Write the kill criteria before you launch

The reason most pivots feel like failures is that nobody wrote down what success looked like before they started. The experiment ran for four months, the numbers were ambiguous, the team got emotionally attached to the idea, and the decision to kill it became a referendum on the whole company rather than a normal closure of a specific bet.

Annie Duke's operator principle, which she applies mostly to poker but translates cleanly: establish kill criteria before you place the bet. Specific benchmarks, specific dates, and a specific action if the benchmarks aren't met.

The pre-mortem template

Before a seed-stage team starts an experiment — a new go-to-market channel, a feature, an ICP expansion — write four lines:

• Thesis: we believe X will produce Y because Z.
• Leading indicator: Y will be visible in [specific metric] by [specific date].
• Kill threshold: if [metric] is below [number] by [date], we stop.
• Action if we kill: who does what in the following seven days to wind this down cleanly.

That's forty minutes of work. The value comes not from what you write but from having to be specific. Most teams, when asked to name the leading indicator, realise they were going to run the experiment on vibes.

Why this works

When the metric misses by the date, the conversation becomes 'the experiment hit its kill threshold, we're winding it down' — which is operationally identical to 'the experiment succeeded, we're scaling it'. No emotion, no referendum, no identity spiral. Just the next thing.

Founders who run this process consistently produce twice the learning per quarter. Not because they're smarter, but because they've removed the sunk-cost tax that every un-killed experiment charges on the next decision.

The meta-rule

If you find yourself unable to write the kill criteria — if every threshold feels arbitrary and every date feels wrong — that's the signal you're not ready to run the experiment. Spend another week on the thesis. Or admit the experiment is actually a hope, and run a different one.

The hardest part of the pre-mortem isn't writing it. It's agreeing to honour it. If you can't agree to the kill in advance, don't agree to the launch.