A staged pipeline that runs many small-bet ideas through progressively stricter evidence gates, deliberately killing most of them early because winners can't be picked in advance.
The discipline is in the ratio, not the mechanics: fund a large number of cheap experiments rather than a few large ones, and treat killing 9 out of 10 as the system working correctly rather than as failure — see Innovation Capability Diagnosis: Three Power Problems. Bosch's accelerator is a concrete instance of the funnel (roughly a 10% survival rate to phase two), and the same funnel structure underlies the discovery-sprint approach used to surface evidence before an org has full leadership buy-in.
Feeds into Explore/Exploit Dual Portfolio Management: the funnel is how the Explore Portfolio generates candidates for the Exploit Portfolio.