What if the most expensive mistake in your product never happens because a team spent two hours asking “what could go wrong here?” That question, asked systematically and written down, is the heart of the Fehler-Möglichkeits-Einfluss-Analyse — better known in English as Failure Mode and Effects Analysis, or FMEA. It is one of the oldest and most durable risk tools in engineering, and it remains the backbone of quality planning in automotive, medical device, aerospace, and machinery industries. This article explains what the method actually is, how the modern version works step by step, where teams go wrong, and how it compares to alternatives.
What the German Term Actually Means
The German compound breaks into three parts: Fehler (failure or fault), Möglichkeit (possibility), and Einfluss (influence or effect), with Analyse at the end. Put together, it describes a structured study of which failures are possible and what influence each one would have.
English-language standards use Failure Mode and Effects Analysis. A related variant adds criticality and becomes Failure Mode, Effects and Criticality Analysis (FMECA). In practice, German-speaking automotive and machinery suppliers say “FMEA” in conversation, and the acronym travels across languages unchanged.
The method’s purpose is narrow and practical. It is not a general brainstorming session about risk. It is a documented, repeatable examination of a product or process that identifies:
- Failure modes — the specific ways something can fail
- Effects — what happens to the customer, the user, or the next process step
- Causes — why the failure mode occurs
- Controls — what currently prevents or detects it
- Actions — what should change
Where FMEA Came From

FMEA has military roots. It was formalised in US military procedure MIL-P-1629, dated 9 November 1949, under the title “Procedures for Performing a Failure Mode, Effects and Criticality Analysis.” That document established the idea that failure analysis should be a formal, documented procedure rather than an informal habit.
NASA adopted the technique during the Apollo programme in the 1960s, at the same time that fault tree analysis was being developed at Bell Labs in 1962 for the Minuteman missile programme. From there, the method migrated into civilian industry, first aerospace, then automotive.
The automotive story matters most for anyone working in German-speaking supply chains. German manufacturers standardised FMEA through VDA, the Verband der Automobilindustrie, while the US Big Three worked through AIAG. For years, suppliers faced two competing handbooks, two sets of forms, and two ways of scoring risk. That ended in June 2019, when the harmonised AIAG & VDA FMEA Handbook was published, replacing the separate AIAG fourth edition (2008) and VDA fourth edition (2006) approaches.
The Seven-Step Approach in the 2019 Handbook
The 2019 handbook did more than merge two documents. It restructured the method into seven steps, which gave FMEA a clearer logical flow than the older form-driven approach.
- Planning and preparation. Define scope, boundaries, team, and timing. Decide whether this is a product (design) FMEA or a process FMEA.
- Structure analysis. Break the system into elements — assemblies, components, process steps — and show how they relate.
- Function analysis. State what each element is supposed to do, including requirements and specifications.
- Failure analysis. For each function, name the failure modes, the effects, and the causes. This is where the classic “failure chain” appears: cause leads to failure mode leads to effect.
- Risk analysis. Score severity, occurrence, and detection, then assign preventive and detective controls.
- Optimisation. Define actions, assign owners and dates, and re-score after implementation.
- Results documentation. Record the outcome so it survives staff turnover and audits.
The seven steps sound bureaucratic on paper, but they solve a real problem. Older FMEA forms let teams jump straight to scoring without first agreeing on what the system does. That produced long lists of failure modes with no clear link to requirements, and reviewers could not tell whether anything important had been missed.
Severity, Occurrence, and Detection
The three scoring dimensions are the engine of the method.
| Dimension | What it asks | Typical scale |
|---|---|---|
| Severity (S) | How bad is the effect if this failure reaches the customer? | 1 to 10 |
| Occurrence (O) | How likely is the cause to happen? | 1 to 10 |
| Detection (D) | How likely are current controls to catch it before it escapes? | 1 to 10 |
Severity is a property of the effect, not of the design. A brake failure scores high severity whether it happens once or a million times. Occurrence and detection are estimates that depend on data, field history, and process capability, which is why they are the numbers most often argued about in review meetings.
