Practical guide · Quality management
What is FMEA? Meaning, DFMEA vs PFMEA, RPN and action priority
Short answer
FMEA stands for failure mode and effects analysis. It is a team method: list how a design or a process step could fail, rate each failure for severity, occurrence and detection, and fix the biggest risks before parts are made.
FMEA (failure mode and effects analysis) is a structured way to find how a product design or a manufacturing process could fail, what each failure would do, what would cause it, and how well today's controls would prevent or catch it. Each failure is rated 1 to 10 for severity, occurrence and detection. The team then acts on the highest risks first, before a part is made or shipped. A design FMEA (DFMEA) looks at the product; a process FMEA (PFMEA) looks at the steps that make it. The ranking used to be the risk priority number (RPN, severity times occurrence times detection); the AIAG-VDA FMEA Handbook of 2019 replaced it with an action priority of high, medium or low.
Last reviewed:
What FMEA means
FMEA is short for failure mode and effects analysis. A failure mode is a way something fails to do its job: a weld that does not hold, a bolt that is missing, a seal that leaks. The effect is what that failure does to the customer, the next process step or the operator. The analysis is the team's work of listing the failure modes, their effects and causes, rating each one, and deciding what to do about the worst.
The point is timing. Root cause tools such as the 5 Whys or a fishbone diagram are used after a failure has happened. An FMEA is done before: before a design is released, before a process starts running, or before a change is made. It asks what could go wrong, not what did.
The method came out of the US military in the late 1940s (procedure MIL-P-1629), was taken up by aerospace in the 1960s, and became standard in automotive through the AIAG FMEA manuals. Today the reference most plants and their customers use is the AIAG-VDA FMEA Handbook, published in 2019, which set the seven-step method and the action priority described below.
| Term | Meaning | Example from a bolted bracket |
|---|---|---|
| Function | What the item or step must do, with its requirement | Clamp the bracket with four bolts at 25 Nm |
| Failure mode | A way the function is not met | One bolt missing |
| Effect | What the failure does downstream | Bracket detaches in service |
| Cause | Why the failure mode happens | Operator interrupted, no bolt count |
| Prevention control | What stops the cause today | None, bolts come from a bulk bin |
| Detection control | What would catch the failure before it leaves | Visual check at end of line |
| S, O, D | Severity of the effect, occurrence of the cause, detection by the controls, each 1 to 10 | 9, 3, 6 |
| RPN | Risk priority number: S times O times D | 9 x 3 x 6 = 162 |
| AP | Action priority (high, medium, low) from the AIAG-VDA tables | High |
The same columns, ready to fill in, are in the free [PFMEA template](/templates/pfmea-template) and [DFMEA template](/templates/dfmea-template).
Types of FMEA
The two you will meet in almost every plant are the design FMEA and the process FMEA. The others are variations for systems, for monitoring during use, and for reliability work.
| Type | What it analyses | When | Who leads |
|---|---|---|---|
| Design FMEA (DFMEA) | How the product design could fail to do its job, assuming it is made to the drawing | During design, before release | Design, test and reliability engineers |
| Process FMEA (PFMEA) | How each process step could fail to make the part to the drawing | During process development, before production, and after every process change | Process and quality engineers with operators and maintenance |
| System FMEA | How subsystems and their interfaces could fail at the level of the whole product | Early in design, before the DFMEAs of the parts | Systems engineering |
| FMEA-MSR | How the product monitors itself and responds to a failure while the customer is using it (added by the AIAG-VDA handbook) | Alongside the DFMEA for products with sensors and software | Design and functional safety engineers |
| FMECA | An FMEA with a criticality analysis added, used in aerospace, defence and reliability engineering | With the DFMEA | Reliability engineers |
A reverse FMEA is not another type: it is a check on the shop floor, after launch, that the controls written in the PFMEA are really in place and that the failure modes it lists cannot happen.
