← Insights

Risk

DFMEA vs PFMEA: which one, when

July 2026 · 6 min read

Two FMEAs, one question that keeps coming up: which one do we need? The short answer is that they analyse different things and, on any product you design and manufacture, you need both. The longer answer - what goes in each, how they hand over to each other, and the mistake that quietly ruins both - is this article.

Same engine, different subject

Mechanically, a Design FMEA and a Process FMEA are identical. Both list failure modes, effects and causes. Both score Severity, Occurrence and Detection 1-10 against published scales. Both produce an RPN and, since the AIAG-VDA handbook of 2019, an Action Priority of High, Medium or Low. If you can run one, you can run the other.

What differs is the subject under analysis:

A one-line test that settles most arguments: if the failure would still occur with a perfect operator on a perfect line, it belongs in the DFMEA. If a design made exactly to print can still end up failing in the customer's hands because of how it was made, that is the PFMEA's territory.

The handover between them

The two documents are not independent - they hand risk to each other in a specific way, and the handover is where mature organisations get their leverage.

Severity flows down. If the DFMEA says a failure of this feature puts the end user at risk - a Severity 9 or 10 - then every process step that produces that feature inherits that severity in the PFMEA. The process team does not get to re-negotiate how much a brake component matters.

Special characteristics flow down. The DFMEA identifies the dimensions and properties where variation threatens function. Those become the special characteristics the control plan must control, and the PFMEA must show how the process protects them.

Process reality flows up. When the PFMEA keeps finding that a feature is genuinely hard to make - a tolerance the process cannot hold, an assembly that invites error - that is design feedback. The best DFMEA reviews have a manufacturing engineer in the room saying "you can design it, we cannot build it".

The classic mistake: one FMEA trying to be both

The most common FMEA defect in the wild is a single document with design failure modes and process failure modes interleaved. It feels efficient. It ruins both analyses:

• The scoring scales stop making sense - Occurrence for a design weakness (rated against design life) and Occurrence for a process slip (rated against parts per million) are different animals wearing the same 1-10 costume.
• Ownership evaporates. Design actions land on process engineers and vice versa, and both quietly stall.
• Auditors notice. IATF 16949 and most OEM customer-specific requirements expect the two analyses as separate, linked documents.

Keep them separate, keep them linked, and let each one be honestly itself.

If you only make, or only design

Contract manufacturers building to a customer's print usually run only a PFMEA - the design risk belongs to the customer, though a sharp supplier still feeds manufacturability findings back. Design houses that outsource all production run the DFMEA and require the PFMEA from their manufacturing partner. The question is never "which is better" - it is "which risks are yours".