper than deals that actually ended, because losers take years to shut down. This tool only uses finished deals, which is why it stops at 2022: the years since are too young to chart. Early signs (PitchBook, Carta) suggest they have been harder than the historical average.
The assumptions the simulator makes
- Money invested
- 100 unitsSplit equally every time. One investment gets all 100; twenty investments get 5 each. The total never changes, so the only variable is how many ways it is divided.
- Portfolios simulated
- 10,000For each combination of settings, from a fixed starting point. The same choices always produce the same numbers, so anything you see can be reproduced.
- Outcome data
- Exits, 2004 to 2022U.S. venture financings whose company exited, by acquisition, listing or shutdown, inside the selected window, grouped into the five buckets in the table above. The years are exit years, not the years the money went in, so a financing still outstanding is in no bucket at all however old it is. Those bucket shares are published figures, not modelled or forecast. What the simulator adds is the shape inside each bucket, which the next two rows set out. The unit is a financing, not a company: a company that raised three rounds appears three times. The simulator uses those financing outcomes as a stand-in for the spread of individual investment outcomes, which is the closest published data there is.
- Within each bucket
- Heavy-tailed (1x to 20x)The published data gives bands, not individual results. For the three buckets between 1x and 20x, the simulator assumes outcomes bunch toward the bottom of each band, with a result twice as big being four times rarer. That puts about 62 out of every 100 results in the 1x to 5x band below 2x. It is a judgement rather than a published figure, and it is the largest single assumption on this page. It is deliberately cautious but not the most cautious possible: on a single investment a flatter assumption would put the doubling figure nearer 37%, and a more sharply bunched one nearer 20%. Even if every result were the worst it could be for its band, the average across all five would still be about 1.6x. The below-1x bucket and the open-ended top bucket are each generated differently, which the next two rows set out.
- Losing deals
- 6 in 10 of that bucketSix of every ten investments in the below-1x bucket return nothing at all. The other four are spread evenly between 0x and 1x on a linear scale, rather than the bunched shape used for the bands above: that shape cannot reach zero, so it cannot describe this bucket. Together that puts the average below-1x outcome at 0.2x. That bucket is more than half of all draws and the four above it carry the rest, which is how the expected multiple across all five works out at about 2.5x.
- The top bucket
- No ceilingThe published top bucket is open-ended: above 20 times cost, with no maximum given. Until October 2026 the simulator closed it at 50 times, which was tidy and wrong, because it made twenty years of venture look as though nothing ever beat 50x. There is no ceiling now. Instead this bucket bunches harder toward 20x than the bands below it, a result twice as big being sixteen times rarer rather than four times, and that is what lets it run upward and still average about 30 times cost, which is what the old ceiling produced. The range of portfolio results still narrows roughly threefold between one investment and forty. About 1 in 600 investments now lands above 50 times cost and about 1 in 5,000 above 100 times. It stays too thin at the very top: returns above 1,000 times cost did happen in this period and effectively never happen here, which understates the best outcomes on purpose.
- Each investment
- Drawn separatelyEvery investment is an independent draw from history, with no link to the others. Real investments move together more than that, so this overstates the benefit of spreading. Tested: if every investment in a portfolio shares a single vintage, the chance of ending below your money across forty investments rises from about 1.3% to about 3.7%.
Glossary
- Deal, or financing
- One investment into one company at one point in time. The historical data counts financings, so a company that raised money three times appears three times.
- Multiple
- How many times the money came back. 1x is break-even, exactly what went in. 0.5x means half came back; 10x means ten times. Every figure on this page is a multiple.
- Portfolio
- The whole set of companies the money is spread across, counted as one combined result rather than company by company.
- Diversification
- Spreading money across many separate investments instead of a few, so that no single outcome decides the result. It is the thing this page is about.
- Total loss
- A deal that returns nothing at all, not even part of the money. It is the most common single outcome in early-stage investing.
- Median
- The middle result: half the simulated portfolios did better and half did worse. It is not the average, and for venture outcomes the two sit far apart, because a few very large winners pull the average up.
- Exit
- The end of an investment, when the company is sold, goes public, or shuts down. Until one of those happens the result is not yet known.
- Marked-up value
- What an investment is said to be worth on paper before it has exited, usually set by the price of a later funding round. It is an estimate, not money received.
- Carry
- The share of the profit the manager keeps. It applies to the gain only, never to the capital returned: on 150 coming back from 100 invested, carry is charged on the 50.
- Illiquid
- Money that cannot be taken back out on demand. An early-stage investment is typically locked up for years, until an exit happens or the company fails.
Important. This is an educational illustration of a statistical idea, that the range of outcomes narrows as you spread money across more independent bets. It is not investment advice, not a forecast, and not a description of any Red Bear Ventures investment, fund or vehicle. The outcome distributions are third-party historical industry data (Correlation Ventures, U.S. venture financings exiting 2004–2022) and are not derived from, and do not describe, any Red Bear Ventures investment. The "All years" view blends two Correlation Ventures datasets weighted by sample size. Data ends in 2022, the most recent year with complete finished-deal outcomes; results for later years are not yet known. Simulated results assume independent outcomes and a modelled, not observed, spread of results inside each published bucket; actual portfolios can be more concentrated in both directions. Past results do not predict future results. Early-stage investments are illiquid and can lose all of their value. Figures are before fees, carry and taxes. How many investments are right for any person depends on their circumstances; see the education guide's section on allocation.
Sources: Correlation Ventures, "We're Still Not Normal" (2023) · Seth Levine's reproduction of the 2004–2013 chart (2014) · Bruce Booth on the 2013 Correlation data · AngelList, what a failed investment returns (2022) · PitchBook on exits since 2022