The Founder Background Signal Reference: What Each Credential Proves

You are diligencing a founder from public sources and need to know what each line on their track record actually predicts. This is the lookup table underneath the composite founder scores: one row per public credential, with the measured base rate, the failure mode, and the shelf life. Jump to the row you need, read what it proves and how it lies, and weight it yourself.

Most guides on founder pedigree argue one side. Pedigree matters, or pedigree is overrated. Neither position helps you at the term sheet. What you need mid-diligence is the number behind the credential, the source, the selection discount, and the counter-signal that would falsify it. That is what each section below gives you.

How to use a founder background signal correctly

A founder background signal is a public line on a track record that shifts the odds of a target outcome, up or down, by a measured amount. It is never a verdict on its own, and it is only as good as the base rate you compare it against.

Before crediting any single signal, anchor to how rare success is. In the VCBench founder-profile benchmark of 9,000 profiles, the baseline outlier-success rate is 9 percent. The Azoulay team's growth work describes the 1-in-1,000 fastest-growing ventures. Every unicorn-founder dataset in this reference describes winners after the fact, so it tells you what champions looked like, not the odds that a founder with the same signal wins. Read every row through that filter.

The signals split into two families that deserve different weights. Execution signals describe what a person has built and run: a prior exit, a prior venture-backed company, years operating inside the category. Brand signals describe association: an elite school, a big-tech logo, an advanced degree. The measured lifts below say the same thing repeatedly. Execution and domain fit predict more than brand, and the strongest single documented gap is a prior modest exit, not a diploma.

Signal reference: prior founding outcomes

A prior venture outcome is the highest-signal line on a founder's resume, but the direction and size depend entirely on what the outcome was. The cleanest published number covers next-venture success by prior outcome.

Table A - Next-venture success by prior outcome (Gompers et al. 2010, over 10,000 venture-backed companies)

Read the gap plainly. A previously successful founder is roughly 8 to 12 points more likely to succeed again than a first-timer. That is real, and it is the answer to "what is the repeat versus first-time founder success rate," but it is modest against a base of low success rates, and it is the most selection-distorted number in this whole reference. More on that discount below.

Notice what previous failure does. A founder who failed before sits at 23 percent, barely above a first-timer at 22 percent and well below a prior winner. Prior failure is close to neutral, not a mark against. Do not penalize a scar you would have to invent a story to justify.

Prior founding is common among the biggest winners but far from required. Tamaseb's Super Founders work, built on roughly 30,000 data points comparing billion-dollar and failed startups, found 40 percent of billion-dollar startups were founded by first-time founders. Around four in ten unicorn founders had previously founded another company. So a majority of unicorn founders had founded before, but a large minority had not. First-time status is not a disqualifier.

Signal reference: the prior small exit

A prior modest exit is the strongest single documented base-rate gap in this reference. Tamaseb found that among the repeat founders of billion-dollar companies, 42 percent had a previous company acquired for around $10 million or with similar revenue, compared with 24 percent in a random venture-backed control group.

Table B - Prior small exit as a predictor of a unicorn (Tamaseb)

This is the row that answers "what does a prior exit predict about a founder." A modest prior exit runs at roughly 1.75 times the control rate among unicorn builders. It beats the raw repeat-founder signal because it filters for people who took something to a real outcome, not just people who raised a round and shut down.

The trap is the definition of "exit." A $10 million outcome can be a genuine acquisition or a soft landing dressed up as one. An acqui-hire, where a larger company buys the team and writes off the product, produces the same LinkedIn line as a real sale but proves nothing about value creation.

Signal reference: elite school and advanced degree

An elite-school line is a weak-to-moderate brand signal, and there is no clean published correlation with the size of a founder's eventual exit. Weight it as a small odds nudge, never as a verdict.

The prevalence numbers deflate the myth. SignalFire's Unicorn Origins report found the top six engineering schools, Stanford, MIT, UC Berkeley, Georgia Tech, Carnegie Mellon, and Caltech, accounted for 15.4 percent of unicorn founders, while the entire Ivy League accounted for 12.6 percent. Endeavor's research found only one-third of unicorn founders attended an elite university, and two-thirds were either self-taught or attended a lower-ranked school entirely.

The causal lift is real but small. Strebulaev's odds-ratio work found Stanford and MIT each raise the odds of producing a unicorn founder by about 60 percent and Harvard by about 50 percent. That is a nudge on top of a low base rate, not a doubling. Tamaseb concluded university ranking mattered less than proximity to Silicon Valley.

Where school matters more is field, not name. Computer science, engineering, and economics together accounted for 53 percent of unicorn founders' majors. An in-domain technical degree tells you something about capability; an ornamental prestige degree in an unrelated field tells you about admissions, not about building.

Advanced degrees are rarer than founder mythology suggests. In Refolk's index of professional profiles, 1,526 US founders explicitly signal "PhD" in their headline, about 0.35 percent of the US founder pool. A PhD is a strong domain-depth signal in deep tech and a near-irrelevant one elsewhere; weight it by whether the research is the company's core.