DFMEA vs PFMEA
The difference between DFMEA and PFMEA is what could fail. A DFMEA asks whether the design would work even if every part were made exactly to the drawing. A PFMEA assumes the design is right and asks how the process could fail to make the part to the drawing. The same leak at a hose joint can appear in both: a groove too shallow for the O-ring is a design cause; an operator fitting the wrong O-ring is a process cause.
| DFMEA (design FMEA) | PFMEA (process FMEA) | |
|---|---|---|
| Object | The product: system, subsystem, component | The process: each step that makes, assembles, tests or moves the part |
| Question | Could the design fail to meet its function? | Could this step fail to make the part to the drawing? |
| Done when | During design, while changes are cheap | Before production, and again at every process change or complaint |
| Typical failure modes | Does not hold the load, leaks, wears early, interferes with another part | Dimension out of tolerance, part missing, wrong part, wrong orientation, not done, contaminated |
| Typical causes | Wrong material, tolerance stack, too little clearance, wrong assumption about use | Setup error, worn tool, operator distracted, wrong material delivered, unclear instruction |
| Detection means | Design reviews, analysis, prototype and validation tests before release | In-process checks, gauges, sensors, poka-yoke, inspection before the next step |
| Team | Design, test, reliability, manufacturing, quality, someone who knows the field | Process engineer, quality engineer, operator, maintenance, supervisor |
| Output | Design changes, verification tests, special characteristics passed to the PFMEA | Poka-yoke devices, process changes, checks that go into the control plan and work instructions |
The handover is the list of special characteristics: the features the DFMEA found critical, which the PFMEA must then protect. In an automotive PPAP both FMEAs are submitted together with the control plan.
Which one you need depends on what you own. A plant that builds to a customer's drawing does the PFMEA; the customer or the design owner does the DFMEA. A company that designs and makes its own product does both, DFMEA first.
Severity, occurrence and detection
Every failure chain (cause, failure mode, effect) is rated three times on a 1 to 10 scale. The rating tables come from the AIAG-VDA handbook or from your customer; the ones in our templates are general guides. What matters most is that the team reads the table criteria aloud and rates against them, not against each other's gut feel.
| Rating | What it rates | 1 means | 10 means | What lowers it |
|---|---|---|---|---|
| Severity (S) | How bad the effect is for the customer, the next step or the operator | No noticeable effect | Affects safe operation or breaks a regulation, possibly without warning | Only a design or process change that changes the effect itself |
| Occurrence (O) | How often the cause is expected to happen, given today's prevention controls | Very unlikely, the cause is prevented | Very frequent | A prevention control: a poka-yoke, a fixture, a capable process, a better setup |
| Detection (D) | How likely today's detection controls are to catch the failure or its cause before it moves on | Almost certain to be caught | No control, or the control cannot catch it | A detection control: a gauge, a sensor, a test, an inspection closer to the step |
Detection is the one people get backwards. A low number is good: 1 means the failure would almost certainly be caught.
Two rules follow. Better inspection never lowers severity: a leak that reaches the customer is just as bad however many you catch. And prevention beats detection: a control that stops the cause lowers occurrence and removes the defect, while a check only finds it after it has been made.
RPN vs action priority (AIAG-VDA 2019)
For decades the ranking was the risk priority number: RPN = severity x occurrence x detection, a figure from 1 to 1,000, highest first. It is simple and still widely used, and many customers still ask for it. Its weakness is that it treats the three ratings as equal. A severity 10 failure with occurrence 2 and detection 2 scores 40 and sinks to the bottom of the list, below a cosmetic scratch rated 3 x 8 x 5 = 120.
The AIAG-VDA FMEA Handbook (2019) replaced RPN with action priority (AP). Instead of multiplying, the team looks up the combination of S, O and D in the handbook's tables and reads off high, medium or low. Severity carries the most weight, then occurrence, then detection. A severity 9 or 10 failure with any realistic chance of occurring comes out high whatever the detection rating.
| RPN | Action priority (AP) | |
|---|---|---|
| How it is worked out | S x O x D | Look up S, O and D in the handbook's tables |
| Result | A number from 1 to 1,000 | High, medium or low |
| Weighting | All three ratings equal | Severity first, then occurrence, then detection |
| What the result asks for | Whatever threshold the plant sets, which hides high severity | High: act, or document why not. Medium: action should be taken. Low: action may be taken |
| Where it is required | AIAG 4th edition FMEAs, many non-automotive plants, older customer forms | Customers working to the AIAG-VDA handbook, mainly automotive |
Both our templates carry an RPN and an AP column. Use the one your customer asks for, and in either case deal with every severity 9 and 10 first.
If you use RPN, do not set a pass mark such as 'act above 100'. Sort by severity first, then by RPN within each severity. That gets you most of what action priority does without a new form.
