Root Cause Analysis — Why Did It Actually Happen?
When a part is out of tolerance or a tool breaks, fixing the immediate symptom (adjusting an offset, indexing a tool) isn't enough. If you don't find the root cause, it'll happen again. This page covers the practical tools: 5 Why, Fishbone, and the corrective action process.
The Two Types of Problems
| Type | What it means | Example |
|---|---|---|
| Special cause | Unexpected, one-off, identifiable event | A chip under the part caused a bad dimension; a tool broke early |
| Common cause | Built into the process, happens regularly | The machine consistently drifts 0.02 mm over a shift; tool wear rate is too high |
Special causes are fixed by addressing the event. Common causes require changing the process. Knowing which type you have determines your corrective action.
5 Why — Keep Asking "Why?"
The simplest root cause tool: ask "why" repeatedly until you reach the underlying cause, not just the immediate symptom.
Example: a bore is out of tolerance
Why 1: Why is the bore 0.03 mm too big? → The boring bar deflected during the cut.
Why 2: Why did the boring bar deflect? → The feed was too high for the overhang.
Why 3: Why was the feed too high? → The operator used the standard feed for a short bar, but this hole requires a long bar.
Why 4: Why didn't the program account for that? → The process plan didn't specify reduced feed for deep boring.
Why 5: Why didn't the plan specify it? → No one considered the overhang when the program was written.
Root cause: process plan doesn't account for tool overhang. Fix: add a feed reduction rule to the setup sheet for deep bores.
5 Why works when you follow the chain. Don't stop at "operator error" — that's always the surface answer. Ask why the operator was put in a position to make that error.
Fishbone Diagram (Ishikawa)
When there are many possible causes, brainstorm them all organized into categories. The standard categories for machining:
| Category | Questions to ask |
|---|---|
| Machine | Is the machine aligned? Spindle worn? Way lubrication OK? |
| Method | Is the process plan correct? Parameters right? Sequence logical? |
| Material | Is the material consistent? Hardness variation? Sawed vs milled stock? |
| Measurement | Is the gauge calibrated? Is the operator measuring correctly? |
| Man (operator) | Trained? Following the sheet? Fatigued? |
| Environment | Temperature variation? Chips on the table? Coolant condition? |
Non-Conformance (NC) Process
When a part is out of spec, you need a documented process:
- Identify: tag the part, mark it non-conforming, segregate it so it doesn't get shipped.
- Contain: check other parts from the same batch — how many are affected?
- Evaluate: can it be reworked? Is it a scrap? Can the customer use it as-is (concession)?
- Root cause: use 5 Why or Fishbone to find why it happened.
- Corrective action: fix the cause so it doesn't repeat.
- Verify effectiveness: after the fix, monitor for 1–2 weeks — did the problem actually stop?
Corrective vs Preventive Action
| Corrective Action (CA) | Preventive Action (PA) | |
|---|---|---|
| Trigger | A problem already happened | A problem could happen |
| Purpose | Fix the cause of the non-conformance | Stop potential non-conformances before they occur |
| Example | "Deep bores need reduced feed" added to setup sheet | "All boring programs over 3× diameter overhang get a feed check" |
Process Qualification
Before a new process runs in production, prove it works consistently:
- First Article Inspection: one good part proves the process can work.
- Capability study (Cpk): 30–50 parts prove it's consistently good.
- Run-off: produce a full batch, measure at intervals, confirm no drift.
- Sign-off: engineer, quality, and production all approve before release.
Practical Tips
"Operator error" is not a root cause. It's the start of the investigation. Ask: why was the operator able to make that mistake? Was the setup sheet unclear? Was the program wrong? Was there no poke-yoke (mistake-proofing)? A good process should make it hard to do the wrong thing.
- Contain first, investigate later. Make sure bad parts don't ship before you spend time on root cause.
- Don't blame people — fix systems. Even if an operator made a mistake, the process should have caught it.
- Document everything. The 5 Why, the fix, and the verification should be written down so the next person learns from it.
- Verify the fix worked. A corrective action that isn't followed up is just a suggestion.
Related: for machining-specific problems (chatter, tool wear, dimensional error), see Chapter 10 — Machining Troubleshooting. For measuring process stability, see Process Capability & SPC.