Analyze structured data, define trustworthy metrics, investigate changes, and turn numbers into decision-ready conclusions with explicit assumptions and caveats. Use when Porter needs KPI logic, SQL-oriented analysis plans, funnel or cohort analysis, experiment readouts, anomaly investigation, forecast framing, business diagnostics, or executive-facing analytical summaries. Do not use for data-pipeline implementation, dashboard layout design, or causal claims that cannot be supported by the evidence.
Answer the business question without lying with precision.
This skill is for analytical work that must stand up to scrutiny: defining metrics correctly, checking whether a movement is real, separating signal from noise, and ending with a recommendation that matches the strength of the evidence.
Use this skill for:
Use this skill when the task needs:
Do not use this skill for:
Before analyzing, identify:
If the metric definition is unstable, fix that before interpreting movement.
Return outputs such as:
A good result should be both numerically defensible and decision-useful.
State:
Do not let analysis drift into curiosity work with no decision owner.
Verify:
Bad metric logic makes every later conclusion suspect.
Work from:
Averages hide the answer surprisingly often.
For every conclusion, distinguish:
If the setup is observational, treat causation as unproven unless assumptions and design actually support it.
Call out limits such as:
Caveats belong in the main analysis, not buried at the end.
End with:
A strong result should:
Use prompt.md for analytical posture and output format.
Use guides/qa-checklist.md before finalizing.
Use examples/README.md for representative analytical asks.
Use meta/skill.json for routing metadata and boundaries.