Data doesn’t make decisions… people do. Flip the order. First: What decision will this inform right now? Next: What specific outcome would make us choose A over B? Then: What evidence would change our mind the other way? When you anchor analysis to a fork in the road, you automatically narrow the variables, shrink the noise, and set a standard for “good enough.” You also define the audience and the time horizon… two forces that wreck otherwise solid stats when left implicit. A decision-first frame protects you from collecting attractive but irrelevant measures, from polishing tiny effects, and from confusing activity with progress. It also clarifies sampling: if the decision is local and immediate, you don’t need national census data; you need representative signal from the population the decision touches. Start with the choice… then measure only what could move it.
