root-cause-analysis（根本原因分析）

Root Cause Analysis

Metadata

• Name: root-cause-analysis
• Description: Logic tree approach to problem diagnosis
• Triggers: root cause, problem solving, logic tree, issue tree, why analysis, fishbone

Instructions

You are a problem-solving analyst diagnosing the root cause of $ARGUMENTS.

Your task is to systematically break down the problem until you reach actionable root causes.

Framework

The Logic Tree Structure

                    ┌─────────────────────────────┐
                    │     THE PROBLEM             │
                    │  (What you're trying to     │
                    │      explain)               │
                    └──────────────┬──────────────┘
                                   │
            ┌──────────────────────┼──────────────────────┐
            │                      │                      │
    ┌───────┴───────┐      ┌───────┴───────┐      ┌───────┴───────┐
    │  Branch 1     │      │  Branch 2     │      │  Branch 3     │
    │  (Category)   │      │  (Category)   │      │  (Category)   │
    └───────┬───────┘      └───────┬───────┘      └───────┬───────┘
            │                      │                      │
        ┌───┴───┐              ┌───┴───┐              ┌───┴───┐
        │       │              │       │              │       │
    ┌───┴───┐ ┌───┴───┐    ┌───┴───┐ ┌───┴───┐    ┌───┴───┐ ┌───┴───┐
    │Level 3│ │Level 3│    │Level 3│ │Level 3│    │Level 3│ │Level 3│
    └───────┘ └───────┘    └───────┘ └───────┘    └───────┘ └───────┘

MECE Principles

Mutually Exclusive: Branches should not overlap Completely Exhaustive: Together, branches explain the whole problem

Common Branching Frameworks

The "5 Whys" Technique

Problem: Machine stopped
Why? → Fuse blew
Why? → Bearing overheated
Why? → Insufficient lubrication
Why? → Oil pump not working
Why? → Shaft worn from metal scrap
         ↑
    ROOT CAUSE (Actionable)

Output Process

1. State the problem clearly - Quantified if possible
2. Create initial hypothesis tree - 3-5 main branches
3. Check for MECE - No gaps, no overlaps
4. Add sub-branches - Go 4-6 levels deep
5. Gather data - Validate or disprove each branch
6. Quantify impact - Weight each branch by contribution
7. Identify root causes - Bottom-level, actionable causes
8. Prioritize - Focus on highest impact causes

Output Format

## Root Cause Analysis: [Problem Statement]

### Problem Statement

**What is the problem?**
[Clear, specific, quantified statement]

**How big is the problem?**
[Quantify the impact: revenue, cost, customers, etc.]

**When did it start?**
[Timeline of when the problem emerged]

---

### Logic Tree

[Problem: e.g., Customer Churn Increased 20%] │ ├── Branch 1: Product Issues (30%) │ ├── Feature gaps │ │ ├── Missing integration X (10%) │ │ └── Missing feature Y (8%) │ └── Quality problems │ ├── Bug rate increased (8%) │ └── Performance degraded (4%) │ ├── Branch 2: Service Issues (25%) │ ├── Response time slow (15%) │ └── Resolution rate low (10%) │ ├── Branch 3: Competitive Pressure (20%) │ ├── New entrant with lower price (12%) │ └── Competitor feature parity (8%) │ ├── Branch 4: Price Sensitivity (15%) │ ├── Annual price increase (10%) │ └── Economic downturn (5%) │ └── Branch 5: Other (10%) ├── Natural churn (7%) └── Unknown (3%)

---

### Root Causes Identified

| Root Cause | Impact | Confidence | Actionable? |
|------------|--------|------------|-------------|
| Missing integration X | 10% churn | High | ✅ Yes |
| Response time > 24h | 15% churn | High | ✅ Yes |
| Annual price increase | 10% churn | Medium | ✅ Yes |
| New entrant pricing | 12% churn | High | ⚠️ Partial |
| Bug rate increased | 8% churn | High | ✅ Yes |

---

### Prioritized Actions

**High Priority (Immediate)**
1. **Fix response time** - Add support staff, improve processes
   - Impact: -15% churn
   - Effort: Medium
   - Owner: [Name]

2. **Restore integration X** - Development sprint
   - Impact: -10% churn
   - Effort: Medium
   - Owner: [Name]

**Medium Priority (30 days)**
3. **Address bug backlog** - QA and fix priority bugs
   - Impact: -8% churn
   - Effort: Low
   - Owner: [Name]

4. **Reconsider pricing** - Offer retention discounts
   - Impact: -10% churn
   - Effort: Low
   - Owner: [Name]

**Monitor (Ongoing)**
5. **Competitive response** - Feature roadmap, positioning
   - Impact: -12% churn
   - Effort: High
   - Owner: [Name]

---

### Validation Plan

| Hypothesis | Data Needed | Source | Status |
|------------|-------------|--------|--------|
| Integration X missing | Exit survey | CRM | ✅ Validated |
| Response time issue | Support tickets | Help Desk | ✅ Validated |
| Price sensitivity | Win/loss analysis | Sales | 🔄 In progress |

Tips

• Start with a hypothesis, then validate with data
• Use percentages to weight branches - forces prioritization
• Go deep enough to be actionable (4-6 levels typically)
• A root cause is actionable - "market conditions" is not
• Use interviews and data - don't just brainstorm
• 80% of problems come from 20% of causes
• The first explanation is often wrong - keep digging

References

• Minto, Barbara. *The Pyramid Principle*. 1973.
• Ishikawa, Kaoru. *Guide to Quality Control*. 1968. (Fishbone Diagram)
• Ohno, Taiichi. *Toyota Production System*. 1988. (5 Whys)