DS-301g · Module 3
Communicating Predictions to Stakeholders
3 min read
A prediction without context is a number. A prediction with confidence intervals, comparison to baseline, and a recommended action is a decision input. Communication framework: the prediction ("revenue next quarter is projected at $2.3M"), the confidence ("with 80% probability of landing between $2.0M and $2.6M"), the comparison ("this is 8% above last quarter and 3% above the same quarter last year"), and the implication ("the range is wide enough that contingency planning for the $2.0M scenario is advisable"). This four-part structure — prediction, confidence, comparison, implication — converts a model output into language that a decision-maker can act on without understanding the model.
Do This
- Present every prediction as a range with an explicit confidence level
- Compare the prediction to relevant baselines — last period, same period last year, target
- State the implication — what should the decision-maker do differently based on this prediction?
Avoid This
- Report a point estimate without a range — it communicates false precision
- Use statistical terminology without translation — "the RMSE is 0.12" means nothing to most stakeholders
- Present predictions without recommendations — information without action guidance creates paralysis