Anyone who has run an angel network for a while has heard the advice given to new members: build a portfolio of 20 companies, or 15, or 25, before judging whether this works. The number moves around a little depending on who is saying it, but it stays in the same band, and it is usually delivered as received wisdom rather than as something with a derivation behind it.
It does have a derivation, and it is not complicated. Early-stage returns are concentrated in a small number of outcomes, which means the question a small portfolio faces is less "how good were my picks on average" and more "did I happen to hold one of the few companies that produced most of the money". Once you frame it that way, the band where the advice lands stops looking arbitrary. It also becomes clear what the advice assumes, and where it breaks.
What the Angel Data Shows, and How Old It Is
The study most often cited here is Robert Wiltbank and Warren Boeker's Returns to Angel Investors in Groups, published by the Kauffman Foundation and the Angel Capital Education Foundation in November 2007. It surveyed members of 86 North American angel groups; 13 percent responded, giving 539 investors who reported 1,137 exits, where an exit meant an acquisition or IPO that returned money or a closure that lost it.¹ Those exits came from 3,097 reported investments, so most of the holdings in the dataset had not resolved when it was collected.²
The headline figure was an average of 2.6 times the money invested over a 3.5 year hold, roughly a 27 percent IRR, with the multiple calculated cash-on-cash per venture. The distribution is the part that matters for portfolio construction. Fifty-two percent of the exits returned less than the capital invested. Seven percent returned more than ten times the money, and those accounted for 75 percent of the total cash returned.¹
Two things about that study need saying before its numbers are used for anything. First, it is old. Only 8 percent of its exits happened before 2000 and 62 percent after 2004, which puts the bulk of the evidence in the three years or so before the report came out in late 2007, nearly two decades ago.¹ Second, the authors flag survivor bias (angels who lost money and left their groups are missing) and self-selection bias (people with good results are more willing to share them). They found no significant difference in returns between groups with high and low response rates, which softens the second concern without removing it.¹ ²
Later data from the same research line points in the same direction, with a heavier tail. Wiltbank's 2016 Tracking Angel Returns study, published by the Angel Resource Institute, reported 2.5 times over 4.5 years, about a 22 percent IRR, with 70 percent of outcomes below 1x and 10 percent of exits providing 85 percent of the total return.³ Tech Coast Angels' own data, published by the Angel Capital Association in 2023 in a post by John Harbison, covered 247 outcomes between 1997 and 2022 and found that eight companies, 3 percent of the outcomes, produced 77 percent of the dollars returned.⁴ The levels differ from study to study. The shape, where a small share of outcomes carries most of the money, has held in every version we could find.
One line in the 2007 report makes the portfolio argument on its own. Only 48 percent of exits returned at least 1x, but 61 percent of investors had an overall multiple of at least 1x, which the authors describe as showing the advantage of making multiple investments.¹
The Arithmetic of Missing the Outlier
The reason small portfolios struggle follows from a simple probability. If a given investment has a chance p of becoming a large outlier, and outcomes are independent, the chance that a portfolio of n companies contains at least one is 1 minus (1 minus p) raised to the power n.
The figures below are our own illustrative calculation, not results from any study. We set p at 7 percent, borrowing the 2007 share of exits above 10x, and treat each company as an independent draw. Both assumptions are generous in one direction and harsh in another: 7 percent is a share of exits rather than of all investments, and startups in the same network, sector and vintage do not fail independently.
At 7 percent, a portfolio of five companies has about a 30 percent chance of holding at least one 10x outcome. Ten companies gets to about 52 percent, roughly a coin flip. Fifteen reaches about 66 percent, 20 about 77 percent, and 25 about 84 percent. Beyond that the gains shrink: 30 companies gets to 89 percent and 40 to 95 percent. If the true rate is closer to 5 percent, the curve shifts right, and 20 companies gives about 64 percent while 25 gives about 72 percent.
That is where 15 to 25 comes from. Below 15, each additional company buys a large increase in the odds of catching the tail, and a member with five or eight companies is mostly measuring luck. Somewhere between 15 and 25 the curve bends, and each additional company buys less. The band is the knee of the curve rather than a threshold where something switches on, and its position depends entirely on how often your deal flow produces outliers.
