FMEA — Failure Mode and Effects Analysis — is the most widely used proactive risk assessment method in quality management. Originally developed by the U.S. military in 1949 and adopted by NASA, Ford, and the automotive industry, it is now embedded in ISO 9001, IATF 16949, and FDA medical device regulations. This guide explains every component of FMEA: the RPN formula, the rating scales, the difference between Design and Process FMEA, a full worked example, and the mistakes that make FMEA fail.
What Is FMEA?
Failure Mode and Effects Analysis (FMEA) is a structured, team-based method that asks three questions about every function in a product or process:
- How could this fail? (failure mode)
- What happens if it fails? (effect)
- How likely are we to prevent, detect, or catch that failure? (risk score)
Unlike reactive methods such as root cause analysis or the 5 Whys — which investigate failures after they happen — FMEA is a prevention tool. You run it before the product ships or before the process goes live. The output is a prioritized list of risks with assigned corrective actions.
FMEA was developed by the U.S. military (MIL-P-1629, 1949), refined by NASA for Apollo and space shuttle programs, adopted by the automotive industry through Ford in the 1970s, and standardized in AIAG’s FMEA manual (now in its 5th edition, developed jointly with VDA). Today it is a mandatory deliverable in automotive APQP, a best practice in medical devices, and widely used in aerospace, healthcare, and software.
The RPN Formula: Severity × Occurrence × Detection
Every failure mode in an FMEA is scored using three independent ratings, each on a 1–10 scale. Multiplied together, they produce the Risk Priority Number (RPN):
Higher RPN = higher risk. Most organizations trigger action at RPN ≥ 100–200, and always for any item with Severity ≥ 9 regardless of RPN.
Important: RPN is a prioritization tool, not an absolute measure. A failure mode with S=10, O=1, D=1 gives RPN=10 — but because the effect is catastrophic (S=10), it still demands attention. Always review high-Severity items separately from the RPN ranking.
FMEA Rating Scales
Severity (S) — How bad is the effect?
| Rating | Description | Example effect |
|---|---|---|
| 1 | No effect | Customer would not notice |
| 2–3 | Minor | Slight customer inconvenience; appearance defect |
| 4–6 | Moderate | Customer dissatisfied; partial loss of function |
| 7–8 | High | Major function inoperable; customer very dissatisfied |
| 9 | Very high | Safety concern; potential regulatory non-compliance |
| 10 | Hazardous | Safety failure without warning; regulatory violation |
Occurrence (O) — How likely is the failure?
| Rating | Description | Approximate failure rate |
|---|---|---|
| 1 | Almost impossible | < 1 in 1,500,000 |
| 2–3 | Low | 1 in 150,000 – 1 in 15,000 |
| 4–6 | Moderate | 1 in 2,000 – 1 in 400 |
| 7–8 | High | 1 in 80 – 1 in 20 |
| 9–10 | Very high | > 1 in 8 |
Detection (D) — How likely are we to catch it?
| Rating | Description | Example control |
|---|---|---|
| 1–2 | Almost certain | Automated 100% inspection; poka-yoke device |
| 3–4 | High | SPC with control charts; automated functional test |
| 5–6 | Moderate | Periodic sampling inspection; manual measurement |
| 7–8 | Low | Visual check only; no standardized method |
| 9–10 | Almost impossible | No detection control in place |
Note: For Detection, lower is better. A rating of 1 means you will almost certainly catch the failure; a rating of 10 means it will reach the customer undetected.
DFMEA vs PFMEA: Two Types of FMEA
The same RPN methodology is used in both types — what differs is the scope and when in the product lifecycle you run it.
DFMEA
Evaluates potential failures in the product design before manufacturing begins. Asks: could this design fail to meet its intended function? Scope: components, sub-assemblies, interfaces. Owned by: design/engineering team.
PFMEA
Evaluates potential failures in manufacturing or assembly processes. Asks: could this process step produce a defect or non-conformance? Scope: process steps, machines, operators. Owned by: manufacturing/process engineering.
