This is why experimentation tries to focus on a singular change at a time -- in short, you're right, and we can focus on iteration to iteration to tease out ful

This is why experimentation tries to focus on a singular change at a time -- in short, you're right, and we can focus on iteration to iteration to tease out fully causal factors. But you know that, based on randomized assignment (or at least representative assignment school by school) an impact that A versus B determines. So you do know E(Y | do(X), Z) to a degree, at least partially.