
Skill: Root Cause Analysis Using Data Patterns and Anomaly Detection
Identify root causes in data anomalies using pattern recognition and statistical analysis
What You Can Do
You can transform unexpected data patterns—revenue drops, churn spikes, performance degradation—into credible root-cause explanations. This skill guides you through statistical anomaly detection, multivariate factor isolation, and evidence-building workflows that produce defensible conclusions your stakeholders will trust and act upon.
Features
Establish statistical baselines and confidence intervals to distinguish meaningful anomalies from natural variation
Break down anomalies by dimension (geography, cohort, device, time period) to identify which variables drive the change
Rank potential causes by likelihood, severity, and controllability to focus investigation effort
Connect observed effects to causal mechanisms with supporting data points and temporal correlations
Translate statistical findings into executive-ready narratives with impact quantification
Calculate percentage change, absolute impact, and financial/operational consequences
Map anomaly timing against deployment windows, marketing campaigns, and external events
Example Output
Example 1: Revenue Drop Investigation
- Detected: 18% YoY revenue decline starting Week 12
- Initial hypothesis: Seasonal pattern vs. structural change
- Factor isolation: 22% drop in new customer acquisition, 8% increase in churn
- Root cause: Competitor launched low-cost offering on March 3rd (aligns with Week 12)
- Recommendation: Price repositioning and product differentiation analysis needed
Example 2: API Performance Degradation
- Detected: Latency increased from 145ms to 410ms (183% increase) on Nov 15
- Correlation check: Deployment #4847 released Nov 15 at 09:22 UTC
- Cohort analysis: Latency spike only affects queries with >1000 result sets
- Root cause: New query optimization logic introduces O(n²) operation for large datasets
- Recommendation: Revert to previous algorithm; refactor optimization logic before redeployment
What's Included
- SKILL.md instruction file with framework overview and investigation methodology:
- Anomaly Detection Checklist: step-by-step guide for quantifying statistical significance
- Factor Isolation Worksheet: template for breaking down changes by dimension and cohort
- Evidence Chain Template: structure for connecting observations to root causes with confidence levels
- Stakeholder Communication Guide: frameworks for translating analysis into business narratives with impact metrics
Who It's For
- Business Analysts investigating metric anomalies and building investigation reports
- Operations Managers troubleshooting performance degradation and system issues
- Product Managers analyzing user behavior shifts and product metric changes
- Data Analysts conducting root cause investigations for cross-functional teams
- Financial Analysts explaining revenue or margin variances to leadership
Best For
- Investigating sudden metric drops (revenue, conversion, retention, engagement)
- Troubleshooting system performance degradation and error rate spikes
- Analyzing customer churn acceleration or behavioral shifts
- Understanding impact of product changes, feature launches, or A/B tests
- Correlating business anomalies with external events or competitor actions







