In business analytics, it’s not unusual to get questions like:
“We launched a campaign last week, and we’d like to know how it performed.”
A request like this can end in a handful of ways:
(A) You’re lucky, and you manage to figure out a meaningful answer that the requester accepts. (B) You hit a dead end because of lack of data, or some assumptions turn out to be invalidating the question.
In the latter case, regardless of the roadblock, you could:
- drop the question because you can’t answer it
- give a partial answer that doesn’t solve the question but might be good enough to move on
- answer an entirely different question, which oftentimes is not a satisfactory outcome for anybody
A better inquiry might be:
“We launch campaigns all the time, is there a way to consistently identify whom we should target so that we don’t waste time and resources on bad prospects?”
A request like this can end in a model, or a tool that provides a solution that is useful today as well as in the future. Any limitation or shortcoming of the instrument can be tuned and improved in future iterations as more information flows through it.