This goes back to the whole correlation does not equal causation thing in statistics. You're likely thinking of a correlated number, the measurement number mus

This goes back to the whole correlation does not equal causation thing in statistics. You're likely thinking of a correlated number, the measurement number must be a causative. Direct improvements on any causative number will by definition affect the outcome. Say we're looking at developer retention. A correlated number is like scrum points. A causative number is like salary. Of course, while the data on the causative and correlative effectiveness of both metrics is dispute-able... you still get my point.