How to do an FMEA: the seven steps
The AIAG-VDA handbook lays out seven steps. They work just as well for a plant that uses RPN. The first three are preparation, the middle three are the analysis, and the last one keeps it alive.
The seven FMEA steps
- 1
Plan and prepare
- 2
Structure analysis
- 3
Function analysis
- 4
Failure analysis
- 5
Risk analysis
- 6
Optimisation
- 7
Document and keep it alive
- 1
Plan and prepare
Set the scope (which product or process, from where to where), pick the team, book the time, and gather the inputs: drawings, the process flow, past complaints and non-conformance reports, and FMEAs of similar parts.
- 2
Structure analysis
Break the object down. For a PFMEA, list every process step in order, including handling, inspection and rework. For a DFMEA, list the system, its subsystems and components.
- 3
Function analysis
Write what each element must do, with the requirement: torque to 25 Nm, seal to 2 bar, hold 200 N. A failure mode is any way that function is not met, so a vague function gives a vague FMEA.
- 4
Failure analysis
For each function, list the failure modes, then the effect of each and its causes. Ask what else could go wrong before moving to the next step; most steps have more than one failure mode.
- 5
Risk analysis
Record today's prevention and detection controls, then rate severity, occurrence and detection against the rating tables. Work out the RPN or read off the action priority.
- 6
Optimisation
For every high priority (and every severity 9 or 10), agree an action with an owner and a date. Prefer actions that remove the cause over actions that add a check. Once the action is in, rate S, O and D again.
- 7
Document and keep it alive
File the FMEA with its before and after ratings, carry the controls into the control plan and work instructions, and reopen it at every process change, customer complaint and 8D.
How to run the session itself, from the 5T preparation to scoring calibration and getting committed actions before people leave the room, is in our FMEA workshop guide.
A worked PFMEA example
This example is illustrative. The process step is operation 40 on a small assembly line: tighten the four M8 bolts that hold a motor bracket to a frame, to 25 Nm, with a click-type torque wrench. Three failure modes were found for the step.
| Failure mode | Effect | S | Cause | Prevention today | O | Detection today | D | RPN |
|---|---|---|---|---|---|---|---|---|
| One bolt missing | Bracket can detach in service | 9 | Operator interrupted mid-cycle; bolts taken from a bulk bin, no count | None | 3 | Visual check at the end-of-line audit | 6 | 162 |
| Bolt under-torqued | Bracket loosens, vibration, motor misaligns, customer complaint | 7 | Wrench drifts out of calibration; operator stops before the click | Wrench calibrated quarterly | 4 | Torque audit on one unit per shift | 7 | 196 |
| Wrench slips, scratches the paint | Cosmetic, visible to the customer | 3 | No guide sleeve on the socket | None | 6 | Visual check at final inspection | 3 | 54 |
RPN check: 9 x 3 = 27, 27 x 6 = 162. 7 x 4 = 28, 28 x 7 = 196. 3 x 6 = 18, 18 x 3 = 54.
Sorted by RPN, the under-torqued bolt comes first at 196 and the missing bolt second at 162. Sorted the AIAG-VDA way, severity leads: the missing bolt is a severity 9 with a realistic chance of occurring, so it is the high priority whatever its RPN. The scratch is low either way, and the team can leave it.
One action covers both of the first two lines: replace the click wrench with a controlled torque tool that logs each tightening and will not release the unit to the next station until four cycles have reached 25 Nm. The bolts also move from a bulk bin to a kitted tray of four per unit, so a missing bolt is visible on the tray before the unit leaves.
| Failure mode | S | O | D | RPN | What changed |
|---|---|---|---|---|---|
| One bolt missing | 9 | 2 | 2 | 36 | Kitted tray prevents it; the tool's cycle count catches it at the station |
| Bolt under-torqued | 7 | 2 | 2 | 28 | The tool stops the cause and checks every bolt, not one unit a shift |
| Wrench slips, scratches the paint | 3 | 6 | 3 | 54 | No action taken, low priority; reviewed at the next revision |
Check: 9 x 2 = 18, 18 x 2 = 36. 7 x 2 = 14, 14 x 2 = 28. Severity has not moved on any line: a missing bolt would still be a severity 9 if it got through, so the line stays in the FMEA.
