← All founder workflows

How to write kill criteria before you fall in love with the idea

Write kill criteria by fixing three things in advance: a date, a measurable threshold, and the person who holds you to it. A usable criterion reads "if I do not have 5 paying customers by 30 November, I stop" — dated, countable, and decided before you are emotionally invested. Vague criteria ("if it isn't working") never fire, because by the time the question arises you can argue any number into hope. Founders who write kill criteria at the start abandon dead ideas months earlier than those who do not.

1 hour $0 to run6 stepsUpdated 2026

The workflow

0/6 done

  1. Kill criteria written at week twelve are negotiated by someone with twelve weeks of sunk cost. Write them in the first hour, in the same document as the idea, when you have nothing invested and can still be honest about what a real signal looks like. This is the entire trick: the decision is made by the version of you who has not yet spent anything.

  2. Paying customers, signed pilots, repeat users in week four. Never impressions, waitlist signups, follower growth or "conversations" — every one of those can be inflated by effort alone, which is exactly why founders reach for them at month five. If the metric can go up while the business goes nowhere, it is not a kill criterion, it is a comfort blanket.

  3. Ninety days is the standard unit and it is usually generous. A shorter window forces the cheap version of the test; a longer one guarantees you will spend the whole thing. Write both the date and what you will have spent by then — hours and dollars — because the second number is what makes the first one feel real.

  4. Most ideas do not fail cleanly; they underperform in a way that suggests a change. Decide in advance which result means "change the audience and rerun" and which means "stop entirely", so you get at most one pivot rather than an unbounded chain of them. Write it as two thresholds: below X, stop; between X and Y, one pivot, same deadline; above Y, continue.

    Kill-or-Continue QuizFull tool
    Input
    90 days, 0 paying customers, 400 waitlist signups, 3 pivots already, still building
    What comes back
    KILL. The signal here is not the zero — it is three pivots inside one deadline with the waitlist as the only rising number. Three pivots means the original thesis was never the constraint. Write the post-mortem and keep the list.

    Run it yourself — free, no signup:

    1. 1.Has anyone paid you actual money for this?

    2. 2.When you stop pushing for a week, what happens?

    3. 3.Of the people who tried it, how many still use it a month later?

    4. 4.Do you have one channel that reliably brings strangers?

    5. 5.Honestly — do you still want to work on this?

    6. 6.In the last month, did you learn something that changed the plan?

    Answer all six for a verdict. None of them ask what you've already spent — that's the point.

  5. Criteria enforced only by you are enforced by the person most motivated to reinterpret them. Send the document to a founder friend with the date in the subject line and ask them to check in on it. This costs one email and is the single highest-leverage step here, because the failure mode is never forgetting the number — it is quietly renegotiating it.

  6. Decide now what you do the day it fires: write a 500-word post-mortem, keep the audience list, cancel the subscriptions, take a week off. A kill with a plan is a normal Tuesday. A kill without one feels like an identity event, and that feeling is what keeps founders paying for dead SaaS subscriptions eighteen months later.

Questions founders ask about this

What makes a good kill criterion?
A date, a countable metric, and a number you set before starting. "Five paying customers by 30 November" works; "if it isn't gaining traction" does not, because traction can be reinterpreted indefinitely by someone who wants to keep going.
How long should I give an idea before killing it?
Ninety days of honest effort is the common unit for a validation test, and shorter is usually better because it forces the cheap version. Write down what you will have spent in hours and dollars by that date — it makes the deadline concrete.
How many pivots should I allow before killing an idea?
One, inside the original deadline. Multiple pivots inside one experiment usually mean the thesis was never the binding constraint, and each one resets your learning while consuming the same runway.
Isn't quitting too early the bigger risk?
Both risks are real, which is why the criteria go in writing at the start. A pre-committed threshold protects you from quitting on a bad week and from grinding for a year on something that never moved.
What should I do the day the kill criteria fire?
Write a short post-mortem, keep the audience and the list, cancel recurring costs, and take a break before starting the next thing. Deciding this in advance turns the kill into a procedure instead of a crisis.

Next, founders usually do this

The tools used above have their own pages — Kill-or-Continue Quiz — and the SOP SOP: Kill a failing venture without sunk-cost bias runs the same ground in more depth. Also worth reading: the Nine Lives Doctrine, and real verdicts from ideas kitty has run this workflow on.

kitty.build runs this entire workflow for you

Every step above — the research, the competitor read, the numbers, the honest verdict — is what nine specialist AI boards do automatically when you feed her an idea. She will tell you to kill it if it deserves killing. First idea is free.

Feed her an idea — free

No card · Failed tasks are free · Your repo, your domain, your revenue