Kill criteria

Kill criteria are the conditions you decide in advance that will trigger you to shut a project, product or experiment down. They're written before you're emotionally invested, while you can still think clearly, so the decision to walk away is made by your past, rational self rather than your present, sunk-cost self.

For a founder running lean or a portfolio of bets, kill criteria are the antidote to the slow death by a thousand maybes. Without them, dying products linger, draining attention and money because nobody set the line. With them, you've already agreed: if we haven't hit X by date Y, we stop. Common forms are a revenue floor, an activation rate, a customer-count target, or a runway limit.

The trick is to make the number something you can actually watch, so the kill date isn't a surprise. Say you're testing a new product wedge and you build the activation funnel in Amplitude: your criterion becomes 'if week-one activation is still under 20% after 60 days, we kill it', and the chart tells you the answer without a debate. Say you're running a cold-email test for a new offer in Instantly: the line might be 'if positive-reply rate stays below 2% across 1,000 sends, the angle is dead, not the channel'. You can even make the trigger fire itself, wiring a Make scenario to ping you in Slack the moment a metric crosses the floor, so you're forced to face the decision.

The discipline is honouring them when the moment comes. The whole point is to remove the in-the-moment negotiation. Good kill criteria are specific, measurable and dated. A vague 'we'll see how it goes' is not a kill criterion, it's a way of never deciding.