Interpret and Report Results
Turning results into insight
Interpretation is where analysis becomes meaning. An observation is what the data shows; an insight explains why it matters and what should change. Candidates must distinguish them and must communicate uncertainty honestly: confidence intervals, effect sizes, sample limitations, and the assumptions the result depends on.
Reporting is audience design. Executives need the decision, the size of the effect, and the confidence - not the method appendix. Operational audiences need segment-level detail and the actions it implies. Good visualizations follow conventions: zeroed axes for bar charts, labelled units, no truncated axes to dramatize small differences, and a chart type that matches the comparison.
Traps. Overclaiming causality from an observational study. Cherry-picking a favourable subgroup or timeframe. Hiding a confidence interval that crosses zero. Presenting a statistically significant but tiny lift as a win. Reporting method detail while omitting the "so what". Failing to state limitations, which destroys credibility when a stakeholder finds the gap.
Rule of thumb - the "so what, who cares, now what" test. Every finding should answer: what did we learn, why does it matter to this stakeholder, and what should they do differently? If a slide cannot answer all three, cut it or rework it.
Worked example. A campaign test shows a 2.1% retention lift with a 95% interval of −0.4% to 4.6%. Correct reporting: "We cannot rule out no effect; the point estimate is promising but the evidence is inconclusive. Recommend extending the test by four weeks rather than rolling out." That is interpretation with integrity, and it sets up the decision domain.