What a Network Owes a Startup It Rejects

The part of the funnel nobody designs

An investor network is, by volume, a rejection machine. The Angel Capital Association's own guidance for angel groups, published in 2007 and still the clearest public statement of how a screening funnel is meant to work, sets out the attrition in plain arithmetic: one in four deals that enter pre-screening make it to the screening phase, and one in three of those go on to due diligence. A group taking in thirty business plans a month will put two or three of them in front of its full membership.¹ Everything else is a no. The document predates most of the platforms and process tooling networks use today, so treat the specific ratios as a structural baseline rather than a current benchmark; nothing about the argument that follows depends on them being exact today.

Then look at where the process design sits. The same guidance describes what each group of companies is told. At pre-screening: "None of the angel groups we spoke with contact the entrepreneurs at this stage, as this is merely a process of elimination." After the screening committee has met: "Companies that have been declined are notified, but detailed feedback is usually not provided at this point." Only the two or three companies that present to the membership get the full treatment, and some groups do it well, collecting written comments from every member in attendance so the founder hears one consolidated view rather than nine contradictory ones.¹

Put those three sentences in order and the rule they describe is simple: the further a company travels, the more it is told. That is defensible on cost, since the people who reach the end are fewer and the network has already spent time on them. It is also upside down relative to volume. The overwhelming majority of the companies a network touches exit through a door that has had almost no thought put into it.

The applicants you turned down are part of your next cohort

Y Combinator, which processes application volume that no angel network will ever see, publishes a number that should give any selector pause. From its own FAQ: "In a typical YC batch, about half the companies applied multiple times before being accepted."²

Half of the batch, in other words, is made up of companies YC had already turned down at least once, which means the rejected pile was not a discard pile at all but a holding pen for a large share of the eventual yes list.

YC is also candid about why it does not explain those declines. Asked whether it gives feedback to rejected applicants, the answer is no unless the company was invited to interview, and the stated reason is volume: "If we did this, we'd spend all our time providing feedback and doing nothing else due to the volume of applications we have to process."²

Both of those facts are true at the same time, and the tension between them is the whole subject. The rejected pile is where a large share of future portfolio companies are sitting. It is also the pile that is most expensive to talk to. Every network resolves this somehow, usually by not deciding at all and letting the resolution emerge from whoever is clearing the inbox that week.

Feedback is an intervention, not a courtesy

Before arguing that networks should say more, it helps to be clear about what saying more does, because the evidence does not all point one way.

Sabrina Howell studied US new venture competitions covering 4,328 ventures. Some competitions disclosed each venture's rank relative to the others; some told founders only whether they had won. Comparing the two, negative feedback reduced the likelihood that a venture continued by roughly nine percentage points, measured against a mean continuation rate of 34 percent, with the effect concentrated in the first six months.³

That is a large behavioral effect from a small piece of information. It also comes with two findings that matter more for process design than the headline does. Founders discounted signals they read as imprecise: they responded less to feedback from a smaller panel of judges, which Howell reads as founders treating a thinner panel as a noisier one and weighting it accordingly. And negative feedback on product and technology scores had no significant effect on whether founders continued, which Howell reads as founders holding better private information about their own product than the judges do.³

Howell's own interpretation is that this feedback is close to costless and improves efficiency, because it lets weak ventures stop sooner. Read from the network's side, the conclusion is narrower and more uncomfortable. A reason attached to a rejection is not a nicety. It changes what the founder does next. Some of what it stops should be stopped. Some of it should not, and the network will never find out which.

The near-miss problem

There is a body of evidence on what happens to applicants who fall just below a funding line, and it comes from grant review rather than venture capital.

Yang Wang, Benjamin Jones and Dashun Wang compared junior scientists whose first NIH R01 application landed just below the funding threshold against those who landed just above it, 623 near-misses and 561 narrow-wins, using applications submitted between 1990 and 2005. The near-miss group had about 11 percent fewer people still active as principal investigators a year later, and that gap held through year seven. Among those who stayed in, the near-misses went on to produce highly cited work at a higher rate than the narrow-wins: 16.1 percent of their papers over the following five years landed in the top 5 percent by citation, against 13.3 percent for the group that got the money.⁴

Scientists are not founders and a study section is not an investment committee, so this is an analogy rather than a transfer. But the structure of the problem is the one every network has. The decisions closest to the line are the ones where the screening signal is weakest, and the population sitting just under the line is not obviously worse than the population sitting just over it. The attrition is the part to notice. The people who left after the near-miss never produced the later result, whatever they were capable of.

What a usable no contains

None of this argues for long letters. It argues for four pieces of information, all of them short.

The first is which stage the decision was made at and on what, recorded at the moment it was taken by the person who took it, rather than reconstructed two weeks later by someone squinting at their own notes.

