Job

Pick which kind of question you are actually asking, from a fixed menu: an amount or trend, a cause, a mechanism, a base rate, whether a source can be trusted, whether a real outcome improved. The type you name dictates which method can settle it.

Pick which kind of question you are actually asking, from a fixed menu: an amount or trend, a cause, a mechanism, a base rate, whether a source can be trusted, whether a real outcome improved. The type you name dictates which method can settle it.

Example

For the onboarding pilot you write: "Job: causal effect, not amount or trend. We are not asking whether signups rose. We are asking whether this flow raised them." That single line rules out a before-and-after chart and forces a holdout group.