# Root Cause Analysis

> Structured incident investigation grounded in the Toyota Production System, Kaoru Ishikawa, James Reason's Swiss Cheese model, Dean Gano's Apollo method, and Google SRE blameless-postmortem culture. Five methods: 5-whys, fishbone, postmortem, fault tree, Kepner-Tregoe IS/IS-NOT. [WHAT] RCA for failure analysis: why did transformation projects fail, OSINT cause chains, policy evaluation. Core axiom: the proximate cause is where analysis begins, not ends. People are never root causes — if a human could make the error, the system permitted it. A cause is "root enough" when it's actionable. [WHEN] Use when: root cause, RCA, 5-whys, fishbone, postmortem, fault tree, "why did this happen", "why does X fail", IS/IS-NOT, defect investigation, pre-launch risk. NOT for: structural loops (use systems-thinking), axiom decomposition (use thinking-firstprinciples). [LANGUAGE] Configurable. Non-blameful tone (the SRE-style "blameless").

- Skill: `carlheath/root-cause-analysis` (Agent Skill)
- Install (CLI): `npx skillmds@latest add carlheath/root-cause-analysis`
- Raw SKILL.md: https://api.skillmd.com/api/skills/carlheath/root-cause-analysis/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: carlheath (https://skillmd.com/u/carlheath)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/carlheath/root-cause-analysis

---


# Root Cause Analysis

**Role:** structured incident investigation.
**Tradition:** TPS / Ishikawa / Reason / Gano / Google SRE.

## Five methods

### 1. 5-Whys

Linear or branched cause chain. Good for **simple, single-thread incidents**.

```
Why did X happen? → Y
  Why did Y happen? → Z
    Why did Z happen? → W
      Why did W happen? → V
        Why did V happen? → U  (root cause)
```

### 2. Fishbone (Ishikawa, 6 Ms or 4 Ps)

Good for **multiple suspected cause-areas at once**.

**6 Ms** (manufacturing tradition):
- Method
- Material
- Machine
- Manpower
- Measurement
- Mother Nature / Milieu

**4 Ps** (service tradition):
- People
- Process
- Policy
- Plant/place

Output as mermaid or hierarchical list.

### 3. Postmortem (blameless timeline)

Good for **complex incidents with multiple actors**. Wraps other methods.

Structure:
1. **Timeline** — what happened, when, by whom (without blame)
2. **Triggering events** — direct triggers
3. **Contributing factors** — structural conditions
4. **Root cause analysis** — uses 5-whys or fishbone
5. **Action items** — what's done differently next time (with owner + deadline)
6. **Lessons learned** — what to preserve in the organisational memory

### 4. Fault Tree (AND/OR gates)

Good for **safety-critical multi-path failures**.

Top event (undesired outcome) → backward via logic gates:
- AND gate: all sub-outcomes must occur
- OR gate: any sub-outcome suffices

Identifies **minimal cut sets** — smallest combinations leading to failure.

### 5. Kepner-Tregoe IS/IS-NOT

Good for **subtle, hard-to-reproduce defects**. Distinction analysis.

| | IS (when the problem occurs) | IS-NOT (when the problem doesn't occur) |
|---|------------------------------|------------------------------------------|
| **What** | ... | ... |
| **Where** | ... | ... |
| **When** | ... | ... |
| **Scope** | ... | ... |

The distinction between IS and IS-NOT reveals what's specific about problem cases.

## Method selection

| Problem type | Recommended method |
|--------------|---------------------|
| Simple cause-effect | 5-Whys |
| Several possible cause areas | Fishbone |
| Complex incident with timeline | Postmortem (wraps others) |
| Safety-critical with multi-path | Fault tree |
| Subtle, hard-to-reproduce | Kepner-Tregoe IS/IS-NOT |

## Domain examples

| Question | Method | Insight |
|----------|--------|---------|
| "Why does digital transformation fail in organization X?" | Postmortem + Fishbone | Multiple contributing factors: leadership, tech, culture, finance |
| "Why wasn't the policy implemented?" | 5-Whys | Political will → funding → administrative capacity → root cause |
| "Why is online discourse polarising?" | Fault tree | AND gate: algorithm + lack of media literacy + social incentives |
| "Project was late — one-off or pattern?" | IS/IS-NOT | Which weeks ship on time? What differs? |

## Core axioms (must be applied)

1. **People are never root causes.** If a human could make the error, the system permitted it. "Why was the person able to make this mistake?"
2. **A cause is 'root enough' when actionable.** Stop when you reach the point where intervention can actually happen.
3. **Proximate cause ≠ root cause.** The visible trigger is where analysis begins, not ends.
4. **Non-blameful tone.** Goal: learn from system failure, not find a scapegoat.

## Integration with other skills

- **thinking-science** for hypothesis generation during investigation ("what if the cause was X?")
- **thinking-redteam** for stress-testing remediation proposals (will it actually work?)
- **systems-thinking** for **structural loops** (complementary — RCA is linear, systems-thinking cyclic)

## Output format

- 5-Whys: linear list with arrows
- Fishbone: mermaid or text hierarchy
- Postmortem: structured markdown with 6 sections
- Fault tree: mermaid
- Kepner-Tregoe: table

---

🎯 COMPLETED: [SKILL:root-cause-analysis] [RCA on event X]
🗣️ CUSTOM COMPLETED: [SKILL:root-cause-analysis] [RCA done]