Three scores
| Score | Its question | Scale |
|---|---|---|
| Severity (S) | If the failure happens, what is the effect? | 1 no effect → 10 hazardous without warning |
| Occurrence (O) | How often is it expected? | 1 remote → 10 very high |
| Detection (D) | Can it be detected before its effect appears? | 1 easily detected → 10 undetectable |
RPN = S × O × D — a result between 1 and 1000.
There is no single threshold
This is the most commonly misrepresented aspect of the method. Published sources use different limits: some take 200 and 500, some 100 and 150, others 100 and 500. So “an RPN of 125 is acceptable” means nothing without saying which set it refers to. The organisation chooses, writes down and justifies its own threshold.
The trap in the product
A failure with severity 10, occurrence 1 and detection 2 has an RPN of 20 — it looks “low”. Yet any failure with severity 10 must be handled separately regardless of RPN. In mature practice RPN is therefore not the only criterion; a separate rule is set for severity.
Where it works well
FMEA is strong on equipment and process reliability: how a component fails, what happens when it does, and when we would notice. It feeds maintenance strategy, critical spare part decisions and inspection frequency directly.
Adding criticality
With criticality analysis added, the method ranks failure modes not only by score but by the weight of the consequence; that is how safety-critical components end up on a list of their own.