Quality in Practice — Lesson 5: FMEA: Catching Defects Before They Occur

Quality in Practice · 5 min read · 20.10.2026

Welcome to the fifth lesson of the Quality in Practice series. So far we have worked with defects that had already occurred — we hunted for the true root cause with 5 Whys and handled complaints as 8D projects. Today we turn that around. You will learn what FMEA is, how to rate a risk with three numbers, and how to run a workshop that yields actions with names and deadlines. Not a spreadsheet nobody opens after the audit.

What FMEA Is and When It Pays Off

FMEA stands for Failure Mode and Effects Analysis. In manufacturing you most often encounter PFMEA, the process variant: you walk the production step by step and at each step ask three questions. What can go wrong here? What will the consequences be? And why does the defect occur — why doesn't inspection catch it?

There is no point doing it all at once. The analysis pays off at four moments: before launching a new product or technology, when a major change is made to the process, tooling or supplier, after taking delivery of a new machine, and after every completed 8D. Record the root cause the team uncovered in 8D in the FMEA where it should prevent recurrence — not only at the point where the defect occurred.

On our projects we see companies dealing with the same defect separately on three lines for years. One joint workshop is enough to reveal that it is a single root cause with three names.

How to Rate a Risk: Severity, Occurrence, Detection

Rate each failure mode with three numbers from 1 to 10. Severity (S) tells you how serious the consequences are — from a scratch on a non-critical surface to a safety hazard. Occurrence (O) estimates how often the cause fails. Detection (D) measures the chance of catching the defect before it leaves the operation. Watch out: the last number works in reverse — 1 means near-certain detection, 10 means the customer is the first to see the defect.

The classic approach multiplies the numbers: RPN = S × O × D. The newer AIAG-VDA handbook ranks risks into H/M/L priorities instead of using the product. Why? Multiplication can mislead. Severity 9 with occurrence 2 gives an RPN of 54 — less than a recurring cosmetic defect. Always address risks with serious or legal consequences, even when the score looks harmless.

An example from the shop floor. CNC machining of aluminium castings for an automotive customer: the operator can clamp the blank into the fixture upside down. Severity 8 — a destroyed milling tool worth €380 plus scrap. Occurrence 4 — five cases a year, mostly after shift changes. Detection 7 — the defect only surfaces two operations later. RPN 224, act on it. The team fitted the fixture with a pin that refuses an upside-down blank: a poka-yoke for less than €50. Occurrence dropped to 1, the RPN to 16, and annual scrap from this cause to zero.

The cheapest defect is the one that occurs at the team table — on paper.

How to Run a Workshop That Leaves Something Behind

We know the most common form of FMEA from our company visits: the quality manager writes it alone the day before the audit, the team signs it, the file is closed. A document like that satisfies the auditor. It won't save the part. The analysis is done by the team, not the quality department.

Rules that have worked for us:

• A cross-functional team of 4 to 6 people. Process engineer, foreman, operator, quality engineer — and, where gauges are a concern, a maintenance technician as well. The operator knows defects that never make it onto any form.
• Blocks of 2 to 3 hours. After three hours, both focus and creativity collapse. Three meetings beat one marathon.
• Analyse at the machine. Take the drawing, the material flow and the latest measurements out to the shop floor, not to the meeting room.
• Data as input. Complaints, completed 8Ds, scrap records. Without them you discuss opinions instead of risks.
• Every action has a name and a deadline. A list of suggestions without ownership is just a wish list.

A 90-employee assembly centre ran three 2.5-hour workshops. The team identified 23 failure modes and assigned actions for the six highest risks — from a poka-yoke interlock to a stricter first-off inspection. Three months later, complaints from the line dropped by 68%. And the document stayed alive: after every change and every 8D, the team returns to it.

This Week's Exercise

1. Pick one operation — ideally the one with the highest scrap or the most complaints over the past year.
2. Spend 60 minutes at the machine with the foreman and the operator. List at least 10 failure modes that genuinely threaten that operation.
3. Rate each one with S, O and D, calculate the RPN and pick the three highest.
4. Check yourself: does the list include the root cause of your latest complaint from this operation? If yes, the method holds up. If not, you have been lucky so far — add the item.
5. For the highest risk, propose one action and assign it with a name and a deadline.

Expect one pattern: for the first twenty minutes the team lists the obvious defects. Then the operator starts talking — and the most expensive items come out.

You will find the complete Quality in Practice series — eight lessons, quizzes and exercises — in our e-learning offering. Next time we move from predicting risks to measuring: the control plan and SPC. Measuring what really matters.

Quiz — test your knowledge