Data Analytics and Reporting
Data analytics and reporting convert data into insight and decisions. The exam tests analytical method selection, metric and report design, validation, reproducibility, and the guardrails around self-service analytics.
Core rules
- Match method to question: descriptive (what happened), diagnostic (why), predictive (what is likely), prescriptive (what to do).
- Define metrics with numerator, denominator, population, time window, and exclusions; a KPI without a denominator invites distortion.
- Validate before publishing: reconcile to source, test edge cases, check cohort definitions, and document assumptions.
- Design reports for the decision: lead with the question, show comparison and uncertainty, avoid chartjunk and vanity metrics.
- Make analysis reproducible: version data, code, parameters, and definitions; peer-review models and monitor drift and bias.
Worked example
A churn dashboard reports 12% monthly churn. The analyst defines churn as no login for 30 days among accounts active in the prior 90 days. Recomputing with a 60-day window changes churn to 8%, revealing the metric was sensitive to an arbitrary threshold. The team publishes the definition, shows the sensitivity, and adds cohort size. The reasoning: the analysis was mathematically correct but decision-unsafe until the definition and uncertainty were visible.
Traps
- Confusing correlation with causation in diagnostic claims.
- Vanity metrics with no decision owner or denominator.
- Stale extracts presented as current performance.
- Self-service with no certified datasets, so definitions diverge.