RPN before and after the actions
Example numbers- One bolt missing: RPN 162 before, 36 after
- Bolt under-torqued: RPN 196 before, 28 after
- Wrench slips, scratches the paint: RPN 54 before, 54 after
Risk priority numbers before and after actions: one bolt missing 162 to 36, bolt under-torqued 196 to 28, wrench slips and scratches the paint 54 to 54 with no action.
The last step is the one most teams skip. The torque tool's pass signal and the kitted tray go into the control plan for operation 40 and into the work instruction, so the control outlives the people who sat in the workshop. The PFMEA template works the RPNs out for you and flags every severity 9 and 10.
What happens after the FMEA
An FMEA that is filed and forgotten was a documentation exercise. Three things should follow it.
- The control plan. Every detection and prevention control the FMEA relies on is written into the control plan for that step: what is checked, how, how often, and what the operator does when it fails. The free control plan template has the columns.
- The work instructions and poka-yoke. Actions that remove a cause usually end up as a fixture, a sensor or a changed sequence. Our poka-yoke guide and the poka-yoke worksheet are the next stop for the high-priority lines.
- Revision. The FMEA is reopened at every process change, every customer complaint and every 8D. D7 of an 8D report asks what in the system allowed the problem; the usual answer is a failure mode the FMEA did not list or a control it rated too kindly.
FMEA vs root cause analysis
FMEA and root cause analysis are two directions on the same road. Root cause analysis starts from a failure that has happened and works back to its cause, with tools such as the 5 Whys, the fishbone diagram and fault tree analysis. FMEA starts from a function and works forward to the failures that could happen.
They feed each other. Every closed root cause investigation is a failure chain the FMEA should contain, with the new control rated honestly. And the FMEA's list of causes is the first place to look when a failure does happen, because the team has already thought about it. The fuller treatment of the after-the-fact tools is in our root cause analysis guide.
Common FMEA mistakes
- Doing it after the fact, to satisfy a customer or an auditor, once the design is frozen or the line is running. The method only pays when changes are still cheap.
- One engineer at a desk. The failure modes operators and maintenance see every week never reach the sheet.
- Rating by feel instead of against the table. Scores then depend on who spoke last, and the FMEA cannot be compared with the one from the next line.
- Ranking by RPN alone and setting a pass mark. A severity 10 with a low RPN disappears below the line.
- Lowering severity because detection improved. Only a change to the design or the process lowers severity.
- Adding inspection as every action. Checks cost money every shift and still let the cause happen. Look for a prevention control first.
- Mixing design and process causes. 'Operator fits the wrong O-ring' belongs in the PFMEA; 'groove too shallow' belongs in the DFMEA.
- Never updating it. A process has changed; the FMEA has not. The next complaint will be a failure mode it does not list.
Go deeper
This page is the short version. These guides cover each part in full.
- FMEA in Manufacturing: Failure Mode and Effects Analysis Complete GuideThe long version: each S, O and D scale in detail, RPN vs action priority, the seven AIAG-VDA steps and where FMEA sits in the lean system.
- FMEA Workshop: Risk Assessment and Prioritization GuideHow to run the session: the 5T preparation, team roles, brainstorming failure modes, scoring calibration and action planning to closure.
- Designing Poka-Yoke Devices: Complete Implementation GuideHow to turn a high-priority FMEA line into a device that prevents the cause rather than a check that finds the defect.
- Fault Tree Analysis: Event-Based Root Cause Analysis for ManufacturingThe top-down companion to FMEA for a single serious event, and when to use each.
- Process Capability: Cp and Cpk Explained for ManufacturingWhere an honest occurrence rating comes from for dimensional failure modes: the capability of the step.
All our articles on this subject are in the quality management topic.
Free templates and tools
- PFMEA template (process FMEA)Failure modes, effects and causes rated for severity, occurrence and detection, with RPN before and after action.
- Design FMEA (DFMEA) templateA design FMEA worksheet with S, O, D and RPN before and after action, a class column and a design rating guide.
- Control plan templateWhat is checked at each step, against which limits, how, how often, and what to do when a result is out.
- Poka-yoke worksheetEach step's errors and what they cost, the device (approach, function, method), payback, and a log that proves each device works.
- Fault tree analysis templateA drawing area with the FTA symbols, an event table with probabilities and minimal cut sets, and the gates worked out bottom-up in Excel.
- 8D report templateD0 to D8 with targets, containment counts, occurrence, escape and systemic causes, verified actions and a worked example.
