Kill Criteria
What result would make us stop, decided before we start? Write the stop conditions down before anything is built, while nobody is invested yet. When the result comes in, the decision has already been made; you only have to read it.
Decide the line, then run the test.
What each step needs
Write them down
Each condition names something you could observe: a behaviour, a payment, a return visit. "If it doesn't feel right" isn't a criterion.
Every condition could be checked by someone who wasn't in the room. Where one needs a number, write down where the number comes from; if it's a judgement, say so and give the reasoning.
Publish them
Put them somewhere the team can't edit quietly: the proposal, the kickoff notes, or in public.
The criteria are dated, and the people who'll make the call have agreed to them.
Run the test
Test with the people the product is for, not with friends or colleagues who want it to work.
The result comes from the users the criteria describe.
Read the result
Compare the result with each condition as written. If it clears them, the next phase starts. If it doesn't, the venture stops, and you keep what you learned.
The decision quotes the criteria, not a new argument.
apprn, and the rule we hold ourselves to.
Own venture
apprn, our agentic salon management venture, is in a live test with salons. Its go or no-go decision is made against kill criteria set before the test began.
We'll publish those criteria before we publish the result, so the call can't be rewritten afterwards. That's the whole point of the framework: a result can't argue with a line drawn before it existed.
Illustrative example
What a criterion reads like, from the clinic booking assistant on our homepage: "It goes to market only if the beta clinics keep using it." It names who, and what they have to do. Anyone can check it.
Where it goes wrong
- Misread
Writing them after the first results
Why Criteria written after the evidence bend to fit it.
- Misread
Using feelings as criteria
Why A criterion is something you can observe, not something you sense.
- Misread
Inventing a threshold to look rigorous
Why If a number can't be sourced, call it a judgement and write down the reasoning.
- Misread
Treating a stop as a failure
Why A diagnosis that stops a bad build is worth more than a build that ships.
Download: the kill criteria worksheet, a free Word file.
In the Journal: Is it worth building? · How to evaluate a venture studio · An AI venture beside your family business
Published 24 September 2026. Graylemon original, from our engagement method.