What's missing in the above is some notion that the "feature machines" should actually do some work to show what the feature will accomplish. Better yet, is th

What's missing in the above is some notion that the "feature machines" should actually do some work to show what the feature will accomplish. Better yet, is that feature more profitable than some other feature? "you can have X in three months or Y in two weeks but if you pick either one, that delays Z by the same amount of time" focuses the attention.