Any metric you can imagine is nothing more then a guess. That's why it's research. The researchers guess and tests. Some guesses can be observably better but that's not quantifiable. At least that's what I think.

Surely you see the hazard in funding expensive wildly speculative projects?

If you fail once, fail again but bigger... who cares it’s not our money and the bigger our project the more we pay ourselves...

That's why scientific projects are still judged by panels. It's not like money is thrown at any random project. It must have merit and be judged to actually be able to test a hypothesis.