FMEA Severity, Occurrence & Detection: Complete Rating Guide
Severity, Occurrence, and Detection are the three ratings that drive every FMEA risk evaluation. Getting them right determines whether your FMEA process identifies real risks or produces misleading results.
💡 Key insight: Severity drives Action Priority — not RPN. A failure with S=9 always gets High AP, regardless of Occurrence and Detection values.
How S, O, D Work Together
In the AIAG-VDA methodology, these three ratings feed into the Action Priority (AP) table:
- Severity (S) — How bad is the effect if this failure reaches the customer?
- Occurrence (O) — How likely is the cause to happen?
- Detection (D) — How likely are current controls to catch the failure before it escapes?
Severity Scale (1–10)
Severity Rating Table
| Rating | Criteria | Example |
|---|---|---|
| 10 | Hazardous without warning | Brake failure — vehicle cannot stop |
| 9 | Hazardous with warning | Steering lock-up with warning light |
| 8 | Very high — inoperable | Engine fails to start |
| 7 | High — performance severely affected | Excessive vibration during operation |
| 6 | Moderate — partial function loss | Intermittent electrical contact |
| 5 | Low — reduced performance | Increased noise level beyond spec |
| 4 | Very low — minor impact | Paint blemish on non-visible surface |
| 3 | Minor — slight inconvenience | Slightly rough surface finish |
| 2 | Very minor — barely noticeable | Cosmetic variation within tolerance |
| 1 | None — no discernible effect | No impact on form, fit, or function |
Key principle: Severity is rated from the customer's perspective, not the manufacturer's. Severity can only be reduced by changing the product design—process changes cannot lower severity.
Occurrence Scale (1–10)
Occurrence Rating Table
| Rating | Criteria | Example |
|---|---|---|
| 10 | Extremely high: inevitable | ≥ 100 per 1,000 (≥ 10%) |
| 9 | Very high | 50 per 1,000 (5%) |
| 8 | High: repeated failures | 20 per 1,000 (2%) |
| 7 | High | 10 per 1,000 (1%) |
| 6 | Moderately high | 5 per 1,000 (0.5%) |
| 5 | Moderate | 2 per 1,000 (0.2%) |
| 4 | Moderately low | 1 per 1,000 (0.1%) |
| 3 | Low | 0.5 per 1,000 (0.05%) |
| 2 | Very low | 0.1 per 1,000 (0.01%) |
| 1 | Remote: failure unlikely | ≤ 0.01 per 1,000 (≤ 0.001%) |
Key principle: Occurrence rates the cause, not the failure mode. Use historical data (scrap rates, PPM, warranty data) whenever available. Prevention controls reduce occurrence.
Detection Scale (1–10)
Detection Rating Table
| Rating | Criteria | Example |
|---|---|---|
| 10 | No detection opportunity | No inspection or test exists |
| 9 | Not likely to detect | Indirect / random check only |
| 8 | Remote chance of detection | Visual inspection of hidden feature |
| 7 | Very low detection | Manual inspection / sampling plan |
| 6 | Low detection | Control chart with delayed feedback |
| 5 | Moderate detection | Gauge / go-no-go check |
| 4 | Moderately high detection | Statistical process monitoring |
| 3 | High detection | Automated in-line measurement |
| 2 | Very high detection | Automated error-proofing (poka-yoke) |
| 1 | Almost certain detection | Failure is prevented by design |
Key principle: Detection rates the current controls' ability to find the failure. Note that the scale is inverted — lower is better for detection.
⚠️ Common confusion: SPC is a prevention control (reduces Occurrence). A final inspection is a detection control. Mixing them inflates or deflates ratings.
Common Rating Mistakes
- Rating severity based on process impact, not customer impact. "Scrap the part" is not a severity criterion.
- Using O=1 for everything. "We've never seen this failure" doesn't mean O=1.
- Consensus bias. Teams average ratings to avoid conflict instead of rating individually.
- Not updating ratings. After implementing corrective actions, re-rate O and D.
Practical Tips for Accurate Ratings
- Use data: scrap reports, customer complaints, warranty returns, process capability studies (Cpk)
- Rate severity first—it rarely changes and anchors the entire analysis
- For occurrence, ask: "How many times per 1,000 units would this cause create this failure?"
- For detection, ask: "If this failure occurred right now, would our current controls catch it?"
- See a complete example: CNC machining P-FMEA
From Ratings to Action Priority
Once S, O, and D are assigned, the AIAG-VDA AP table determines whether the risk requires action. The Qhubio FMEA tool applies this table automatically, ensuring consistent AP assignment across your entire analysis.