As data scientists, we need to resist the urge to be content with accurate predictions and extend analytics to embrace the business problem to ensure that beneficial outcomes are realized and sustained by stakeholders. Statistically, sound predictions can lead to poor decisions, poor outcomes, and shattered trust between analytics teams and the decision-makers they hope to assist. The problem lay in the assumption that “If I can predict ‘X’, then I will make a better decision about ‘Y.’”
Good predictions are not automatic precursors to good decisions. There can be a gap between prediction “X” and decision “Y” that needs to be filled.