Types of FMEA: DFMEA, PFMEA, FMECA and AIAG-VDA Compared
Once you have grasped the basic method, the obvious next question is which version of it you are supposed to be using. Ask three practitioners and you may get three different answers, because FMEA is not one method but a family of closely related ones that grew up in different industries.
This guide sorts them out. If you have not yet met the core method, start with our complete guide to FMEA and come back here.
Two questions that separate the variants
Most of the confusion clears up once you realise the variants differ along two independent lines:
- What are you analysing? A whole system, a product design, or a manufacturing or service process. This gives you System FMEA, Design FMEA and Process FMEA.
- How do you prioritise? By Risk Priority Number, by Action Priority, or by criticality. This is what separates classic FMEA, the AIAG-VDA method and FMECA.
You pick one answer from each line. A Process FMEA scored with Action Priority is a perfectly ordinary combination.
Variants by what you are analysing
System FMEA
Looks at a whole system and the interactions between its subsystems. Used earliest in a project, when the architecture is still being decided and changing it is still cheap. Failure modes tend to be about interfaces: things that fail because two parts do not work together properly, rather than because either part is defective.
Design FMEA (DFMEA)
Analyses a product design before it is built, asking how the design itself could fail to do its job. The output is design change: different materials, tolerances, redundancy, a different mechanism. Because it runs early, a DFMEA has the greatest leverage of any variant. Fixing a design flaw on paper costs a fraction of fixing it after tooling.
Process FMEA (PFMEA)
Analyses the process that makes or delivers the thing, asking how the process could fail to produce what the design specified. The output is process change: different controls, poka-yoke devices, revised work instructions, additional inspection.
PFMEA is the variant most non-manufacturing organisations should reach for first, because almost every business runs processes even if it does not design products. Onboarding a customer, closing the month-end accounts and dispatching an order are all processes with failure modes.
Concept FMEA
Sits before DFMEA, comparing candidate concepts against each other before one is chosen. Ford's original handbook included it. It is less commonly used today but genuinely valuable when the decision is which approach to take rather than how to execute a chosen approach.
FMEA and FMECA: what the C adds
FMECA stands for Failure Mode, Effects and Criticality Analysis. The additional step is criticality analysis, and it comes from the original military line of descent rather than the automotive one.
The distinction matters because the terms get used interchangeably in casual conversation when they should not be. An FMEA identifies failure modes and their effects. An FMECA does that and then additionally ranks each failure mode by criticality, which combines the severity of the consequence with how likely that failure mode is to occur.
Criticality analysis comes in two forms, and the gap between them is large.
Qualitative criticality
Each failure mode is assigned a severity class and an occurrence level, then plotted on a criticality matrix, typically with severity along one axis and probability along the other. Failure modes landing in the top-right corner get attention first.
This is achievable by any organisation. It needs judgement, not data, and the matrix is a genuinely good communication tool because a senior manager can read it without training.
Quantitative criticality
This is the version that appears in MIL-STD-1629A, and it calculates a criticality number from real reliability data. The inputs are the part failure rate, the proportion of that part's failures which take this particular failure mode, the conditional probability that the failure mode produces the loss in question, and the operating time.
The output is a number with actual meaning, which is a considerable step up from an RPN. The catch is the data requirement. You need failure rate figures for your components, and most organisations outside defence, aerospace and high-volume manufacturing simply do not have them. Attempting quantitative criticality with invented failure rates produces false precision, which is worse than honest judgement.
A sensible rule: use qualitative criticality unless you genuinely hold reliability data, in which case use the quantitative form and be glad you can.
A note on the standards
MIL-STD-1629A remains the reference document for FMECA in defence and aerospace work, despite having been formally cancelled in 1998. Cancelled standards do not vanish from practice, and contracts still cite it. The current international standard covering the same ground is IEC 60812, updated in 2018, which addresses both FMEA and FMECA.
The AIAG-VDA method
Published in 2019, this is the harmonised approach that replaced the separate American AIAG and German VDA methods. If you supply the automotive industry, this is almost certainly what your customer expects.
It restructures the work into seven steps: planning and preparation, structure analysis, function analysis, failure analysis, risk analysis, optimisation, and results documentation. The first three steps are the significant change in emphasis. Where classic FMEA tends to jump fairly quickly to listing failure modes, the AIAG-VDA method insists you first break the subject into a structure and articulate what each element is supposed to do. Failure modes then fall out as ways those functions are not delivered, which produces a more complete list and fewer vague entries.
