FMEA for Practitioners: Anticipate Failures Before They Occur

Run a process FMEA the AIAG-VDA way: the 7 steps, severity, occurrence and detection scales, action priority, and a sealant application example.

17 min read 3 sections Guide 14 of 20, Lean Master Practitioner Updated

S × OD
FMEARPN = severity × occurrence × detection

New to this topic?

Read these first to get the most out of this guide:

S×O×D
Classic RPN
1 to 10
Rating Scale
H/M/L
Action Priority
Living
Document Status

The FMEA Process: 7 Steps

Poka-yokeMake the wrong way impossible, not just warned againstThe Poka-yoke guide

Step 1: Define Scope

What process or product are you analyzing? A process FMEA (PFMEA) examines manufacturing steps. A design FMEA (DFMEA) examines the product design. For lean practitioners, PFMEA is most common. Define the boundaries: which operations are included?

Step 2: Identify Process Steps

List every operation in the process. For each operation, identify: its function (what it should do), its requirements (how you know it is done correctly).

Step 3: Identify Failure Modes

For each operation: how could it fail? What could go wrong? A failure mode is the way the operation fails to meet its requirement. Example: “Hole drilled undersize,” “Sealant applied with insufficient fillet,” “Fastener installed with incorrect torque.”

Step 4: Determine Effects and Severity

For each failure mode: what is the impact on the customer (next operation, final assembly, or end user)? Rate Severity on a 1 to 10 scale. S=1 is no effect. S=9 to 10 is safety hazard or regulatory violation. In aerospace, structural failures and FOD are typically S=9 or 10.

Step 5: Identify Causes and Occurrence

For each failure mode: what causes it? Rate Occurrence (likelihood) on a 1 to 10 scale. O=1 means nearly impossible. O=10 means almost certain. Base ratings on actual data (defect rates, historical occurrence) when available, not gut feel.

Step 6: Identify Current Controls and Detection

What controls exist to prevent the cause or detect the failure? Rate Detection on a 1 to 10 scale. D=1 means the control will almost certainly detect. D=10 means no control exists or detection is impossible. SPC improves detection. Poka-yoke improves both prevention and detection.

Step 7: Determine Action Priority and Assign Actions

Use the S×O×D Action Priority table (or RPN if your organization uses the older method) to prioritize. Assign specific actions with owners, deadlines, and verification methods. After implementation, re-rate to confirm the risk has been reduced.

Rating Scales

SafetyHazard tape lays across the aisle and the walkway lines mark the safe pathThe Safety guide
RatingSeverityOccurrenceDetection
1No effectNearly impossible (≤1 in 1,000,000)Almost certain to detect
2 to 3Minor annoyanceVery unlikelyHigh chance of detection
4 to 6Moderate impact, rework neededOccasionalModerate detection capability
7 to 8Major impact, customer affectedFrequentLow detection capability
9 to 10Safety hazard / regulatoryVery frequent / certainNo detection / cannot detect

Worked Example: Sealant Application PFMEA

STEP · KEY POINT · WHY
Standard workThe current best way, written down, so it can be improvedThe Standard work guide
Process FMEA: Wing Panel Sealant Application Aerospace
Failure ModeEffectSCauseOCurrent ControlDRPNAction
Insufficient fillet radiusCorrosion path at joint; rework7No standard work for technique6Visual inspection at next station5210Create standard work + TWI training
Sealant uncuredJoint failure in service; safety9Shop temp below cure minimum3Inspector checks adhesion4108Add temp monitoring + auto-alert
Foreign object in sealantFOD risk; corrosion initiation8Dirty application area4Pre-seal clean per SOP6192Add poka-yoke clean verification step
Wrong sealant type appliedChemical incompatibility; rework8Multiple sealant types at station2Operator reads work order7112Color-code sealant tubes by type; poka-yoke storage

Priority order (by RPN): Insufficient fillet (210) → FOD (192) → Wrong type (112) → Uncured (108). The highest-RPN item gets action first: standard work and TWI training for sealant application technique.

The FMEA Is a Living Document

The most common FMEA failure: it is created during APQP, filed in a binder, and never updated. A real FMEA is updated when: (1) A new failure mode is discovered in production. (2) A customer complaint reveals an undetected failure. (3) A process change adds new failure possibilities. (4) Countermeasures are implemented and ratings need to be re-scored. Review the FMEA at least quarterly or whenever the process changes.

