How to calculate RPN and Action Priority
Every failure mode in an FMEA gets three ratings from 1 to 10, with 10 always the worst case:
RPN = S × O × D
| Rating | Question | 10 | 1 |
|---|---|---|---|
| S, severity | How bad is the effect? | a safety risk to the operator or the end user | no noticeable effect |
| O, occurrence | How often does the cause happen? | no prevention control | prevention rules the cause out (for example, by fixture design) |
| D, detection | Do current controls catch it? | no inspection or test | the failure cannot be produced, or is always caught |
The 2019 AIAG-VDA handbook’s Action Priority (AP) uses the same three ratings, but through a table instead of a multiplication. Severity decides first, then occurrence, and only then detection:
| S | O | D 7 to 10 | D 5 to 6 | D 2 to 4 | D 1 |
|---|---|---|---|---|---|
| 9 to 10 | 6 to 10 | H | H | H | H |
| 9 to 10 | 4 to 5 | H | H | H | M |
| 9 to 10 | 2 to 3 | H | M | L | L |
| 7 to 8 | 8 to 10 | H | H | H | H |
| 7 to 8 | 6 to 7 | H | H | H | M |
| 7 to 8 | 4 to 5 | H | M | M | M |
| 7 to 8 | 2 to 3 | M | M | L | L |
| 4 to 6 | 8 to 10 | H | H | M | M |
| 4 to 6 | 6 to 7 | M | M | M | L |
| 4 to 6 | 4 to 5 | M | L | L | L |
| 2 to 3 | 8 to 10 | M | M | L | L |
| any | 1 | L | L | L | L |
Combinations not in the table (S 1, S 2 to 3 with O up to 7, S 4 to 6 with O up to 3) are all L.
Worked example
The values already in the tool come from a PFMEA of a cell that drills, deburrs, assembles and tightens a bracket, then leak tests, paints and labels it:
| Failure mode | S | O | D | RPN | AP |
|---|---|---|---|---|---|
| Washer missing | 7 | 5 | 7 | 245 | H |
| Torque below spec | 9 | 4 | 2 | 72 | H |
| Burr left in the hole | 5 | 7 | 6 | 210 | M |
| Scratched paint | 4 | 6 | 7 | 168 | M |
| Badly printed label | 3 | 7 | 10 | 210 | L |
| Hole above tolerance | 6 | 5 | 5 | 150 | L |
| Leak not caught by the test | 8 | 3 | 3 | 72 | L |
Take the washer: 7 × 5 × 7 = 245. In the AP table, S 7 to 8 with O 4 to 5 and D 7 to 10 gives H.
Now compare two rows. The label has an RPN of 210, one of the highest on the list, but AP L: the worst that happens is the customer typing the code by hand. The torque has an RPN of only 72, but AP H: a bolt that works loose in use is severity 9. With an “RPN over 100” cutoff, the torque would get no action and the label would eat the team’s time. AP turns that around.
The torque stays H even with detection 2. With severity 9 and occurrence 4, only detection 1 (the failure cannot get through) would lower the AP. The sounder route is to bring occurrence down to 3 or less: with D 2, the AP drops to L. Catching the failure is not enough, you have to stop the cause.
How to read the result
- High AP (H): the team needs to act to improve prevention or detection, or document why current controls are enough.
- Medium AP (M): the team should act to improve the controls, or justify why not.
- Low AP (L): the team could act, but it is not required.
- Within each AP, the list is sorted by RPN and then by severity, which act as tie-breakers.
- To lower the AP, work in the table’s order: severity only changes with a design or process change, occurrence drops with prevention (fixtures, poka-yoke, maintenance) and detection drops with automatic in-station inspection.
After the actions, rate the failure again and check whether the AP dropped. To choose which failures to analyze first, a Pareto chart of defects or complaints helps. When detection depends on a measurement, check process capability in the Cpk calculator and monitor it with a control chart.
Common mistakes
- Using a fixed RPN cutoff. S 10 × O 2 × D 5 and S 2 × O 10 × D 5 both give an RPN of 100, but one is a safety risk and the other a frequent nuisance.
- Lowering the detection rating to dodge the action. A visual check is not detection 2. Use the handbook scale and how the control really performs.
- Rating occurrence by how often the defect reaches the customer. In a PFMEA, occurrence is about the cause, given the prevention controls in place today.
- Lumping several effects into one row. Rate the severity of the worst effect of each failure mode, or split it into rows.
- Filling it in and filing it away. An FMEA is only worth it if the recommended actions get done and the ratings are reviewed afterward.