Most people who run an angel group, an accelerator or a small fund carry the same working assumption: a company that arrives through someone you trust is a better bet than one that arrives through the application form. The assumption is usually right about conversion. Referred deals get further through the process. Whether they turn out better, and what the referral filter quietly removes on its way through, are separate questions, and the research on them is more interesting than the assumption.

Where Closed Deals Come From

The most cited figures come from Gompers, Gornall, Kaplan and Strebulaev's survey of 885 institutional venture capitalists at 681 firms, run between November 2015 and March 2016 and published in the Journal of Financial Economics.¹ Their sourcing table reports the percentage of deals closed in the previous 12 months that originated from each source, based on the 446 respondents who answered that question. Professional networks accounted for 31 percent, proactively self-generated deals for 28, referrals from other investors for 20, inbound approaches from company management for 10, referrals from portfolio companies for 8 and quantitative sourcing for 2.¹ In the authors' words, "few VC investments, therefore, come from entrepreneurs who beat a path to the VC's door without any connection."¹ The NBER's write-up of the working paper reports the same pattern in rounded figures.²

Two details in that table are easy to miss. The first is what it measures: shares of closed deals, not conversion rates. It tells you that one closed deal in ten started as an inbound approach, and nothing about how many inbound approaches it took to produce that one. A channel can be a large share of the pipeline and a small share of closed deals, and this table cannot distinguish that from a channel that is small at both ends.

The second is the performance split. The authors divided firms with at least ten exits in the past decade into those above and below the median IPO rate. The sourcing mix of the two groups is almost the same: inbound was 11 percent at the high-IPO firms and 10 at the others, referrals from other investors 21 and 20, professional networks 30 and 33.¹ They conclude that "the type of the sources is less important than sometimes claimed," and suggest that what separates the better firms may be "the quality of their referral network" rather than the channel itself.¹ For an operator, knowing which referrers are good may matter more than having referrals at all.

Conversion Is One Question, Outcomes Are Another

Pipeline data that includes rejected companies is rare, because most investors never track the deals they passed on. The closest thing available is a 2025 NBER working paper by Young Soo Jang and Steven Kaplan that uses the full deal records of one early-stage venture firm investing in pre-seed and seed companies in the US Midwest.³ The firm logged 8,815 deals between August 2015 and December 2021, scored 366 of them and invested in 114, and the authors followed every company, funded or not, in PitchBook through March 2024.³

Of the 6,581 deals with a recorded source, 58 percent were outbound (found by the firm itself, including at accelerators and events), 25 percent inbound and 17 percent referred by co-investors, meaning other venture firms considering the same company.³ The definitions matter here, because the paper's "inbound" category mixes cold emails with referrals from the firm's personal network, its LPs and service providers. Only the co-investor channel is a clean referral category.

Conversion followed the expected order. The firm scored 13.5 percent of co-investor deals, 5.3 percent of inbound deals and 2.7 percent of outbound ones.³ The later outcomes, which cover all sourced companies including those the firm rejected, also favoured the referrals: 50 percent of co-investor deals went on to raise at least 1 million dollars of venture funding, against 29 percent of inbound and 28 percent of outbound, and 16 percent raised at least 10 million against 8 percent for both other channels.³

Read that carefully before treating it as proof that referrals find better companies. A co-investor referral is by definition a company another venture investor is already thinking of funding, so a higher rate of later venture funding is partly built into the category. Once the firm's own scores were included, deal source did not significantly predict whether it invested, and among those scored deals, co-investor referrals were significantly less likely to go on to raise 50 million dollars or more.³ Across all sourced deals, about 2 percent of each channel reached that level.³ In this one firm's data, the referral channel looks better at producing companies that clear the next funding bar and no better at producing the outliers.

There is a volume point too. As illustrative arithmetic of our own from the paper's shares (93 percent of scored deals had a source): about 30 percent of roughly 340 sourced, scored deals is around 100 outbound companies, against roughly 150 from co-investors.³ The channel with the lowest conversion rate still supplied a large part of the shortlist, because it supplied most of the funnel.

What a Warm Introduction Carries

It would be wrong to treat warm introductions as pure bias. Shane and Cable's study of 202 seed-stage investors and 50 high-technology ventures found that direct and indirect ties shaped which ventures were funded "through a process of information transfer."⁴ A referral from someone who has worked with a founder carries information an application cannot: how the person behaves under pressure, whether they finish what they start.

