Most startups do not end on a decided day. They end the way a fire goes out, slowly, then all at once, with nobody able to point to the moment it happened.
One founder described the cost of that better than any framework. Looking back on a company that bled out over two years, he wrote: "Set a kill date. We didn't. We just drifted. 24 months later, still drifting."
Read that again. The damage was not one bad decision. It was the absence of a decision, repeated every morning for two years. Each day the choice was the same: keep going, or stop. And each day "keep going" was the easier answer, because stopping meant admitting the thing was over, and going meant only one more day. So he chose one more day, twenty-four months in a row.
This piece is about the decision that prevents that. Not a pep talk about "knowing when to quit." A mechanical instrument: a date you set in advance, with conditions you write down while you can still think straight, that fires whether or not you feel ready when the day comes. It is duller than inspiration and far more useful, because the whole problem with drift is that it never feels like a decision while it is happening.
Why "I'll know when it's time" is a trap
The plan most founders actually run is unwritten: "I'll know when it's time to stop." It feels responsible. It is the single most reliable way to drift for two years.
Here is why it fails. The moment you would need to read the situation clearly is the exact moment you are least able to. By the time a company is in trouble, the founder has been under months of money pressure and lost sleep, and that state measurably degrades decision-making. A 2024 meta-analysis pooling 256 effects from nearly 112,000 people put the drag of financial scarcity on cognitive performance at roughly half a standard deviation (Financial scarcity and cognitive performance: a meta-analysis, Journal of Economic Psychology, 2024). In plain terms: the brain you are counting on to "know when it's time" is running with a chunk of its capacity quietly missing, precisely when the call matters most.
Then add the second force, the one that has a name in the research: the sunk cost effect. People keep funding a losing course of action specifically because of what they have already poured in, not because of what is left to gain (Arkes and Blumer, The psychology of sunk cost, Organizational Behavior and Human Decision Processes, 1985). The more you have spent, the harder it is to stop, which is exactly backwards from how a clear-eyed outsider would decide. The two years you have already lost become the reason you lose the third.
So "I'll know when it's time" asks a fogged, sunk-cost-loaded brain to make the hardest call of all on the worst possible day. It almost never makes it. It just chooses one more day.
A kill date is a decision made by the version of you that can still think
The fix is to move the decision earlier, to a moment when your judgment is intact, and have it fire later, when your judgment is not.
A kill date is exactly that. It is a date, plus a small set of pre-committed conditions, written down in advance, that together say: if we reach this date and these things are not true, we stop. Not "we reconsider." We stop. You make the hard call once, while you can still make it well, and then you let it hold for you on the day you no longer can.
This is not a new idea. Pilots use it. They set decision altitudes and bingo fuel before they ever need them, because they know that in the moment, under stress, the temptation to push one more mile is strongest and the judgment to resist it is weakest. The number is set on the ground, calm, and obeyed in the air, scared. That is the whole design.
A kill date does the same job for a company. It takes the one decision exhaustion is guaranteed to get wrong and makes it in advance, in writing, by the version of you that still has full capacity.
How to make it concrete enough to actually fire
A kill date only works if it is decidable. "We'll stop if things aren't going well" is not a kill date. It is a feeling, and feelings drift. To be an instrument, it needs three parts.
A date. An actual calendar day, not "around the end of the year." Vague timing is what lets one more day stretch into one more quarter into one more year. Pick the day.
Pre-committed conditions. The two or three things that must be true by that date for continuing to make sense, written now, in advance. Make them observable, not hopeful. Not "if it feels like there's momentum," but "if we have not reached X paying customers," or "if revenue has not covered Y," or "if I have personally gone Z weeks without a clear reason to believe the curve bends." The point of writing them now is that present-you sets the bar honestly, before future-you, exhausted and sunk-cost-loaded, is tempted to quietly lower it.
A tripwire. The mechanism that brings the date back to your attention on the day, instead of letting it slide past unnoticed the way the AWS bills and the auto-renewing domain do. Drift is largely a failure of attention: the deadline arrives and nobody is looking. The tripwire is whatever guarantees someone is looking.
Runway is a countdown timer, not a safety net.
The trouble is that a countdown only protects you if you are actually watching the count, and the fog is very good at making sure you are not. The kill date is how you commit, in advance, to watching it, and to acting on what it shows.
Why this has to be held outside your own head
Here is the part that the willpower version of this advice always skips. You can write the most honest kill date in the world, and then, on the day, talk yourself out of it. Move the date. Soften the condition. Decide the tripwire does not really apply this quarter. Sunk cost does not stop pulling just because you wrote a note to yourself; if anything, by the deadline it is pulling harder than ever.
A commitment you hold only in your own head is a commitment guarded by the exact faculty that is compromised. That is why a kill date works better when it lives somewhere outside you. A co-founder who agreed to the conditions and will hold you to them. A written agreement you cannot quietly edit. Something with enough independence that, when exhaustion reaches to move the date, the date does not move just because you want it to.
That is the function an instrument serves. Not to decide for you, but to hold the decision you already made, steady, while the part of you that made it is temporarily offline. The same logic runs through the whole first 90 days after a shutdown: the calls that matter most are the ones your judgment is least able to make on the day, so you commit to them in advance and let something outside your own head keep them honest.
What this is, and what it is not
A kill date does not promise the next thing works, and it does not make stopping painless. What it does is narrow, and worth being precise about: it keeps the decision to stop from being quietly stolen by exhaustion, one easy day at a time. It converts a two-year drift into a single decision, made once, in clear weather.
And it is honest about the bigger picture. Setting a clean kill date does not buy you an outcome. The base rates are sobering and worth saying plainly: of the founders who close a company and start another, only about one in five see the next one succeed, and many never start another at all. A kill date does not change those odds by itself. What it changes is how much of you is left, and how much you have learned, when the date arrives, instead of two depleted years later.
You are not failing by setting one. You are doing the one thing the drifting founder wishes, in hindsight, he had done. You are deciding while you still can.