Older guidance sat lower. A 2004 Angel Capital Association best-practice note by Bill Payne said many felt 10 to 12 angel deals was adequate diversification.⁵ AngelList's own education material now puts the marker higher, noting that the majority of funds on its platform with at least 20 investments have a TVPI greater than 1x.⁶
What the Larger Datasets Add
AngelList's research team has published the most direct tests of this, and it is worth being specific about which paper says what, because they get blended together.
In a December 2019 paper, Startup Growth and Venture Returns, Abraham Othman modelled seed returns as a power law and argued that "investors increase their expected return by indexing as broadly as possible at the seed stage (i.e., by putting money into every credible deal)". He limits the claim to seed, and says the case for selectivity grows at later rounds.⁷
In April 2020, Nigel Koh and Othman looked at 10,665 limited partner portfolios on AngelList in How Portfolio Size Affects Early-Stage Venture Returns. Median annual IRR rose by 9.0 basis points for each additional investment, and mean IRR by 6.9 basis points, so a typical 100-investment portfolio returned almost 9 percentage points a year more than a single-investment one.⁸ ⁹ They are candid about the limits: good investors are more likely both to make many investments and to pick well, and portfolio size explains only part of the difference between investors, with r-squared values of 0.23 and 0.18.⁸
A June 2020 AngelList post by Othman, drawing on 1,808 pre-Series C platform investments, simulated managers making ten investments each. The typical manager, it found, "fails to pick any outsize winners in their 10 chances, whereas the market portfolio is assured of selecting all of the return-driving winners".¹⁰
Correlation Ventures, working from its own database of US venture financings since 2000, reached a subtly different conclusion. In a 2022 Monte Carlo analysis, David Coats found that expected returns stayed the same across portfolio sizes and only the spread changed. An eight-company portfolio had a 28 percent chance of returning less than 1x and a 10 percent chance of returning more than 5x; at 100 companies those fell to 6 percent and 2 percent.¹¹
The disagreement matters less than it looks. In a random draw from a fixed distribution, adding companies does not change the average outcome; it moves the typical outcome toward the average. For a power-law distribution, where the median sits far below the mean, that is most of what an individual investor experiences. The AngelList empirical result, where the mean also rose, is consistent with the selection effect its own authors describe.
The Honest Counterpoint
Diversification only helps if the pool you are diversifying across contains the outliers. Othman's recommendation is to index every credible deal, and the qualifier carries the weight. Twenty investments drawn from a deal flow that rarely produces a 10x outcome will reliably deliver the average of that deal flow, which may be below 1x. Spreading a member's money thinly across weak deals makes the loss more predictable without making it smaller.
There is also a direct tension with depth. The 2007 study found a median of 20 hours of due diligence per investment, and investors above that median had an overall multiple of 5.9x against 1.1x for those below it.¹ That is an association, not proof that diligence caused the gap, but it points to a trade-off: a member trying to reach 25 companies in a few years has less time for each one, and a network pushing volume can dilute the diligence that makes its deal flow worth diversifying across.
The concentration argument is legitimate on its own terms. Correlation's numbers show that a small portfolio raises the chance of a very large overall multiple along with the chance of a loss.¹¹ An investor who has an edge in picking, and can bear the variance, can rationally choose that. Most individual angels have no evidence yet about whether they have that edge, which is part of why the diversified approach is the default recommendation. The market is also moving the other way: the ACA's 2026 Angel Funders Report summary says angel groups are writing larger cheques while funding fewer companies.¹²
What This Means for Running a Network
The practical problem for an operator is pacing. A member who writes two or three cheques a year needs seven to ten years to reach 20 companies, which means most of your newer members spend their first several years in the zone where results are mostly noise. That affects retention as much as returns: a member who has lost money on their first five investments has seen an ordinary outcome for a five-company portfolio, but may not know it.