Social proximity can also help after the cheque. Hegde and Tumlinson studied almost all US venture deals from 1991 to 2010, covering about 22,000 venture partners and 85,000 startup executives.⁵ A single co-ethnic pairing between a VC partner and a company executive was associated with a 16 percent higher probability that the VC invested, and co-ethnic investments had a higher probability of an IPO or acquisition, which the authors attribute to better communication and coordination between investor and founder after the investment.⁵ ⁶ So a network that tries to remove every personal connection from its process could lose information it was using.

The same body of research shows the cost side. Gompers, Mukharlyamov and Xuan found that venture capitalists strongly prefer to co-invest with people who share their background, and that this kind of affinity hurts results: the probability of success fell by 17 percent when two co-investors had worked at the same company, by 19 percent when they shared an undergraduate school and by 20 percent when they belonged to the same ethnic minority group.⁷ ⁸ The authors trace the penalty mainly to poor decision-making in high-affinity syndicates after the investment rather than to weaker deal selection.⁷ Referrals between investors tend to travel along the same lines of similarity.

What the Warm Introduction Filters Out

The cleanest evidence on who gets into the referral stream comes from Howell and Nanda, who used Harvard Business School's New Venture Competition between 2000 and 2015.⁹ Participants were assigned to judging panels with different numbers of venture capitalists on them, and that variation was effectively random. For men, each additional VC judge increased the likelihood of later starting a VC-backed company by 25 percent relative to the effect among women.¹⁰ Women saw no benefit from exposure to male or female VCs.⁹ The authors read this as evidence that "access to investors impacts fundraising independent of the quality of ideas."⁹

The authors' follow-up survey points to why. In it, male participants were nearly twice as likely as women to proactively reach out to VCs after the competition (81 percent more likely, in the Harvard Kennedy School summary), and men and women reported no difference in how often VCs responded to that outreach.¹⁰ ¹¹ The filter was on who sent the follow-up email, a trait of the founder rather than the company. A network whose second meetings come mostly from founders who chase the judges afterwards is selecting partly on that.

Compare the cold channel. Gornall and Strebulaev sent 80,000 pitch emails about fictitious but promising startups to 28,000 venture capitalists and angels, randomly varying the founder's apparent gender and race.¹² Female founders received 9 percent more interested replies than men pitching identical projects, and Asian founders 6 percent more than White founders.¹² The authors conclude that investors did not discriminate against those groups when evaluating unsolicited emails, and suggest that research on investor bias look at networks and in-person interactions instead.¹² Put next to Howell and Nanda, the evidence is consistent with this reading: at least at the first-reply stage, the measured disadvantage sits more in who reaches the investor than in how a written pitch is judged.

What Open Applications Filter Out

Open applications have their own costs. They bring volume, and volume costs screening time. In Jang and Kaplan's data the inbound channel converted at well under half the rate of co-investor referrals.³ An application form also selects on its own traits: knowing the network exists, writing well in the network's language, and having time to fill in the form. Founders outside the operator's circles may never see the call, so "open" is relative to who hears about it. And a form carries none of a referrer's direct experience of the founder.

So each channel has a characteristic blind spot. Referrals tend to under-represent founders who are less connected or less inclined to chase investors, and they carry homophily. Open applications tend to under-represent founders who never hear about the call, and they arrive without a referrer's context. The research does not support saying either produces better portfolios.

What a Network Operator Can Do

The most useful step is also the cheapest: record the source of every deal at the moment it enters the pipeline, with categories precise enough to separate the cases that matter. Jang and Kaplan's mixed "inbound" bucket shows why.³ An operator can do better with a short fixed list (member referral, with the member named; co-investor; portfolio founder; event or outbound; open application; cold email) and a rule that the field cannot be left blank.

Then track each source through every stage: screened, met, diligence, invested, and what happened afterwards, including follow-on funding and survival. Jang and Kaplan could follow the companies the firm turned down; most networks never look, which means they cannot see whether a channel's rejects went on to do well.³

Look at referrers one by one as well as by channel. If Gompers and coauthors are right that the quality of a referral network differs more than its type, the per-person hit rate is the number that tells you whom to listen to.¹

Run referred deals through the same structured application as everyone else. The introduction can buy a company a faster read; it should not substitute for the fields every other company fills in.