Action Priority instead of RPN
The most consequential change is that Action Priority replaces the Risk Priority Number.
Rather than multiplying severity, occurrence and detection, each combination of the three is looked up in a table which returns a priority of High, Medium or Low. The table is weighted so that severity dominates: a high-severity failure lands in the High band even when occurrence and detection scores are low.
The reasoning is exactly the weakness described in our main FMEA guide. Multiplying ordinal scores produces a number with no real meaning, lets very different risks share a score, and buries rare but catastrophic failures. Decades of arguments about where to set the RPN action threshold were symptoms of the same problem.
The counter-argument is worth knowing. RPN, whatever its theoretical flaws, gives you a continuous number you can track over time, which makes it easy to show that risk went down after you did something. Action Priority gives you three bands, which is better for deciding what to do and worse for demonstrating progress. Plenty of organisations run both: Action Priority drives the decision, RPN tracks the trend.
FMEA-MSR
The AIAG-VDA handbook also introduced supplemental FMEA for Monitoring and System Response. It addresses failures that occur while the customer is actually using the product, and asks whether the system detects them and responds safely in the moment. It matters most for anything with software and sensors that has to fail safe in real time. If your product is purely mechanical, you can skip it.
Sector variants worth knowing
Healthcare FMEA
Healthcare adapted the method for clinical processes, where failure modes are things like a medication being given to the wrong patient. The best-known version uses a hazard scoring approach combining severity and probability, and adds decision logic about whether a hazard needs action at all given existing controls. The emphasis shifts from components towards process steps and human factors.
Medical devices and ISO 14971
Medical device manufacturers work to ISO 14971 for risk management. FMEA is commonly used as one technique within that framework rather than as the framework itself. Do not confuse the two: satisfying ISO 14971 requires more than producing an FMEA.
Where FMEA sits in the wider family
ISO 31010 catalogues risk assessment techniques and positions FMEA and FMECA among alternatives including fault tree analysis, bowtie analysis and HAZOP. It is a useful reference when you need to justify why you chose FMEA rather than something else.
Comparison at a glance
| Variant | Analyses | Prioritises by | Typical home |
|---|---|---|---|
| System FMEA | Subsystem interactions | RPN or Action Priority | Early-stage engineering |
| Design FMEA | Product design | RPN or Action Priority | Product development |
| Process FMEA | Manufacturing or service process | RPN or Action Priority | Operations, any sector |
| FMECA | Components and systems | Criticality | Defence, aerospace |
| AIAG-VDA | Design or process, seven steps | Action Priority | Automotive supply chain |
| Healthcare FMEA | Clinical processes | Hazard score | Hospitals, care providers |
How to choose
In practice the decision is usually made for you, and the order of questions is this:
- Does a customer, regulator or standard specify a method? If you are in the automotive supply chain, use AIAG-VDA. If you are on a defence contract citing MIL-STD-1629A, use FMECA. This question overrides everything below.
- Do you have reliability data? If yes, quantitative criticality is available and worth using. If no, do not pretend otherwise.
- Are you analysing a design or a process? Choose DFMEA or PFMEA accordingly. If in doubt, and you are not designing a physical product, it is a PFMEA.
- Is anything on your list potentially catastrophic? If yes, use a severity-first prioritisation such as Action Priority, or at minimum add a rule that any severity of 9 or 10 gets reviewed regardless of its RPN.
If none of those apply and you are simply trying to get better at managing risk, use the classic method described in our main guide. It is the easiest to teach, the easiest to run in a workshop, and it delivers most of the value.
One practical warning about scales
The detailed rating scales in the AIAG-VDA handbook, the SAE standards and the IEC and ISO documents are copyrighted material published by those bodies. You cannot lawfully copy them out of a colleague's PDF into your own template.
If you are contractually required to use a particular method, buy the handbook. If you are not, write your own scale descriptions in your own words, anchored to examples from your own business. Locally written anchors produce more consistent scoring than generic ones anyway, because "a complaint from a key account" means something concrete to your team in a way that an abstract severity description does not.
Where to go next
Whichever variant you land on, the mechanics of running the analysis are broadly the same, and the failure modes of the method itself are identical: doing it alone, discarding rare events, scoring detectability backwards and never reviewing the result. Those are covered in the main FMEA guide.
We work through FMEA and FMECA in detail, alongside the wider KPI and measurement toolkit, on the KPI Black Belt programme.
Learn FMEA properly
FMEA and FMECA are covered in depth on the KPI Black Belt programme, alongside the rest of the measurement toolkit.