Several design choices follow. Ticket minimums set how many companies a member can afford to reach, and a lower minimum on syndicated deals can be the difference between a member building a portfolio and a member making a few large bets. A sidecar or pooled vehicle gives members exposure across every deal the network backs, which is the closest a network can get to the "every credible deal" approach Othman describes. Encouraging members to spread their commitments across several years also protects them against one bad vintage.
Follow-on reserves deserve their own treatment, and we will come back to them separately. Briefly: Payne's 2004 note said prudent angels reserve 50 to 100 percent of invested capital for follow-ons, and AngelList notes that many fund managers reserve up to half their fund, though smaller funds often keep more for initial cheques to maximise the count.⁵ ⁶ The 2007 study found that follow-on investments were associated with lower returns, which is a reason to treat reserves as a decision made deal by deal rather than an automatic top-up.¹
The number an operator most needs to know, and usually does not, is where each member sits on the curve. A network where most members hold four companies behaves very differently from one where most hold 20, even if the deal flow is identical. Pynn is built partly so that count is visible per member, alongside the deal flow it came from, instead of sitting in each investor's own spreadsheet.
Sources
1. Robert Wiltbank and Warren Boeker, Returns to Angel Investors in Groups, Ewing Marion Kauffman Foundation and Angel Capital Education Foundation, November 2007. https://www.angelcapitalassociation.org/data/Documents/Resources/AngelGroupResarch/1d%20-%20Resources%20-%20Research/6%20RSCH_-_ACEF_-_Returns_to_Angel_Investor_in_Groups.pdf
2. David Teten, A Closer Look At The Quality Of Angel Returns Data, TechCrunch, 26 January 2013. https://techcrunch.com/2013/01/26/a-closer-look-at-the-quality-of-angel-returns-data/
3. UK Business Angels Association, Latest US Study Tracking Angel Returns presented at ACA Summit Philadelphia (reporting Robert Wiltbank, Tracking Angel Returns 2016, Angel Resource Institute), 16 May 2016. https://ukbaa.org.uk/blog/2016/05/16/latest-us-study-tracking-angel-returns-presented-at-aca-summit-philadelphia/
4. John Harbison, Angel Returns Beat All Asset Classes But Pose Greater Risk, Angel Capital Association, 25 May 2023. https://angelcapitalassociation.org/blog/angel-returns-beat-classes/
5. William H. Payne, Asset Allocation and Portfolio Strategy for Angel Investors, Angel Capital Association best practices, 14 February 2004. http://www.angelcapitalassociation.org/data/Documents/Members%20Only/BestPractices/E3c%20-%20Best%20Practices/4%20BestPractices_PortfolioStrategy.pdf
6. Matthew Speiser, Jess Toy and Kate Bridge, The Basics of Venture Capital Portfolio Construction and Management, AngelList Education Center, 20 January 2022, updated 13 May 2026. https://www.angellist.com/learn/portfolio-construction
7. Abraham Othman, Startup Growth and Venture Returns, AngelList, December 2019. https://angel.co/pdf/growth.pdf
8. Nigel Koh and Abraham Othman, How Portfolio Size Affects Early-Stage Venture Returns, AngelList, April 2020. https://angel.co/pdf/lp-performance.pdf
9. Kauffman Fellows, Seed portfolio performance increases with the number of investments, says AngelList, Kauffman Fellows Journal, 29 April 2020. https://www.kauffmanfellows.org/journal/seed-portfolio-performance-angellist
10. Abe Othman, What AngelList Data Says About Power-Law Returns In Venture Capital, AngelList, 17 June 2020. https://www.angellist.com/blog/what-angellist-data-says-about-power-law-returns-in-venture-capital
11. David Coats, Thiel vs. AngelList. Who is Right?, VC by the Numbers, Correlation Ventures, 28 April 2022. https://medium.com/correlation-ventures/thiel-vs-angellist-who-is-right-e5a4209ba8ec
12. Angel Capital Association, ACA Publishes 2026 Angel Funders Report, 13 July 2026. https://angelcapitalassociation.org/blog/aca-publishes-2026-angel-funders-report/