Other variants include System FMEA (SFMEA, for complex system-of-systems), Machinery FMEA (MFMEA, for equipment reliability), and Software FMEA (for code and system software). In healthcare, a close cousin called Failure Mode and Effects and Criticality Analysis (FMECA) adds a criticality matrix on top of RPN scoring.
How to Run an FMEA in 7 Steps
- Define scope and assemble a cross-functional team. Decide DFMEA or PFMEA. Assemble 4–8 people across design, manufacturing, quality, and customer service. FMEA done solo is always incomplete — diverse perspectives surface failure modes that siloed experts miss.
- List all functions and requirements. For each component (DFMEA) or process step (PFMEA), write down what it is supposed to do. This is the anchor: a failure mode is any way this function could fail to be performed.
- Identify potential failure modes. Brainstorm every way each function could fail: complete failure, partial failure, failure at wrong time, unintended function, intermittent failure. Pull from warranty data, customer complaints, and similar past projects. Aim for completeness, not elegance.
- Determine effects and rate Severity (S). For each failure mode, identify the worst reasonable effect on the end user. Rate S 1–10 using the scale above. Severity describes the consequence, not the likelihood — rate it as if the failure actually occurred.
- Identify root causes and rate Occurrence (O). For each failure mode, list the most likely root cause. Rate O 1–10 based on historical failure data or team judgment. If you have field data, use it — it produces more accurate Occurrence ratings than estimates alone.
- List current controls and rate Detection (D). Document what exists today to prevent the cause or detect the failure before it reaches the customer: inspections, tests, SPC charts, poka-yoke devices. Rate D 1–10. If no control exists, D = 9 or 10.
- Calculate RPN, prioritize, and act. RPN = S × O × D. Sort descending. Review all items with RPN ≥ your threshold (commonly 100–200) AND any item with S ≥ 9. Assign corrective actions, owners, and due dates. After implementation, re-score O and D to confirm risk reduction.
FMEA Example: Injection Molding Process (Full Worksheet)
The following PFMEA covers selected process steps for an injection molding operation producing structural plastic brackets. This simplified example uses five failure modes to illustrate how the scoring and prioritization work in practice.
| Process Step / Function | Potential Failure Mode | Potential Effect | S | Potential Cause | O | Current Controls | D | RPN | Recommended Action |
|---|---|---|---|---|---|---|---|---|---|
| Injection / Fill cavity completely | Short shot (mold not fully filled) | Defective part — scrap, production delay | 7 | Insufficient injection pressure | 4 | Visual inspection 1/shift | 4 | 112 | Add in-cavity pressure sensor; set alarm at threshold |
| Injection / Fill cavity completely | Flash (excess material at parting line) | Dimensional non-conformance; assembly failure | 5 | Excessive injection pressure; worn parting line | 3 | Dimensional check (CMM), 2/shift | 3 | 45 | Calibrate pressure controller quarterly; inspect tooling monthly |
| Cooling / Cool part evenly to spec | Warpage (>0.5 mm flatness deviation) | Assembly failure; returned parts; customer complaint | 6 | Uneven cooling channel flow; blocked water line | 5 | CMM flatness check, 1/hour | 3 | 90 | Flush cooling channels weekly; install flow meters on cooling circuit |
| Ejection / Release part without damage | Part sticking to mold | Part damage; production stop; extended cycle time | 6 | Insufficient draft angle; worn ejector pins | 3 | Operator visual check each cycle | 4 | 72 | Apply mold release each 500 shots; add ejector pin wear PM schedule |
| Material prep / Dry pellets to <0.02% moisture | Wet pellets entering barrel | Splay defects; weakened part strength; scrap | 7 | Dryer failure; incorrect drying time; wrong material lot | 3 | Moisture meter spot check before run | 5 | 105 | Add dryer temperature alarm; require moisture log per lot before start |
Action summary: Two items exceed RPN 100 (short shot: 112, wet pellets: 105) and both receive immediate action. Warpage at RPN 90 is borderline — the team assigns a lower-priority action. Flash at RPN 45 is monitored but no urgent action needed.