Signal reference: big-tech and apprenticeship pedigree

Ex-FAANG or big-tech tenure works as a founder-pipeline and training-ground effect, not a cleanly measured fundraising-versus-execution split. It concentrates in a very small number of companies, which is what gives the logo its signal, and its trap is that a logo hides the scope of what the person actually did.

Strebulaev's Stanford GSB team analyzed 2,791 founders behind 1,110 US unicorns and found they had previously worked at 6,109 different organizations, but only 33 entities produced 15 or more future unicorn builders. Google alone produced 96 unicorn founders. SignalFire frames the mechanism as apprenticeship: founders are increasingly shaped by a small set of companies that function as modern apprenticeship systems rather than by the universities they attended.

So "does FAANG experience predict startup success" is really a question about which company and which role. Ex-IDF personnel, in the same dataset, were roughly three times more likely to build US unicorns. The signal is concentrated. The failure is treating "ex-Google" as uniform. A founding product manager who owned a launch and a two-year individual contributor who touched one feature carry the same logo and very different evidence.

Signal reference: domain tenure and founder age

Domain match beats raw serial status, and it is more predictive than either school or generic experience. Founders with three or more years in the same industry as their startup are about twice as likely to build a unicorn as someone with no relevant background.

The age evidence points the same way and corrects the young-founder myth. The Azoulay, Jones, Kim, and Miranda paper in AER: Insights found the mean age at founding for the 1-in-1,000 fastest-growing new ventures is 45.0, with similar findings in high-technology sectors, entrepreneurial hubs, and successful exits. A 50-year-old founder is roughly 1.8 times more likely to hit that top tier than a 30-year-old. The mechanism is prior experience in the specific industry, not age itself.

On the serial-founder question, the advantage does not grow with count. Secondary analysis reports it peaks at the second venture and then declines, with figures around 34 percent at the second venture and 28 percent at the third. Treat that per-venture split as unverified, but the direction matters: a fourth-time generalist can under-perform a first-time domain expert. Category continuity across ventures is the thing to read, not the number of logos.

A first-time founder deep in the category will out-predict a fourth-time founder who keeps switching industries.

The example queries below show how to isolate exactly these people from public data: prior modest exits in the same industry, multi-year operators in a specific domain, second-time founders in one vertical. This is the kind of compound search that public profile fields make possible.

Refolk lets me ask for that population in plain English rather than reconstructing it field by field. When the signal you trust is "prior small exit plus domain continuity," a search that expresses both conditions at once is the difference between a shortlist and a week of manual profile reading.

The seven-step read: from raw resume to weighted verdict

The procedure is the same whether you screen ten founders or one. Work down the steps in order; the order matters because the base rate and the selection discount both come before you assign weight. Pedigree-first funds screen school and employer at extraction; data-driven funds explicitly de-prioritize school and lead with prior-founder experience and domain tenure. Use the order below, which puts base rates ahead of brand.

How this goes wrong: failure modes and false positives

Every signal in this reference has a way of lying, and most of the lies come from survivorship. This is the section to read twice. Each row below pairs the false positive with the check that catches it.

The deepest failure sits under all of these. The repeat-founder persistence number is measured only on serial founders who received funding. Gompers' own authors warn that previously successful entrepreneurs who get funded may not be better than other entrepreneurs, but simply funded by better-informed VCs. They add a range restriction that quietly undercuts the whole signal: the very best and the very worst entrepreneurs do not become serial entrepreneurs. So the pool you are measuring is already filtered twice.

There is a labor-market artifact that runs the other way and is worth knowing. Former founders receive 43 percent fewer job-market callbacks, and successful ones fare 33 percent worse than failed ones, a selection story from the hiring side rather than a statement about founder quality. It is a reminder that any number attached to founders carries the fingerprints of who selected them.

The correction for every pedigree and celebrity story is the same two figures: two-thirds of unicorn founders are non-elite, and 40 percent of billion-dollar startups were first-time-founded. Whenever a single credential starts to feel decisive, put it back next to those numbers.

Sizing the pool and keeping the read current

The population you are drawing from is large, which is why individual credentials have to be weighted rather than treated as gates. In Refolk's index, the US founder pool is 437,600, roughly four times the UK's 110,259, and PhD-signaling founders are a thin slice of it.

Table C - Founder pool sizing, Refolk's index

Two things follow. First, a scarce signal like a PhD headline, present in about one in 300 US founders, carries more differentiating power in a search than a common one, but only in domains where research depth is the product. Second, because the pool is so large, the same background line means different things across categories, so the read has to be built per founder, not from a lookup alone.

Keep the read current by watching shelf life. Execution signals decay when the operating role goes stale or the category shifts. A prior exit five years ago in the same vertical the founder is building in now still holds its weight; the same exit a decade ago in an unrelated field has decayed to a story. Brand signals like an elite degree or a big-tech logo do not decay in existence, but their predictive weight stays low regardless of age. Before you finalize, walk this checklist.

The point of a reference is that you stop arguing about whether pedigree matters and start reading each line for what it measures. A prior modest exit in the same category, with the team intact, is the row to trust most. An unrelated elite degree is the row to trust least. Everything else sits between those two, weighted by the number, discounted by the selection story, and dated by its shelf life.