Howell and Nanda's result also suggests a design change: if access depends on who follows up, build the follow-up into the process. Their own conclusion points towards "encouraging and potentially formalising networking opportunities."¹⁰ A pitch event where every participant is offered a slot with an investor afterwards, instead of waiting to see who writes in, removes one selection step that has little to do with the company.

One caveat: an angel group closing ten or fifteen deals a year will not reach statistically meaningful source-to-outcome comparisons for years, so treat the numbers as descriptive. They are still better than an impression, and they are the only evidence specific to the network's own members.

Whatever tool a network uses, Pynn included, the source-to-outcome view only exists if source and referrer are recorded on every deal from the day it arrives; nobody can reconstruct them reliably two years later.

Sources

1. Journal of Financial Economics, How Do Venture Capitalists Make Decisions?, Paul A. Gompers, Will Gornall, Steven N. Kaplan and Ilya A. Strebulaev, 2020 (NBER Working Paper 22587, September 2016). https://www.nber.org/papers/w22587

2. NBER Digest, How Do Venture Capitalists Make Decisions?, National Bureau of Economic Research, December 2016. https://www.nber.org/digest/dec16/how-do-venture-capitalists-make-decisions

3. NBER Working Paper 33483, Venture Capital Start-up Selection, Young Soo Jang and Steven N. Kaplan, February 2025. https://www.nber.org/system/files/working_papers/w33483/w33483.pdf

4. Management Science 48(3), Network Ties, Reputation, and the Financing of New Ventures, Scott Shane and Daniel Cable, March 2002. https://ideas.repec.org/a/inm/ormnsc/v48y2002i3p364-381.html

5. ifo DICE Report, Ethnicity and Venture Capital (summary of Does Social Proximity Enhance Business Partnerships? Theory and Evidence from Ethnicity's Role in U.S. Venture Capital, Management Science, 2014), Justin Tumlinson and Deepak Hegde, September 2016. https://www.ifo.de/DocDL/dice-report-2016-3-tumlinson-hegde-september.pdf

6. NYU Stern, Venture Capitalists Do Better Investing in Start-Ups Run by Entrepreneurs of Same Ethnicity, NYU Stern School of Business, February 2014. https://www.stern.nyu.edu/experience-stern/faculty-research/hegde-venture-capitalists

7. Journal of Financial Economics 119(3), The Cost of Friendship, Paul Gompers, Vladimir Mukharlyamov and Yuhai Xuan, 2016 (NBER Working Paper 18141). https://www.nber.org/papers/w18141

8. HBS Working Knowledge, In Venture Capital, Birds of a Feather Lose Money Together, Carmen Nobel, June 2014. https://www.library.hbs.edu/working-knowledge/in-venture-capital-birds-of-a-feather-lose-money-together

9. Journal of Financial and Quantitative Analysis 59(6), Networking Frictions in Venture Capital, and the Gender Gap in Entrepreneurship, Sabrina T. Howell and Ramana Nanda, September 2024 (online June 2023). https://www.cambridge.org/core/journals/journal-of-financial-and-quantitative-analysis/article/abs/networking-frictions-in-venture-capital-and-the-gender-gap-in-entrepreneurship/AA981C3615556F057D1B6D69FEFA1088

10. VoxEU (CEPR), Networking Frictions in Venture Capital and the Gender Gap in Entrepreneurship, Ramana Nanda and Sabrina Howell, 29 February 2020. https://cepr.org/voxeu/columns/networking-frictions-venture-capital-and-gender-gap-entrepreneurship

11. Harvard Kennedy School Gender Action Portal, Networking Frictions in Venture Capital and the Gender Gap in Entrepreneurship (research summary), Harvard Kennedy School. https://www.gap.hks.harvard.edu/networking-frictions-venture-capital-and-gender-gap-entrepreneurship

12. Management Science 71(6), Gender, Race, and Entrepreneurship: A Randomized Field Experiment on Venture Capitalists and Angels, Will Gornall and Ilya A. Strebulaev, June 2025. https://ideas.repec.org/a/inm/ormnsc/v71y2025i6p5308-5327.html

Sourcing

Warm Introductions Versus Open Applications: What Sourcing Data Shows