Analysis, Insights, and Recommendations
What this domain tests
Moving from numbers to decisions. Concepts: descriptive vs. diagnostic vs. predictive analysis, correlation vs. causation, segment comparisons to isolate drivers, anomaly detection, and writing recommendations tied to evidence. Know how to structure an insight: observation → evidence → so-what → recommended action.
Traps to watch for
- Presenting a metric without interpretation ("traffic dropped 10%") and calling it an insight.
- Claiming causation from a correlation in a report.
- Ignoring seasonality or a tracking change when explaining a shift (check data quality first).
- Recommending a fix without naming the owner or success metric.
Worked example
Observation: revenue fell 12% month over month. Diagnostic: segment by channel - paid search conversions dropped 40% while organic held. Evidence: paid search spend also fell. Insight: the revenue decline is driven by reduced paid search investment, not demand. Recommendation: restore budget or shift to a channel with better ROAS, and track conversion rate weekly against the 3% target.
Rule of thumb
Before explaining a change, rule out tracking issues and seasonality. Then segment to find the driver. An insight without a recommended action and a success metric is just a number.