The Bottom Line

FMEA is proactive quality: anticipating failures before they occur instead of reacting to defects after. The 7-step process identifies failure modes, rates their risk (Severity × Occurrence × Detection), and prioritizes prevention actions. Focus on high-severity items first (regardless of occurrence), drive occurrence down with process improvement, and drive detection up with SPC and poka-yoke. Keep the FMEA alive by updating it with field data and new discoveries. Next: Poka-Yoke Design, error-proofing techniques that eliminate failure modes at the source.

Frequently asked questions

What is FMEA?

Failure Mode and Effects Analysis is a systematic method for identifying how a process or product can fail (failure modes), what the consequences are (effects), and what causes the failure (mechanisms). Each failure mode is rated on Severity (impact), Occurrence (likelihood), and Detection (ability to catch it). The ratings drive prioritization of prevention and detection actions.

What is the difference between RPN and Action Priority?

The traditional approach multiplied Severity × Occurrence × Detection to get a Risk Priority Number (RPN). The newer AIAG/VDA methodology uses an Action Priority (AP) table that considers the interaction between S, O, and D more intelligently. For example, a high-severity failure mode always gets high priority regardless of occurrence, because the consequence is unacceptable even if rare. RPN could mask this (S=9, O=1, D=1 = RPN of 9, which appears low).

When should FMEA be done?

Before production starts (during process design/planning), and updated whenever the process changes, a new failure mode is discovered, or a quality escape occurs. An FMEA that is written once and filed is worthless. A living FMEA that is updated with field data and drives ongoing improvement is invaluable.

Share this guide: LinkedIn X Follow SymplProcess on LinkedIn

Take this to a room

The running order

For an FMEA team. They should leave with the seven steps, honest ratings, and a plan to keep it alive.

7 beats · 13 min
  1. 1

    Seven steps, in order

    Scope, steps, failure modes, effects, causes, controls, actions. Skipping any one weakens all of them.

    • Process FMEA for manufacturing steps. Design FMEA for the product.
    • Define the boundaries first: which operations are in and which are out.
    • Each operation needs its function and its requirement stated before you can say how it fails.

    Ask the room Do we have a current PFMEA for this process?

  2. 2

    A failure mode is specific

    Vague failure modes produce vague actions, and this is where most FMEAs go soft.

    • Hole drilled undersize. Sealant applied with insufficient fillet. Fastener installed without washer.
    • Not quality problem or process issue.
    • If you cannot picture it happening, it is not written specifically enough.
  3. 3

    Severity comes first and stands alone

    High severity items get attention regardless of how rarely they happen.

    • S of 9 or 10 is a safety or regulatory hazard. Structural and flight-critical items live here.
    • You cannot reduce severity by making the failure rarer, only by changing the design.
    • Work the severity list before the RPN list.
  4. 4

    Rate from data where it exists

    Occurrence and detection are the two ratings people invent, and inventing them makes the ranking meaningless.

    • Occurrence from actual defect rates and history.
    • Detection from what the control genuinely catches. Operator training is a weak control; rate it that way.
    • SPC and error-proofing improve detection and occurrence respectively.
  5. 5

    Action priority, not just RPN

    The newer AIAG/VDA method ranks by the combination rather than the product, and it is better.

    • RPN treats 9x2x2 and 2x9x2 as identical. They are not.
    • Action Priority high, medium, low reflects severity properly.
    • Use whichever your organisation uses, but know why the newer one exists.
  6. 6

    Actions need owners and verification

    An action without a name and a date is a note.

    • Specific countermeasure, named owner, deadline, and how it will be verified.
    • After implementation, re-rate. Did occurrence actually fall?
    • Verification with data, not with an assertion that it was done.
  7. 7

    Keep it alive

    The most common FMEA failure is that it is created during launch and never opened again.

    • Update when a new failure mode appears in production.
    • When a customer complaint reveals something undetected.
    • When the process changes. When a countermeasure lands and ratings need re-scoring.
    • Review quarterly at minimum.

    Ask the room When was ours last updated, and what triggered it?