root-cause-investigation

name	root-cause-investigation
description	Systematically diagnose why a metric changed unexpectedly. Activate when a KPI moves significantly, a stakeholder asks "why did X drop/spike?", or a post-incident review requires evidence-based root cause documentation.

When to use

• A key metric moved significantly and the team needs an explanation
• A stakeholder challenges a number and you need to trace the cause
• Post-incident review requires documented root cause analysis
• Distinguishing real signal from noise in a reported metric change

Process

1. Validate the change — confirm the metric movement exceeds normal variance (use rolling average ± 2σ); rule out data quality issues before proceeding
2. Establish timing — plot the full metric history; identify exactly when the change began; distinguish sudden shift from gradual trend
3. Decompose the metric — break the metric into its components (e.g. revenue = volume × price × mix); use scripts/drilldown_analyzer.py to identify which component(s) drove the change
4. Drill down by dimension — compare performance across geography, platform, user segment, product; rank contributors by impact (absolute + % of total change)
5. Generate and test hypotheses — correlate the change timing against a list of known events (releases, campaigns, incidents); use references/hypothesis_testing_guide.md
6. Document findings — fill assets/rca_report_template.md with: change magnitude, timing, component analysis, dimension drilldown, validated hypothesis, and tiered recommendations

Inputs the skill needs

• Required: metric name, time period, and magnitude of change
• Required: historical data sufficient to establish a baseline
• Optional: dimensional breakdown data (by segment, geography, channel)
• Optional: event log of potential causal events (releases, campaigns, outages)

Output

• scripts/drilldown_analyzer.py — metric decomposition and dimensional contribution analysis
• references/rca_framework.md — structured RCA methodology and decision tree
• references/hypothesis_testing_guide.md — common root cause patterns and diagnostic tests
• assets/rca_report_template.md (filled) — evidence-based findings with tiered recommendations