Explanation
It works because attention decays predictably during uninterrupted footage, and a change resets that decay. This is why a static talking head with no cuts loses people at a rate that has little to do with what is being said.
The technique is heavily overused, which has produced its own problem: videos cut so aggressively that nothing is on screen long enough to be understood. The interrupt is meant to buy attention for the content, not to substitute for it.
The common mistake
Adding cuts to fix low retention. If the drop is in the first two seconds it is a hook problem, and faster cutting further in cannot reach it.