From RPN to Action Priority
For decades, teams multiplied the three scores to get the Risk Priority Number, or RPN: S × O × D, with each factor typically scored from 1 to 10. The RPN produced a single number that could be sorted from highest to lowest, which made it popular and, in some organisations, badly misused.
The 2019 handbook de-emphasises the RPN in favour of an Action Priority (AP) table. Instead of a product, the table maps each combination of severity, occurrence, and detection to High, Medium, or Low priority. The change reflects a simple insight: a severity of 9 or 10 should never be buried under a moderate RPN just because occurrence and detection happen to be low.
In practice, this means teams should stop treating a threshold like “RPN above 100 requires action” as a rule. The AP table gives a defensible answer for each line, and it removes the temptation to game scores to push a line below a cutoff.
Why the RPN Gets Criticised
The RPN has been criticised for years in reliability-engineering literature, and the criticisms are worth knowing because they still apply to any scoring scheme.
Non-uniqueness is the biggest problem. Different combinations of severity, occurrence, and detection produce the same product. A failure with scores of 9, 2, and 5 has the same RPN as one with 2, 5, and 9. The first is a severe hazard with a rare cause; the second is a minor annoyance that happens often. Treating them as equal is clearly wrong.
Equal weighting is the second issue. Multiplication assumes each factor contributes proportionally, but severity is not interchangeable with detection. A high severity cannot be offset by good inspection.
Arbitrary scale boundaries are the third. Where a team draws the line between a 4 and a 5 on occurrence is a judgement call, and different teams draw it differently. The numbers look precise but carry uncertainty.
The Action Priority table addresses the first two criticisms reasonably well. It does not eliminate the third, which is why calibration across teams and reviewers remains essential.
Practical Tips That Actually Work
Teams that run FMEA well tend to do a handful of things differently from teams that produce thick documents nobody reads.
Start with functions, not failures. If the team cannot state what a component or process step is supposed to do, in measurable terms, the failure analysis will be guesswork. Write the requirement first, then ask how it could fail.
Keep the team small and cross-functional. A design engineer, a process engineer, a quality specialist, and someone close to the customer or field service is usually enough. Large committees produce long meetings and shallow analysis.
Use real data for occurrence. Warranty claims, field returns, process capability studies, and test results are better than opinion. When data is missing, say so in the document rather than inventing a number.
Score severity once and leave it alone. Severity follows from the effect, and the effect does not change because the team is under time pressure. If a severity score moves during optimisation without a design change, something is wrong.
Treat detection as a design problem. The most common weak spot is a control that relies on an operator noticing something. Human inspection is a detection control with a poor track record, and the scoring should reflect that honestly.
Close the loop on actions. An FMEA with open actions and no owners is a wish list. Every recommended action needs a name, a date, and a verification step.
Common Mistakes to Avoid
- Copying last year’s FMEA and changing the date. Carry-over is fine for structure; it is not fine for failure analysis when the design or process has changed.
- Scoring in the meeting without evidence. Occurrence and detection scores should be prepared before the session and defended during it.
- Confusing FMEA with a fault tree. FMEA is bottom-up: component failure to system effect. Fault tree analysis is top-down: undesired event to contributing causes. They complement each other.
- Ignoring interfaces. Many real failures occur where two subsystems meet, and interface failures are exactly what element-by-element analysis tends to skip.
- Letting the FMEA live in a spreadsheet nobody opens. The document is a living record. If it is not updated when the design changes, it becomes fiction.
FMEA and Broader Standards
FMEA does not exist in isolation. IEC 60812 covers FMEA and FMECA procedures, and its third edition was published in 2018, replacing the 2006 second edition. That standard is the reference point for general industrial applications outside automotive.
