Sample frameDay 320·Innovation · Discipline

Morning. Every bet you're running has an ending. Almost none of them have one written down.

Innovation.

Kill Criteria

5 min read·Apply by lunch

The question

Name the bet you're currently running with real money in it, and then say what would have to be true — and by what date — for you to walk away from it: is there an answer, or have you just discovered that the criterion you're actually using is that the bet ends when the money does?

The idea

Annie Duke won a National Science Foundation fellowship to study cognitive psychology at the University of Pennsylvania, left the doctoral programme to play poker professionally, and spent two decades as one of the best in the world at deciding when to fold. Quit (2022) is her book about the half of that skill nobody teaches, and its central claim isn't that people quit too easily. It's that they almost never quit too soon. Ask an operator to name a bet they killed early and regretted and you'll wait; ask for one they carried a year too long and you'll get three. Her instrument is the kill criterion, and its shape is exact: the best quitting criteria, she writes, combine a state and a date. Not 'if this isn't working we'll rethink it', but 'if we don't have forty paying customers by 30 April, we stop'. Why it can't be written later is a sunk-cost problem and well enough understood. Duke's contribution is when it has to be written instead. The trap is that the only moment you can see a bet clearly is the moment you know least about it — the day you commit, before a single result is in. Write it then, or you'll be writing a negotiation with yourself.

The framework

01
A state and a date — and the state has to be checkable by a stranger.
The criterion has two halves and both are non-negotiable. The state is something checkable by a person with no stake in the answer — forty paying customers, a signed pilot, a unit cost under a named figure. The date is a real date in a real calendar. 'If it's not working by the middle of next year' fails both halves, because you'll be the one deciding what working means and when the middle of next year has arrived. And a kill criterion isn't a prediction that the bet will fail. It's the evidence that would settle the question either way.
02
Written at the moment of maximum excitement, not maximum information.
The window is short and it opens exactly once. Kill criteria have to be set at the point of commitment — the funding decision, the first hire, the first line of code — which is also the point at which you know least about whether the bet will work. That feels backwards, and it is the whole method: information accumulates from there, but so does your investment in one particular reading of it, and the second curve is much the steeper. A month in, your name's on the thing. Six months in, you're defending a version of yourself, and every bad month starts arriving with a story attached.
03
What quitting late actually costs you.
The bill is almost never the money burned on the bet that limped — that's usually smaller than it feels, and it's recoverable in the telling. The bill is everything you didn't start, because the people and the attention were already committed elsewhere. Quitting on time is what buys the next attempt, which means a portfolio full of things nobody will end isn't a portfolio at all; it's one long bet you never actually chose to make. So count your dying projects in attempts blocked rather than in dollars sunk, and the arithmetic changes shape.
04
The single extension — and why there isn't a second.
Criteria trip, and occasionally you'll genuinely want to continue because something real has changed. Allow that exactly once, and make it cost something: a new state, a new date, both written down, both harder than the pair they replace. What this prevents is the rolling extension, where the line moves a quarter at a time and the conversation never actually happens. A bet that's been extended twice has stopped being a bet — it's a commitment nobody has examined, funded out of the money that was meant to buy you options.

Members only · 4 principles + template + AI mentor

364 more frameworks are waiting.

Kill Criteria is Day 1 of 365. One framework every morning for a year — across strategy, sales, negotiation, leadership, and more.

$1 /day

Billed $365/yr · cancel any time

365 frameworks

12 topics · templates · AI mentor

Reserve your place — $365/yr →

One payment a year · the whole library · no auto-renewal.

More Innovation frameworks

All 24 innovation frameworks →