More FMEA Examples by Industry
Automotive — Brake Caliper Assembly (PFMEA)
Failure mode: Brake fluid leak at caliper banjo fitting
Effect: Reduced braking force; potential accident — S = 9
Cause: Under-torqued fitting (operator skips torque wrench) — O = 2
Control: 100% torque audit station with torque wrench click confirmation — D = 2
RPN = 36 — Low RPN but S = 9 triggers immediate review. Action: add electronic torque tool with pass/fail interlock — if no confirmation signal, line stops.
This example illustrates why Severity ≥ 9 must always be reviewed regardless of RPN. A low RPN gives false comfort when the consequence is a safety failure.
Healthcare — IV Medication Administration (PFMEA)
Failure mode: Tenfold dose error (decimal point misread)
Effect: Patient serious harm or death — S = 10
Cause: Handwritten order with ambiguous decimal; no trailing zero (e.g., “.5 mg” read as “5 mg”) — O = 3
Control: Single-nurse check before administration — D = 6
RPN = 180 — High. Actions: (1) mandate printed labels from pharmacy system, (2) enforce independent double-check policy for high-alert medications, (3) add leading zeros to all doses (“0.5 mg”). After actions: D improves from 6 to 2 → new RPN = 60.
Software — Payment Processing API (SFMEA)
Failure mode: Checkout transaction timeout (no response within 10s)
Effect: Customer sees error, may retry — risk of double charge; lost sale — S = 7
Cause: External payment gateway latency spike (no retry logic) — O = 4
Control: Error log only; no automated alert — D = 6
RPN = 168 — High. Actions: (1) add exponential-backoff retry with idempotency key to prevent double charge, (2) add P99 latency alert in monitoring dashboard. After: O drops to 2, D to 2 → new RPN = 28.
FMEA vs Other RCA and Quality Tools
FMEA is proactive — it prevents failures. Other tools are primarily reactive — they investigate failures that have already occurred. Understanding the difference helps you pick the right tool at the right time.
| Tool | Timing | Purpose | Best used for |
|---|---|---|---|
| FMEA | Proactive | Identify and score all potential failures before they occur | New product/process launch; design changes; regulatory submissions |
| 5 Whys | Reactive | Drill to root cause of a single known failure | Recurring defects; incident investigations; quick RCA sessions |
| Fishbone diagram | Reactive / proactive | Brainstorm all possible causes across categories | Team-based cause brainstorming; complex multi-department problems |
| Pareto chart | Reactive | Prioritize which defects or causes to fix first (80/20 rule) | Quality data analysis; prioritizing improvement projects |
| PDCA cycle | Framework | Structure iterative improvement cycles | Wraps around all other tools; drives continuous improvement |
How they work together: In an automotive APQP project, the team runs a DFMEA during design to eliminate high-risk failure modes before tooling is cut. Once the process is in production, they run a PFMEA and use Pareto charts on early production data to find which failure modes are actually occurring most. When a new defect escapes to the customer, they run a 5 Whys to find the root cause, then update the PFMEA with the new cause and improved controls.
6 Common FMEA Mistakes
- Running FMEA solo or with a single department. FMEA is a team exercise. A design engineer alone will miss process failure modes. A process engineer alone will miss design intent. The cross-functional team is not optional — it is the method.
- Rating Severity, Occurrence, or Detection by gut feel without data. Occurrence is especially prone to optimistic underestimation. Use historical warranty data, field failure rates, or similar project data wherever available. If you have no data, say so and use conservative estimates.
- Ignoring high-Severity items because the RPN looks acceptable. An item with S=9, O=1, D=1 gives RPN=9 — but the effect is a safety failure. Always review S ≥ 9 items with a separate “critical characteristics” lens, regardless of RPN.
- Not updating the FMEA after design changes or process changes. An FMEA that does not reflect the current design or process is worse than useless — it gives false confidence. FMEA is a living document; update it whenever the scope changes.
- Treating FMEA as a compliance checkbox rather than a decision tool. If the team fills in the FMEA after the design is frozen, they are writing a post-hoc justification. FMEA must be run early enough that findings can influence decisions.
- Assigning actions without owners or due dates. An FMEA with high-RPN items and no assigned corrective actions is paperwork. Every recommended action needs a name, a target date, and a follow-up review to confirm the RPN improved after implementation.
