> How do you actually know an idea has failed, though?

The same way you know if a clinical trial in unsuccessful. You set milestones in advance, and if you don't meet those milestones then you've failed.