data-driven process improvement

tons of causal inference

       Levers
         ⬇
IN ➡ {Process} ➡ OUT
         ⬆
       Environment

Out = Process(In, Lever, Environment)

$Lever^* = \argmax_{Lever} \mathbb{E}_{Env} [Value(Out) - Cost(In)]$

How do we understand the process well enough to maximize the output value per input cost?