Calculate RPN and Build Your FMEA Worksheet
Our free FMEA Calculator is coming soon — meanwhile, use our suite of root cause analysis tools to complete your risk management workflow.
Explore All Free Tools →Frequently Asked Questions
What does FMEA stand for?
FMEA stands for Failure Mode and Effects Analysis. It is a structured, proactive risk assessment method that identifies every way a product or process could fail, evaluates the consequences of each failure, and prioritizes action to reduce risk. Originally developed by the U.S. military in 1949, it was later adopted by NASA, Ford, and the automotive industry, and is now embedded in ISO/IATF and FDA standards.
What is a good RPN score in FMEA?
RPN ranges from 1 to 1,000. There is no universal threshold — different industries set their own action limits, commonly RPN ≥ 100–200. More importantly: any item with Severity ≥ 9 should receive immediate attention regardless of RPN, because a catastrophic safety or compliance failure justifies action even if the probability is very low. Low RPN on a high-Severity item is not permission to ignore it.
What is the difference between DFMEA and PFMEA?
DFMEA (Design FMEA) evaluates potential failures in the product design before manufacturing — it asks “could this design fail to meet its requirements?” PFMEA (Process FMEA) evaluates potential failures in manufacturing or assembly processes — it asks “could this process step produce a defect?” Both use the same RPN methodology. DFMEA is owned by the engineering team; PFMEA is owned by manufacturing and process engineering.
When should you use FMEA?
Use FMEA proactively: (1) when designing a new product or process, (2) when modifying an existing design or process, (3) before launching a new production line, (4) after a significant customer complaint or field failure, or (5) as part of a regulatory submission (FDA, automotive APQP). FMEA works best before failures happen — the later in the product lifecycle you run it, the more expensive any changes become.
How is FMEA different from the 5 Whys?
The key difference is timing and direction. FMEA is proactive — it asks “what could go wrong?” before a failure occurs and scores all failure modes by risk. The 5 Whys is reactive — it asks “why did this happen?” after a specific failure has already occurred. They complement each other: FMEA prevents failures; 5 Whys resolves failures that FMEA did not predict or prevent. After a 5 Whys investigation, the findings should always be fed back into the FMEA to update Occurrence and Detection ratings.
Is FMEA required by ISO 9001 or IATF 16949?
ISO 9001:2015 does not explicitly require FMEA, but its risk-based thinking requirement (clause 6.1) is commonly satisfied through FMEA. IATF 16949 (automotive quality standard) does explicitly require FMEA as part of Advanced Product Quality Planning (APQP). Medical device manufacturers under ISO 13485 and FDA 21 CFR Part 820 also use FMEA for design risk management under ISO 14971. Food industry HACCP plans are essentially a form of FMEA applied to food safety hazards.
Can FMEA be used outside of manufacturing?
Yes — FMEA is widely used across industries. In healthcare: surgical procedures, medication administration, and medical device design. In software engineering: service reliability analysis and API failure mode mapping. In financial services: transaction processing and compliance controls. In aerospace: system safety analysis. In food safety: HACCP is a direct application of FMEA principles. The RPN framework applies anywhere you can define a function, a potential failure mode, and a consequence.
Related Resources
- Root Cause Analysis: A Complete Guide (2026) — the full RCA process from problem statement to corrective action
- FMEA Examples: 6 Real-World Cases Across Industries — full worksheets from manufacturing, healthcare, software, and aerospace
- Free FMEA Template — Excel, Google Sheets & Online — blank worksheet with all 12 columns and step-by-step Excel setup
- Fault Tree Analysis: The Complete Guide (2026) — top-down Boolean logic, AND/OR gates, quantitative probability
- RCA Tools Compared: 5 Whys, Fishbone, FTA, FMEA & Pareto — when to use each method
- Pareto Analysis: The Complete Guide — prioritize defects with the 80/20 rule
- PDCA Cycle: The Complete Guide — continuous improvement framework that wraps around FMEA
- Free 5 Whys Tool — investigate failures that your FMEA did not prevent