FMEA
Prevent Before It Happens
RPN
Risk Priority Number
S×O×D
Severity × Occurrence × Detection
Living
Document, Not One-Time

What Is FMEA?

Failure Mode and Effects Analysis is a structured method for identifying what could go wrong in a process or product design, assessing the risk of each potential failure, and prioritizing actions to prevent the highest-risk failures. It is proactive problem solving — finding and fixing problems before they happen, not after.

FMEA was developed by the U.S. military in the 1940s, adopted by NASA, and became a standard requirement in automotive (AIAG) and aerospace manufacturing. Today it is used across every industry where failure has consequences.

Two Types of FMEA

TypeFocusWhenWho Leads
Design FMEA (DFMEA)What could go wrong with the product design?During product development, before release to productionDesign / product engineering
Process FMEA (PFMEA)What could go wrong in the manufacturing process?Before production launch, updated throughout production lifeManufacturing / process engineering

This guide focuses on Process FMEA, which is what most manufacturing teams work with daily. The methodology is the same for both types.

The FMEA Structure

For each process step, the team identifies:

Potential Failure ModeWhat could go wrong? How could this step fail? Examples: part loaded backwards, hole drilled off-location, wrong torque applied, weld incomplete, label missing.
Potential EffectIf this failure happens, what is the impact? On the next operation, on the final product, on the customer, on safety. Be specific: "Bearing fails in the field within 6 months" not just "quality issue."
Potential CauseWhy would this failure occur? What in the process allows it? Wrong fixture, worn tool, operator error, material variation, machine setting drift. This is where root cause thinking applies.
Current ControlsWhat is currently in place to prevent or detect this failure? Poka-yoke devices, inspections, SPC charts, visual checks, automated sensors. Be honest — "operator training" is a weak control.

RPN Scoring

Each failure mode is scored on three dimensions, each rated 1-10:

RatingSeverity (S)Occurrence (O)Detection (D)
1No effectExtremely unlikely (<1 in 1M)Almost certain to detect
2-3Minor annoyance / cosmeticVery low (1 in 100K)High probability of detection
4-6Moderate — customer dissatisfied, partial function lossModerate (1 in 1,000)Moderate detection capability
7-8High — major function loss, customer very dissatisfiedHigh (1 in 100)Low detection — likely to escape
9-10Safety/regulatory — hazard without warningVery high (1 in 10+)No detection — will reach customer

RPN Formula

RPN = Severity × Occurrence × Detection

Range: 1 to 1,000. Higher RPN = higher risk = higher priority for action. But do not use RPN blindly — a failure with Severity 10 (safety) and low occurrence still needs attention regardless of the total RPN.

FMEA Example

StepFailure ModeEffectCauseSODRPN
Install gasketGasket missingOil leak in field, warranty claimNo poka-yoke, operator skips step847224
Torque boltUnder-torquedJoint loosens, noise, potential safetyTorque wrench not calibrated935135
Apply labelWrong labelCustomer confusion, recall riskMultiple labels at station, no verification756210

Taking Action

The FMEA is useless without actions. For high-RPN items, the team assigns:

Action TypeTargetExample
Reduce SeverityChange design to make failure less impactfulAdd redundant seal so single gasket failure does not cause leak
Reduce OccurrencePrevent the cause from happeningInstall poka-yoke fixture that will not close without gasket
Reduce DetectionMake the failure easier to catch before it leavesAdd vision system to verify label after application

Prioritize Occurrence Over Detection

The best FMEA actions reduce occurrence (prevent the failure) rather than improve detection (catch it after it happens). Installing a poka-yoke (O drops from 4 to 1) is always better than adding an inspection step (D drops from 7 to 3). Prevention beats detection. See jidoka.

Running an FMEA Session

Assemble the right teamCross-functional: process engineer, quality engineer, operator, maintenance, design engineer if available. The operator's input is critical — they see failure modes that engineers miss.
Walk the process at the gembaDo not do FMEA from a conference room. Walk the actual process step by step. Observe. Ask "what could go wrong here?" at every station.
Score honestlyThe biggest FMEA failure is sandbagging scores to avoid work. If detection is weak, score it high. If occurrence is frequent, admit it. Honest scoring drives honest action.
Assign actions with owners and datesTop RPN items get specific countermeasures, specific owners, and specific due dates. Track them like any other action item.
Re-score after actions are completeDid the RPN actually drop? Verify with data. If the poka-yoke was installed, re-score occurrence. The FMEA is a living document — update it as the process changes.
✅ FMEA That Drives Improvement
  • Cross-functional team with operator input
  • Done at the gemba, not in a conference room
  • Honest scoring — high RPNs are opportunities
  • Actions prioritize prevention over detection
  • Living document updated with process changes
❌ FMEA Theater
  • Done by one engineer alone at a desk
  • Scores kept low to avoid creating work
  • Actions = "train operator" for everything
  • Completed once for PPAP and never updated
  • Sits in a binder for the auditor, not used by the team

