Structured method for diagnosing why a metric, system, or process changed — drops, spikes, anomalies, discrepancies, regressions, or "why did X happen" questions. Use this skill whenever the task involves explaining an observed change or difference, even if the user doesn't say "root cause": questions like "why did conversions drop," "what's driving this spike," or "something broke, figure out what" all qualify. Do NOT use for open-ended exploration of unfamiliar data (use an EDA skill), for cross-system metric discrepancies like "why don't these two tools agree" (use metric-reconciliation), or for forward-looking "what will happen" questions.