In medical devices, ISO 14971 governs the application of risk management, and its third edition was published in 2019. FMEA is commonly used as one tool within an ISO 14971 risk management file, though it is not the only one and not always the best fit for every hazard.
Aerospace and systems engineering programmes often reference FMEA alongside other analyses in systems engineering handbooks, where it appears as one of several reliability techniques rather than the single answer.
How FMEA Compares to Other Risk Tools
| Method | Direction | Best for | Main limitation |
|---|---|---|---|
| FMEA | Bottom-up | Component and process failure modes | Long, repetitive for complex systems |
| Fault tree analysis | Top-down | Tracing causes of one specific bad event | Needs a clearly defined top event |
| Hazard analysis | Event-driven | Safety hazards and harm to people | Less detail on process variation |
| Bow-tie analysis | Cause and consequence | Communicating barriers around a major hazard | Coarse for component-level design |
The practical answer for most teams is not to choose one. A design FMEA plus a fault tree for the highest-severity events covers more ground than either alone, and the two cross-check each other.
Building FMEA Into Daily Work
The organisations that get value from the Fehler-Möglichkeits-Einfluss-Analyse treat it as part of how work is done, not as an audit deliverable. That means the analysis starts early, when changes are still cheap, and continues as the design matures. It means failure modes found in the field feed back into the document. And it means the scoring conversation is about evidence, not about hitting a number.
For teams working across German and English supply chains, the terminology is the least of the challenges. The method, the seven steps, and the Action Priority logic are the same on both sides. What differs is discipline: whether the analysis is done with real data, real owners, and real follow-through. Readers who want to see how structured quality and process documentation fit together in practice can find related material at occcy.com.
Conclusion
The Fehler-Möglichkeits-Einfluss-Analyse has survived since 1949 because it asks a question every engineering team should ask anyway: what could fail, how bad would it be, and what are we doing about it? The 2019 AIAG & VDA handbook modernised the method with seven clear steps and replaced the flawed RPN with an Action Priority table. The tool is only as good as the evidence behind the scores and the follow-through on the actions. Teams that treat FMEA as a living record of design decisions get real protection. Teams that treat it as paperwork get a thick document and the same failures they always had.
What does Fehler-Möglichkeits-Einfluss-Analyse mean in English?
It translates as Failure Mode and Effects Analysis, or FMEA. The German term is used in automotive and machinery engineering, and the acronym is the same in both languages. A related variant that adds criticality is called FMECA, for Failure Mode, Effects and Criticality Analysis.
When was FMEA first formalised?
FMEA was formalised in US military procedure MIL-P-1629, dated 9 November 1949, titled “Procedures for Performing a Failure Mode, Effects and Criticality Analysis.” NASA used the technique during the Apollo programme in the 1960s, and it later spread into automotive and medical device manufacturing.
What changed in the 2019 AIAG & VDA FMEA Handbook?
The handbook, published in June 2019, merged the separate AIAG and VDA approaches into one method. It introduced a seven-step approach covering planning, structure, function, failure, risk, optimisation, and documentation. It also de-emphasised the Risk Priority Number in favour of an Action Priority table that classifies each line as High, Medium, or Low.
Why is the Risk Priority Number criticised?
The RPN multiplies severity, occurrence, and detection, which means different combinations can produce the same number. A severe hazard with a rare cause can score identically to a minor issue that happens often. Critics also point to the equal weighting of the three factors and the arbitrary boundaries between scores.
What is the difference between FMEA and fault tree analysis?
FMEA works bottom-up: it starts with component or process failure modes and traces their effects. Fault tree analysis works top-down: it starts with one undesired event and traces the contributing causes. The two are complementary, and many teams use both for high-severity risks.
Which standards cover FMEA?
IEC 60812 covers FMEA and FMECA procedures, with its third edition published in 2018. In medical devices, ISO 14971 governs risk management and its third edition was published in 2019; FMEA is often used as one tool within that framework. Automotive teams follow the AIAG & VDA handbook.