The second is whether this is a "not now", a "not us", or a "not fundable as presented". Those three have almost nothing in common, and most decline emails collapse all of them into the same sentence. A company told "not now" should come back. A company told "not us" should stop spending cycles on this network and go find one whose thesis it fits. A company told the third thing has a problem that no amount of re-application will fix, and it deserves to know that before it spends another six months finding out.

The third is the disqualifying fact, where there is one. If the network does not do pre-revenue hardware, will not lead, or has a conflict in the category, saying so takes a line and saves both sides a cycle. It is the cheapest information a network holds and the most frequently withheld.

The fourth is whether re-application is welcome and what would need to be different. YC's number suggests this is not a soft question. If a large share of a future cohort comes out of the rejected pile, the terms on which people come back are part of how the portfolio gets built.

The cost objection to all of this is real, but it is a cost of capture rather than a cost of writing. If the reason exists as a structured field on the screening record at the moment the decision is taken, the note is nearly free to produce and it accumulates into something the network can query later. If it has to be reconstructed from memory, it is expensive, it is late, and it is probably wrong. That makes this a question about how the pipeline is instrumented before it is a question about manners.

Precision beats warmth

The hiring literature has looked at the same problem for longer than venture has. A meta-analytic review by Truxillo and colleagues in 2009 found that giving candidates explanations is among the more consistently supported ways to improve how they perceive a selection process, and practitioner summaries of that work recommend explaining the decision and setting it in context, such as the size of the applicant pool.⁵ It is a different setting with a different power balance, so treat it as a pointer rather than a proof.

The finding that transfers more cleanly is Howell's one about precision. Founders discount signals they read as noisy. "Not a fit at this time" is a noisy signal by construction, and it will be discounted, which means it does no work in either direction. A specific reason can be acted on. It can also be wrong in a way the founder can see and push back on, which is a feature rather than a defect, because a network that is wrong in public about a category gets to find that out.

The case against, which is not weak

There are defensible reasons networks stay quiet, and they should be stated properly.

A written reason is a written record, and in some jurisdictions and for some grounds that carries legal exposure. Counsel will say so, and counsel is not wrong. A reason also invites a reply, and a proportion of founders will treat it as an opening rather than a close, which adds load to the people with the least spare capacity. And a reason can be wrong and still be believed: Howell's evidence shows founders act on these signals, so a network with noisy screening that issues confident explanations is doing damage at scale. In that case the thing to fix is the screening, not the note.

The resolution that holds up is that feedback quality should track screening quality. A network that can state clearly why it declined something has a process worth describing. A network that cannot has learned something about itself rather than about the company, and that is the more useful finding of the two.

The measured part and the rest

Angel groups reporting to the ACA deployed $491.3 million in 2025, up from $437 million the year before.⁶ The applications behind that figure ran into many multiples of the deal count, and almost none of that activity is recorded anywhere the network can inspect.

The yes path gets the memos, the diligence files and the portfolio reviews. The no path gets an inbox rule. A network that captures the reason for each decline is building a record of its own judgment that it can check against outcomes later, which is the only way to find out whether the screen is any good. A network that does not has one dataset to learn from, and it is the portfolio, which is both the smallest sample it holds and the one selected to make it look right.

Sources

1. Angel Capital Education Foundation / Angel Capital Association, "Best Practice Guidance for Angel Groups: Deal Screening," Shira Cohen, Columbia University, July 2007. https://www.angelcapitalassociation.org/data/Documents/Resources/AngelCapitalEducation/ACEF_BEST_PRACTICES_Screening.pdf

2. Y Combinator, "Frequently Asked Questions." https://www.ycombinator.com/faq

3. Sabrina T. Howell, "Learning from Feedback: Evidence from New Ventures," Review of Finance, vol. 25, no. 3, 2021, pp. 595-627. https://academic.oup.com/rof/article-abstract/25/3/595/6136191 (working paper version: https://www.nber.org/system/files/working_papers/w23874/revisions/w23874.rev0.pdf)

4. Yang Wang, Benjamin F. Jones and Dashun Wang, "Early-Career Setback and Future Career Impact," Nature Communications, vol. 10, article 4331, 2019. https://www.nature.com/articles/s41467-019-12189-3

5. Society for Industrial and Organizational Psychology, "What We Know About Applicant Reactions to Selection," citing Donald M. Truxillo et al., "Effects of Explanations on Applicant Reactions: A Meta-Analytic Review," International Journal of Selection and Assessment, 2009. https://www.siop.org/wp-content/uploads/2024/07/SIOP-Applicant_Reactions_to_Selection_final.pdf

6. Angel Capital Association, "ACA Publishes 2026 Angel Funders Report." https://angelcapitalassociation.org/blog/aca-publishes-2026-angel-funders-report/