Root cause analysis is one of those skills everyone in quality claims to have and few teams do consistently well. The tools aren't the problem. Fishbone diagrams, 5-Why, Is / Is Not and fault trees all work. What goes wrong is how teams use them under time pressure, with a customer waiting and a production manager asking when the line can restart.
Below are seven common mistakes in root cause analysis, why they happen, and how to avoid them. The examples come from automotive, aerospace and medical device settings. They're illustrative scenarios based on typical situations, not specific named incidents.
1. Jumping to the solution
Someone in the first meeting says "it's the new supplier's resin" and the whole team spends the next week proving it. This is confirmation bias, and it is the single biggest threat to a good analysis.
Example (medical devices): A pouch sealing line starts producing seals with channel leaks. The team quickly blames a new lot of film and requests a supplier investigation. Two weeks later the supplier proves the film is in spec. The actual cause was a worn heater band on one sealing jaw, which a temperature profile would have shown on day one.
How to avoid it:
- Write down every hypothesis, including the popular one, as a hypothesis
- Use an Ishikawa diagram to force the team to consider all six categories before choosing
- Ask, for every favorite theory: what data would prove this wrong? Then go get that data
2. Confusing the symptom with the cause
"The part was out of tolerance" is not a root cause. It's the problem restated. Neither is "the seal leaked" or "the bolt was loose."
A useful test is to ask whether your root cause tells you what to fix. "Bolt was loose" doesn't. "The torque tool's clutch was worn, so it slipped at 70% of target torque, and the tool isn't on a calibration schedule" does.
How to avoid it: keep asking why until the answer points to something in the process, equipment, material or system that you can physically change.
3. Analyzing from the conference room
A lot of root cause analyses are written entirely from memory, opinions and spreadsheets. Nobody goes to look at the parts, the machine or the process running.
Example (automotive): A team investigating scratched painted panels spends two meetings debating paint shop parameters. When they finally walk the line, they find a rack with a missing rubber protector that touches every fifth panel at the same spot. The defect location on returned parts matched it exactly.
How to avoid it: go to the gemba, the place where the work happens. Look at the defective parts and good parts side by side. Watch the process run. Talk to the operators on the shift that made the defect. Ten minutes on the floor often saves ten hours of discussion.
4. Finding only one root cause
Most defects that reach a customer have at least two causes: why it was made wrong (occurrence) and why it wasn't caught (escape or non-detection). Many also have a systemic cause: why the management system allowed both.
Example (aerospace): A fastener is found without a torque stripe during final inspection at the customer. The team's root cause is that the mechanic skipped the step. That's only the occurrence. Why did the in-process inspection buy-off pass without the stripe? Why does the work instruction allow the stripe step to be signed off in a batch at the end of the shift? Those are the escape and systemic causes, and they're where the durable fix lives.
How to avoid it: structure every analysis with at least two legs, and expect your customer to ask about the third. Our guide on writing a D4 root cause shows how.
5. Not verifying the root cause
A root cause that hasn't been verified is a hypothesis. It might be right. It might be one of three contributing factors. It might be wrong.
The strongest verification is to turn the defect on and off: reintroduce the cause and reproduce the defect, then remove it and show the defect disappears. When that isn't possible, use lab analysis, correlation with process data in the same time window, or a controlled trial.
How to avoid it: add a verification line under every root cause, with data. "Verified: 0 of 300 parts failed with the new filter; 7 of 1,000 failed with the blocked filter." If you can't write that line, you're not done.
6. Weak corrective actions
The analysis can be perfect and the defect can still come back if the corrective action is weak. The usual suspects:
- Retrain the operator
- Remind the team at the shift meeting
- Add a caution note to the work instruction
- Increase inspection
These actions depend on people remembering, or on catching the defect rather than preventing it. They decay over time.
A better approach is to think in a hierarchy, from strongest to weakest:
| Strength | Type of action | Example |
|---|---|---|
| Strongest | Eliminate the cause | Change the design so the wrong orientation is impossible |
| Strong | Error-proof (poka-yoke) | Fixture that won't accept a part loaded backwards |
| Strong | Automatic detection with reaction | Sensor that stops the machine on low vacuum |
| Medium | Standardize and control | Parameter locked in the recipe, under change control |
| Weak | Inspection | 100% visual check |
| Weakest | Training and reminders | Retrain the operator |
Training is still sometimes necessary. It just shouldn't be the only action for a cause that reached a customer.
7. Closing too early, and only fixing one spot
Two related mistakes happen at the end of an investigation.
The first is declaring success too soon. A week of good parts doesn't prove much for an intermittent problem. Define the validation period and the success criterion before you start, long enough to cover the conditions that caused the problem, like a full material lot or a seasonal humidity swing.
The second is fixing only the place where the defect showed up. If a blocked vacuum filter caused dropped parts on one pick-and-place station, how many other stations use the same filter with no maintenance plan? This read-across, called yokoten in lean language, is often where the biggest savings are.
How to avoid it:
- Set a validation window and a measurable target up front
- Update the PFMEA, control plan and work instructions, not just the machine
- List every similar product, line and process, and check each one
Our free FMEA template and control plan template make those document updates quicker.
A quick self-check for your next analysis
Before you call a root cause analysis complete, ask:
- Did we write down and test more than one hypothesis?
- Does each root cause tell us exactly what to fix?
- Did we look at the actual parts and process?
- Do we have both an occurrence and an escape cause?
- Is every root cause verified with data?
- Are our actions stronger than training and inspection?
- Did we validate long enough, and read across to similar processes?
Seven yeses and you have an analysis worth putting your name on. If the 5-Why is where your team usually gets stuck, read why most 5-Why analyses fail. And if you want a second pair of eyes on a report before it goes to the customer, Vantage 8D's evaluator scores each discipline and flags weak root causes and actions like the ones above.