Innovation.

The framework

1. 01

   Write the idea down as beliefs, one to a card.

   Pull the plan apart into the statements that have to be true for it to work: they want this enough to change what they do now; we can deliver it at this quality for this cost; they'll pay this much; the channel we're counting on will carry us. Write each on its own line, in the present tense, phrased so that it could be false. A belief you can't phrase so it might be wrong isn't an assumption, it's a preference. Most ideas produce twenty or thirty of these, and a good few will be things nobody in the room has ever said aloud.

2. 02

   Which of the three risks are you quietly avoiding?

   Bland and Osterwalder sort every card into one of three families. Desirability: do they want it. Feasibility: can we actually build and deliver it. Viability: does the money work once we do. Nearly every team over-tests the family it enjoys — engineers reach for feasibility, because the answer is inside the building; commercial founders reach for viability and build a model, because a model can be finished on a Sunday. Desirability is the one that gets deferred, since it lives outside with people who owe you nothing. Sort your cards by family and count each column; the thin one is telling you something.

3. 03

   The axes are stakes and evidence — nothing else goes on the map.

   Draw two axes. Up the side: how much of the idea dies if this belief is wrong. Across the bottom: how much evidence you hold — plenty on the left, none on the right — where evidence means something that happened rather than something that was said in a meeting. Every card takes a position. The value of doing it this way is procedural as much as analytical — once the map is on the wall, the choice of what to test next stops being a matter of who argued hardest, and the person defending their favourite belief has to defend its position on the grid instead.

4. 04

   One test at a time, in the order the map gives you.

   The queue runs top-right first, and that ordering does real work: if the belief everything rests on turns out to be false, the cards beneath it never needed testing, and you've saved the quarter it would have taken to work through them. Run one, take the result honestly, and re-plot: evidence moves a card left, and a surviving belief almost always spawns two you couldn't see before. You've already run two columns of this map without calling them that — the business model written as guesses three mornings ago, and the beliefs you listed in the search ledger back in the elective's first week — and two more are coming, one with numbers instead of cards and one with an owner beside every line. They're columns on this map, not new instruments, so keep one wall and add to it.