Your Runway Math Is Two Years Stale

Graduation rate is a conversion metric. Take every company that raised a seed round in a given year, count how many raised a Series A inside some window, divide. Seed cohort on the bottom, Series A on top, window usually twenty-four months. That is the whole calculation. It is the same arithmetic as trial-to-paid, and VCs quote it the way growth teams quote activation.

The number going around right now is grim. Crunchbase ran the cohorts in May: of the companies that raised a seed of $1M or more in 2023, 24% have progressed. The 2024 cohort sits at 16%. Through 2020, that figure was typically 55% or higher.

So is seed broken? Should we all pack it in and go get real jobs?

I went looking for the window instead, because the window is where conversion metrics go to lie.

The window is shorter than the wait.

Carta's book puts the median time from seed to Series A at 2.1 years. In 2019 it was 1.5. For AI companies it runs 1.9.

Now hold that against a twenty-four-month measurement window. The window is shorter than the median time to finish. More than half the companies that will eventually raise a Series A have not raised it yet on the day we take the picture, and we publish the picture anyway.

Statisticians call this right-censoring. On a cohort dashboard it shows up as the last bar being short. The bar is short because the cohort is young, it fills in over the following quarters, and every quarter somebody screenshots it before it does. The 2024 seed cohort is the last bar.

Run the cohort out four years and half of them make it.

Carta again: roughly 50% of seed companies reach Series A by the sixteenth quarter. The 2019 Q1 cohort landed at 49.1% by Q16.

49.1% against a pre-2021 norm of 55% or better. That is a real decline, and it is a decline of about six points, not thirty-nine. Both numbers are honestly computed. They measure different things, and the one that travels is the one that fits in a headline.

Frankly, this is the most boring possible explanation for a scary chart, and I would happily stop here and tell you the panic is manufactured. The windowing bug hides something more expensive than it exaggerates.

Seed sped up. Series A stayed where it was.

Here's the deal. Censoring explains why the recent number reads low. It leaves the stretch from 1.5 years to 2.1 completely unexplained, and that stretch is the part with a dollar figure attached.

PitchBook has first financings on track to pass 7,000 this year, a record by more than 1,300 deals. AI pushed the cost of starting a company down far enough that more people start one, and multistage funds carrying more capital than seed strategy have been writing seed checks at volume.

The producer sped up. The consumer did not.

That is an unbounded queue. Seed enqueues at a record rate, Series A dequeues at whatever rate partnerships can actually run diligence and hold an IC vote, and nothing in the system pushes back on the producer. No backpressure anywhere in the pipe. So the queue grows, and what eventually drains it is a timeout. Runway expires, the company falls off the back, and it never appears in the numerator of anybody's conversion metric.

Unlock a focused set of AI strategies built to streamline your work and maximize impact. This guide delivers the practical tactics and tools marketers need to start seeing results right away:

Stay ahead of the curve with these top strategies AI helped develop for marketers, built for real-world results.

Download the Free Report

You are selling a fifth of the company to buy seven months.

The market already repriced the wait, and it did it in public, on the term sheets. Carta's median seed raise is $4M. The 95th percentile is $16.6M. Median seed valuation is $20M post-money. Andy McLoughlin at Uncork put his own firm on the record: the average check went "from $2.5 million or less, to $4.5 million."

A $4M seed into a $20M post is about a fifth of the company. Founders are selling roughly that fifth to buy twenty-five months of runway where eighteen used to cover it. The bigger seed is the market pricing the longer queue. That is a rational trade, and the founder pays for it in ownership.

The failure I keep watching is founders taking the 2026 check and keeping the 2019 plan. Crunchbase is blunt about the consequence: budget a twelve-month runway to your Series A and you are scheduling a cash crunch, because the median is past two years. You raised for the new latency and staffed for the old one.

I have made the same mistake from the other side of the table. I have sat in a Monday partner meeting and asked why a portfolio company had not raised its A yet at month eighteen. In 2019 that was a fair question. In 2026 it is asking why the last bar is short.

A founder asked me two weeks ago which number to plan against, the 16 or the 49. She had both in a doc, sourced, windows labeled, better cited than most of what crosses my desk. She wanted me to pick one.

I didn't pick. Both numbers are averages over a population she is not a member of, and a graduation rate describes a cohort while telling you nothing whatsoever about a company. She has fourteen months of cash and a growth number that has not moved since April. That is the entire forecast, and she already had it.

She wanted a number because a number would have let her put the question down for a week. I gave her a window instead, and a shrug dressed up as rigor.

Here is what it costs me to say it depends: nothing. I do not have a payroll run on the fifteenth.

— SWEdonym