Any portfolio built before consensus is mostly wrong. That is not a failure of judgement, it is the arithmetic of working early. The useful question is not how to be wrong less often — it is how much each wrong answer is permitted to cost.
That cost is set at the beginning, and it is set by structure rather than by intent. A thesis tested with four hard conversations and a working prototype is cheap to walk away from. The same thesis tested with a senior hire, a launched brand, and a year of runway is not — financially, and more importantly psychologically, which is usually the currency that runs out first.
So we design the disproof before we design the build. What would we have to see in the next eleven weeks to stop? If nobody in the room can answer that, we are not investigating an idea. We are decorating one.
Losing cheaply is what makes it possible to keep looking. The firms that run out of nerve are rarely the ones that were wrong most often. They are the ones that let each wrong answer get too expensive to survive.