AIAG & VDA FMEA: The Harmonised Method

For years North American and German automotive supply chains ran two different FMEA conventions. Suppliers serving both did the analysis twice. The harmonised AIAG-VDA method exists to end that.

The seven steps

The most visible change is that FMEA is now described as a seven-step process rather than a form to fill in — an attempt to stop it being treated as a document produced for an audit.

1. Planning and preparation — define scope, boundaries, the team, and what this analysis is and is not covering. Skipped constantly, and the cause of most FMEAs that sprawl.
2. Structure analysis — break the system or process into its elements: the system, subsystem and component levels, or the process, station and work-element levels.
3. Function analysis — what each element is supposed to do. You cannot describe a failure without first stating the function it fails to deliver.
4. Failure analysis — failure modes, effects and causes, at the levels the structure established.
5. Risk analysis — existing prevention and detection controls, and the evaluation of severity, occurrence and detection.
6. Optimisation — actions to reduce risk, with owners, dates, and evidence of what the action achieved.
7. Results documentation — communicating what was found and decided.

Action Priority instead of RPN

The old approach multiplied severity, occurrence and detection into a single risk priority number, and teams then set a threshold — "act on anything over 100". Two problems with that.

First, the multiplication is arithmetically meaningless: the scales are ordinal, not ratio, so their product is not a quantity. Second, and more damaging in practice, very different situations produce the same number. Severity 10, occurrence 2, detection 5 gives 100. So does severity 4, occurrence 5, detection 5. The first could hurt somebody. Under a threshold rule they receive identical attention.

Action Priority instead uses a lookup across the three ratings that produces High, Medium or Low — and it weights severity properly, so a high-severity item cannot be argued down to Low by claiming good detection. It is a table, not a formula, which is precisely the point: it removes the opportunity to game a number.

What it means for FMEAs you already have

Existing FMEAs do not become invalid. The usual pragmatic approach is to convert at the next substantive revision rather than reworking an archive that is not changing — but check your customer’s customer-specific requirements, because some specify the harmonised method for new programmes explicitly, and that overrides your preference.

Also worth planning for: the structure and function analysis steps require information many existing FMEAs simply do not contain. Converting is not a column remapping exercise. It is usually a re-analysis, and it is usually worth it, because the missing steps are the ones that made the original analysis vague.

What software can and cannot do here

Software can hold the structure, enforce the workflow, keep revisions straight, and carry the results into the control plan. It cannot supply engineering judgement, and any tool suggesting it can generate your risk analysis for you should be treated with suspicion — an FMEA is a record of what your team decided and why, and that is the part with the value.

One practical note on licensing: the AIAG-VDA rating tables are copyrighted material licensed to you. Reputable software implements the method and the workflow and lets you reference your licensed tables; it does not ship the manual’s content inside the product.