Innovation.

The framework

1. 01

   The list has to include the bets nobody declared.

   Write down every live bet, starting with the funded ones and then going after the rest: the pilot still running for a single customer, the experiment two engineers keep alive on Fridays, the partnership that never formally ended, the feature you're still maintaining because killing it would feel like an admission. Anything that has taken more than a week of somebody's time this quarter goes on the list, with the days actually spent beside it rather than the days you intended. The list is almost always longer than the founder expects, and that length is the first finding of the review.

2. 02

   Cost and evidence are two different columns.

   Against each bet put what it has consumed, in all three of the currencies its affordable loss was set in: money; time, counted in the days of the people you would least like to lose; and standing, counted in misses the same customers and staff have watched. Any one of the three past its stop puts the bet over budget, whatever the other two say. Then, separately, put what it has proved. Proved is not produced. A bet that has shipped a beautiful prototype and tested no assumption has produced a great deal and proved nothing, while a fake door that killed a bad idea in a fortnight proved something worth paying for. Name the specific belief each bet has removed from your assumption map this quarter. If a bet removed none of them while consuming real money, you're funding activity and calling it innovation.

3. 03

   Which of these would you start today, knowing what you now know?

   A gate asks it late and asks it partially — only of the projects that have reached a gate, each on its own schedule, and never of the bets nobody declared. So rank the bets against each other on the same page — value if it works, evidence so far, and which tier each belongs to, core or adjacent or transformational — and you'll see whether the money actually sits where you said you'd put it. Then test capacity without flattering yourself: count the live bets against the people you can genuinely staff, and cut until the number fits. Too many bets produces the same result as too few, only more slowly and at greater cost.

4. 04

   Read the kill conditions out loud.

   Beside each bet sits the condition you wrote down before you were emotionally inside it. Read it against the current state and give the verdict in the room — continue, hold, recycle or kill — then record it with a date, because a verdict nobody wrote down isn't a verdict. If a bet has breached its condition and you find yourself arguing for it anyway, that argument is the finding. Then put next quarter's date in the diary while everyone's still there: a review that has never ended anything gets priced by your team as a status meeting inside two quarters.