Interpreting FMEA Results: Why High RPN Is Not Always the Biggest Risk

Two rows on the same PFMEA. Row A has Severity 10, Occurrence 1, Detection 1 — RPN of 10. Row B has Severity 2, Occurrence 10, Detection 8 — RPN of 160. The team prioritizes Row B because the RPN is higher. Six months later the rare event in Row A actually happens, and the safety incident shows up in warranty data with a much higher cost than the steady-state nuisance Row B addressed. This is the most damaging pattern in legacy RPN-based FMEA work, and it is why the AIAG-VDA Handbook replaced RPN with Action Priority in 2019. Practitioners interpreting FMEA results need to recognize when high RPN is misleading and when it accurately represents the biggest risk.

This guide covers what high RPN with low Severity actually means, why the multiplication-based ranking misranks safety-critical failures, the AP-vs-RPN interpretation rules, and the specific patterns that show up when a team has been working from RPN-only thinking.

The RPN Math Problem

RPN is calculated as Severity × Occurrence × Detection, each rated 1–10. Range: 1–1000. The intent: a single number that ranks failure-mode risk for resource allocation. The flaw: multiplying three ordinal scales treats them as equally weighted, which conflates fundamentally different risk profiles.

Two Failure Modes With Identical RPN

Failure A: S=10, O=1, D=1 → RPN = 10. Catastrophic effect; nearly impossible occurrence; almost-certain detection.
Failure B: S=2, O=10, D=1 → RPN = 20. Negligible effect; persistent occurrence; almost-certain detection.
Failure C: S=2, O=2, D=10 → RPN = 40. Negligible effect; rare; no detection method exists.

RPN ranks Failure C highest. AP ranks Failure A highest. AP is correct: a Severity-10 failure means safety-critical or regulatory exposure, and the rare-but-detected probability does not change the consequence if it occurs.

The legacy advice from AIAG 4th Edition (2008) was to set RPN action thresholds (e.g., “take action on any RPN above 100”), which institutionalized this exact mistake. AIAG-VDA 2019 dropped the RPN threshold model and replaced it with the AP logic-based table.

How AP Reweights the Same Inputs

Action Priority categorizes each S/O/D combination as High (H), Medium (M), or Low (L) using a logic-based lookup that prioritizes Severity first, then Occurrence, then Detection. The key behaviors:

Severity 9-10 — almost always High AP regardless of Occurrence and Detection. A safety-critical or regulatory failure cannot be deprioritized by “rare” or “detected.”
Severity 7-8 — usually High or Medium AP unless both Occurrence and Detection are very low.
Severity 1-3 — usually Low AP regardless of how high Occurrence and Detection get. A negligible-effect failure that happens often and goes undetected is still a negligible-effect failure.

The AP table’s logic encodes the practitioner judgment that severity is not commutative with occurrence or detection. Multiplying S=10 by O=1 and D=1 produces a small number, but the actual risk is large because the effect is catastrophic. Multiplying S=2 by O=10 and D=10 produces a large number, but the actual risk is small because the effect is negligible.

Common Interpretation Error “Sort the FMEA by RPN descending and work from the top.” This was the AIAG 4th Edition workflow, and many quality teams still apply it from habit. Under AIAG-VDA, sort by AP (H, M, L) first, and within each tier sort by Severity descending. A Severity 10 / Low AP row outranks a Severity 4 / Medium AP row in any case where both cannot be addressed.

How to Interpret a High-RPN Low-Severity Row

When the FMEA shows RPN 200 with Severity 4, the row represents a frequent or hard-to-detect failure with a non-critical effect. The right interpretation:

Check the AP rating. A high-RPN low-Severity row will typically be Medium AP, not High. The Medium classification is correct — the row deserves action, but not at the expense of any High AP row regardless of the latter’s RPN.
Look for shared causes. High-RPN low-Severity rows often cluster around the same root cause (a process capability shortfall, an inadequate detection method). One action targeting the shared cause can drop multiple Mediums simultaneously.
Quantify the cost of inaction. Negligible-effect failures still cost something at scale — rework, scrap, inspection burden, customer dissatisfaction. The business case for action is in the cost-of-quality numbers, not the RPN.
Do not treat it like a Severity-9 row. Resources are finite. A Medium-AP row with RPN 200 has lower priority than a High-AP row with RPN 60.

Common Patterns From RPN-Only Thinking

Severity-9 rows hidden in Low-RPN territory. Teams skip them because the RPN is “only 36.” Audit finding waiting to happen.
Inflated Detection ratings to lower RPN. Some teams game the rating to bring RPN below the action threshold rather than do the underlying work. AP closes this gap because Severity drives the classification regardless of Detection.
Action-threshold theater. The team takes action on RPN over 100, ignores everything below, and reports closure rate. Auditors read the FMEA and find Severity-9 / RPN-50 rows the team marked “below threshold.”
Equal weighting of S, O, D in team discussion. Time spent debating whether Detection is 6 or 7 when Severity is the dominant axis. The AIAG-VDA correction: spend the time getting Severity right, then Occurrence, then Detection — in that order of effort.

When High RPN Really Is the Biggest Risk

RPN is not always wrong. When Severity, Occurrence, and Detection are all in the upper range (S≥7, O≥5, D≥5), high RPN is a real signal — this is a high-Severity failure that occurs often and is hard to detect. AP and RPN agree at this end of the distribution. Where they diverge is in the asymmetric cases (catastrophic-but-rare, frequent-but-negligible) and that is where the practitioner has to read the FMEA on the AP scale, not the RPN scale.

Practical Rule For any Severity-9 or Severity-10 row, the RPN is irrelevant. You take action regardless. For any Severity-1 to Severity-3 row, the RPN is also largely irrelevant — the effect is too small to drive action even at high Occurrence. RPN is most informative in the middle Severity band (4-8), and even there AP gives the cleaner ranking.

Communicating Results to Non-Practitioners

Quality managers, design engineers, and customer auditors who haven’t read the AIAG-VDA Handbook may anchor on RPN out of habit. Three communication patterns that work:

Lead with AP H/M/L when reporting status (“eight High AP items, twelve Medium, thirty-five Low”), not with RPN distribution.
Report Severity-9-10 row count separately. These are the safety-critical chains regardless of how the team classifies overall risk.
When stakeholders ask for RPN, provide it but always alongside AP. “Row 27: RPN 80, AP High — Severity 9 with low Occurrence and good Detection” tells the right story; “RPN 80” alone tells the wrong one.

For the underlying methodology of how Severity, Occurrence, and Detection ratings combine, see how RPN is calculated and what S, O, D ratings mean. For the AP-vs-RPN methodology shift in detail, see why AIAG-VDA replaced RPN with Action Priority and how to apply the AP table. For the Severity scale that drives the special-characteristic threshold, see the 1-10 Severity scale for manufacturing failures.

Summary

High RPN does not equal biggest risk. RPN multiplies three ordinal scales as if they were equally weighted, which lets catastrophic-but-rare failures hide in low-RPN territory while frequent-but-negligible failures top the action list. AP corrects this by prioritizing Severity first — a Severity-10 failure is High AP regardless of Occurrence and Detection. When interpreting FMEA results, sort by AP first, Severity second, and use RPN as a secondary signal in the middle Severity band. Teams still working from RPN-only thinking ship Severity-9 rows in “low priority” territory and find out about the misranking through warranty data, recalls, or audit nonconformances.

To recompute AP from S/O/D inputs without manually walking the AIAG-VDA lookup table, use the RPN and Action Priority calculator.