What problem does it solve? When metrics drop, funnels leak, or retention declines, teams often jump to conclusions without verifying data quality or isolating the real cause. This Skill turns raw metric changes, funnel data, retention figures, and cohort breakdowns into structured analysis reports where every conclusion carries an evidence grade (A data-supported, B strong inference, C speculation), so decisions rest on verified findings rather than guesses. ## Core Features & Use Cases - Five-Step Diagnostic Workflow: Align on the analysis goal, run a data health check (period completeness, metric definitions, sample size, baseline validity), decompose the metric one dimension at a time, locate the phenomenon via trend/comparison/structure/correlation, then produce graded conclusions and actions. - Evidence-Graded Reporting: Outputs a structured report (conclusion first, data health check, decomposition tree, phenomenon location, prioritized action list, hypotheses to verify) using the template in references/analysis-report-template.md. - Action Prioritization: Separates recommendations into stop-loss actions, validation experiments, and long-term structural fixes with priority levels. - Use Case: Your DAU dropped 10% last week. Paste the dashboard numbers or CSV export, and the Skill decomposes the decline by channel, user segment, and funnel stage, flags that the drop coincides with a push-notification strategy change (graded as B inference), and recommends a P0 rollback plus an A/B test to verify. ## Quick Start Ask the AI to use operations-data-analysis to diagnose why your key metric dropped, pasting your data, dashboard screenshots, or SQL query results.