🎯 Key Takeaway

FMEA is your crystal ball — it lets you see and prevent failures before they reach the customer. Build it with a cross-functional team at the gemba, score honestly, prioritize prevention over detection, and keep it alive as a working document. The best time to do an FMEA is before production launch. The second best time is today. Pair it with poka-yoke for the highest-RPN items and SPC for ongoing monitoring.

Interactive Demo

Adjust the Severity, Occurrence, and Detection ratings for each failure mode. Watch the RPN scores update and the priority ranking shift in real time.

⚡
Try It Yourself
FMEA Risk Assessment
▼
Rate each failure mode on Severity (1-10), Occurrence (1-10), and Detection (1-10). RPN = S × O × D. Higher detection scores mean HARDER to detect (worse). Focus on the highest RPNs first.
Failure ModeSeverityOccur.Detect.RPNRisk
Sensor drift
Incorrect readings
168Medium Risk
Seal leak
Product contamination
160Medium Risk
Bearing seizure
Line stoppage
135Medium Risk
Software glitch
Batch error
112Medium Risk
Belt wear
Reduced speed
72Low Risk
RPN Rankings
Sensor drift
168
Seal leak
160
Bearing seizure
135
Software glitch
112
Belt wear
72
168
Highest RPN
0
High Risk Items
Sensor drift
Top Priority
Ready for the full knowledge check? Test your understanding with guided scenarios and data export.
PROTake the Pro Knowledge Check →
Free forever · Every feature included

Stop reading, start modeling

Model your process flow, run simulations, optimize staffing with TOC math, and test your knowledge with 107 interactive checks — all in one platform.

Open Workbench → Pro Knowledge Check

Take this to a room

The running order

For a cross-functional team about to run an FMEA. They should leave knowing to score honestly, to walk the process, and to attack occurrence before detection.

7 beats · 12 min
  1. 1

    Failures before they happen

    FMEA is the one quality tool that runs before there is anything to investigate.

    • Design FMEA: what could go wrong with the product, during development.
    • Process FMEA: what could go wrong in manufacturing, before launch and forever after.
    • It is a living document. A one-time FMEA is a filed document.

    Ask the room When was our PFMEA last updated, and by what event?

  2. 2

    The four columns that matter

    Failure mode, effect, cause, current controls. Everything else is arithmetic.

    • Effect must be specific: bearing fails in the field within six months, not quality issue.
    • Cause is what in the process allows it - fixture, worn tool, setting drift.
    • Be honest about controls. Operator training is a weak control.
  3. 3

    RPN is severity times occurrence times detection

    Three one-to-ten scales multiplied. The scales are defined, so use the definitions.

    • Severity 9 to 10 means safety or regulatory, hazard without warning.
    • Occurrence 7 to 8 means about one in a hundred.
    • Detection 9 to 10 means it will reach the customer.
  4. 4

    A worked line

    One row from a real process makes the whole method concrete.

    • Install gasket. Failure: gasket missing. Effect: oil leak in the field, warranty claim.
    • Cause: no error-proofing, operator can skip the step.
    • S 8, O 4, D 7. RPN 224 - the top item on that sheet.
  5. 5

    Attack occurrence, not detection

    This is the recommendation people most often get backwards.

    • A poka-yoke that drops occurrence from 4 to 1 beats an inspection that drops detection from 7 to 3.
    • Prevention removes the failure. Detection just catches it later, at cost.
    • Reducing severity means changing the design - use it when you can.

    Ask the room Of our current countermeasures, how many are inspections?

  6. 6

    Score honestly or do not score

    The biggest FMEA failure is sandbagging the numbers to avoid the work they imply.

    • If detection is weak, score it high.
    • If occurrence is frequent, admit it.
    • Honest scoring is the only thing that makes the ranking mean anything.
  7. 7

    How to run the session

    Cross-functional, at the gemba, with owners and dates - then re-score.

    • Process engineer, quality, maintenance, and the operator. The operator sees modes engineers miss.
    • Walk the process step by step. Do not do this in a conference room.
    • Top RPN items get named owners and due dates.
    • Re-score after the action, with data. Did the RPN actually drop?

    Ask the room Who from the floor is in the room when we do this?