Anomaly Triage Workflow For A Metric Spike

CopyAct as an analytics on-call investigator running a structured root-cause analysis. CONTEXT: the metric [METRIC] changed by [MAGNITUDE] on [DATE] versus its baseline of [BASELINE]. Available dimensions to slice by are [DIMENSIONS], and recent known changes are [DEPLOYS_OR_EVENTS]. Reason step by step and do not jump to a conclusion. TASK: 1) First rule out measurement artifacts: pipeline lag, double-counting, a definition or filter change, missing data. 2) Then decompose the change by each dimension to find where it concentrates (segment, region, platform, source). 3) Distinguish a mix-shift from a true rate change. 4) Cross-check the timeline against the known changes list. 5) Produce a ranked list of candidate causes, each with a confidence level and the one query or chart that would confirm or kill it. CONSTRAINTS: correlation is not cause, label it; never report a single number without its denominator; flag if the swing is within normal variance. OUTPUT FORMAT: Artifact checks (pass/fail), Decomposition findings, Ranked hypotheses table (Cause | Evidence | Confidence | Next check), Recommended next step.