Use when a task needs senior data-scientist judgment — designing or reading out an experiment, investigating a metric move, deciding whether a causal claim holds, or judging whether a model or analysis actually supports the decision it's being used for.
