The most reliable seed-stage fundraising signal may be having done grunt work at a top-tier fund

Deedy has updated his founder-signal chart. The question is straightforward: which founder backgrounds make top-tier VCs more likely to invest at seed or Series A? The data comes from Harmonic and covers 8,525 U.S. startups between 2019 and 2026, using a logit regression to examine 63 signals and control variables. The sample is roughly twice as large as the previous version. The previous version covered 4,836 U.S. startups from 2021 to 2024 and 59 signals.

One of the top signals in the new chart is having worked at a top-tier fund. In the main chart, former top-tier fund employees have an odds ratio of 2.83, and former frontier AI lab employees 2.16. An odds ratio can be roughly understood as a change relative to the baseline: around 1.0 means little difference, above 1 increases the odds, and below 1 decreases them. It is not a success rate; the horizontal lines in the chart also show 95% confidence intervals.

Further down, former Tesla/SpaceX employees are at 1.83, Technion or Tel Aviv U at 1.81, having a Tier 1-2 lead at the previous company at 1.73, having worked at a company backed by Tier 1-2 at 1.58, a CS degree at 1.45, a top university at 1.41, former Google at 1.39, top-company alumni at 1.38, Harvard at 1.37, and former scaleup experience at 1.36. Among negative signals, PhD is 0.75, fintech company 0.74, founders over 50 0.66, and any MBA 0.60. These are the values from the 15 strongest signals in the previous version. The new version moves former frontier labs and cybersecurity higher, turns IIT into a positive signal, and turns being educated in China into a negative signal.

Why the top-ranked signal is high is more worth looking at than how high it is. Deedy explains that former top-tier fund employees include GPs, venture partners, EIRs, investors, scouts, fellowship or researcher program participants, and even interns. The common thread is being close to the process and knowing people. This explanation is not romantic, but it fits the reality of seed rounds: the company has no data yet, so investors can only read people, networks, and past positions.

Someone in the comments asked how to distinguish a top-tier fund's 'ability to pick winners' from having 'been at a top-tier fund' itself. Deedy replied that the analysis only looks at seed and Series A rounds; most companies at this stage do not yet have much performance data, so it is impossible to directly pick winners. That answer draws the boundary clearly. The chart predicts the probability of raising money, not a company's ultimate success or failure. Commenter Iris晴柔 put it more bluntly: most of these are VC pattern matching, predicting who gets invested in, not who wins.

A few more details got the comments going. Caltech is the strongest university signal, while Waterloo is a negative signal; Technion is strong, but being educated in Israel is a negative signal; former Stripe is a negative signal, which Deedy said he has not figured out; being a former founder, even with an exit, is less important than expected, partly because many are small acquihires. Commenter Shaun Gold's quip hits the mark: seed investing is increasingly like résumé fantasy football; some good founder without these badges may be quietly building the real winner. That judgment may not be right, but it is a reminder that the signal chart is a mirror of the fundraising market, not a verdict on startups.

Another boundary is geography. Someone asked whether the same holds in Europe, and Mladen Lotar also pointed out that the sample covers only the U.S. Deedy confirmed this. Before applying this chart elsewhere, hold off.

Original tweet: Deedy's update