- Cost of poor quality (COPQ) calculatorTotal COPQ and cost of quality, and each as a share of revenue, from the four quality cost categories.
- Process capability (Cp, Cpk) calculatorCp and Cpk against your specification limits, plus Pp and Ppk from pasted measurements, with a distribution chart.
- First pass yield (FPY) calculatorFirst pass yield at each step and rolled throughput yield (RTY) across the whole process.
Doing this in LeanSuite
Quality Improvement
Log quality issues from the floor, trace each one to its root cause and follow the corrective action through to closure.
See Quality ImprovementAI Root Cause Analysis
Describe a problem, answer the AI's questions, and get the likely root causes and an action plan.
See AI Root Cause AnalysisNon-Conformance (NCR)
Moves non-conformance reports into an automated workflow and links each defect report to root-cause problem solving.
See Non-Conformance (NCR)
See it on your own lines
Bring the problem you are working on to a 45-minute demo. Or look around with sample data first, or pilot LeanSuite on your own lines for 90 days, with the setup done by our team.
FAQ
Questions about fMEA
FMEA stands for failure mode and effects analysis. A failure mode is a way a product or a process step fails to do its job; the effect is what that failure does to the customer, the next step or the operator. The analysis is the team's rating of each failure for severity, occurrence and detection, and the actions that follow.
FMEA is the method. DFMEA (design FMEA) applies it to a product design and asks whether the design would work if every part were made to the drawing. PFMEA (process FMEA) applies it to the manufacturing process and asks how each step could fail to make the part to the drawing. A company that designs and makes its own product does both, DFMEA first.
The two main types are design FMEA (DFMEA) and process FMEA (PFMEA). Others are system FMEA, for the whole product and its interfaces early in design; FMEA-MSR, for how a product monitors itself and responds to failures in use; and FMECA, an FMEA with a criticality analysis added, used in aerospace and reliability work.
RPN is the risk priority number: severity multiplied by occurrence multiplied by detection, each rated 1 to 10, giving a number from 1 to 1,000. A failure rated 9, 3 and 6 has an RPN of 162. Higher means higher risk. Its weakness is that it weights the three ratings equally, so a catastrophic but rare and well-detected failure can score low.
Action priority is the ranking the AIAG-VDA FMEA Handbook (2019) introduced in place of RPN. Instead of multiplying the ratings, the team looks up the combination of severity, occurrence and detection in the handbook's tables and reads off high, medium or low. Severity carries the most weight. High means act or document why not, medium means action should be taken, low means action may be taken.
There is no safe RPN threshold, and setting one is the most common FMEA mistake. An RPN of 40 can be a severity 10 failure that is rare and well detected, and it still needs action. Sort by severity first and deal with every 9 and 10, then work down the RPNs within each severity, or use action priority, which builds that rule in.
The AIAG-VDA handbook's seven steps are: planning and preparation, structure analysis, function analysis, failure analysis, risk analysis, optimisation, and results documentation. The first three set the scope and break the product or process into functions, the middle three find and rate the failures and agree actions, and the last records the results and keeps the FMEA current.
A process FMEA needs the people who designed, run, inspect and maintain the process: a process engineer, a quality engineer, an experienced operator or supervisor and a maintenance technician, plus a facilitator who runs the session rather than supplying the answers. A design FMEA needs design, test, manufacturing and quality engineers and someone who knows how the product is used in the field.
Before the design is released or the process starts running, while changes are still cheap, and again at every change: a new machine, material, supplier, tool or cycle time. It is also reopened after a customer complaint or a serious internal defect, to add the failure mode that was missed and re-rate the controls. An FMEA done only to satisfy an auditor after launch changes nothing.
A failure mode is a specific way a function is not met. For a step that tightens four bolts to 25 Nm, the failure modes are a bolt missing, a bolt under-torqued, a bolt over-torqued and the wrong bolt fitted. For a hose connector that must seal to 2 bar, the failure mode is a leak at the joint. A failure mode is not the cause (the wrench drifted) or the effect (the bracket fell off).
Root cause analysis works backwards from a failure that has happened to its cause, using tools such as the 5 Whys, a fishbone diagram or a fault tree. FMEA works forwards from a function to the failures that could happen, before they do. Each closed root cause investigation should be added to the FMEA as a failure chain with its new control.
Both use it. Six Sigma uses FMEA in the Analyze and Control phases of DMAIC to rank risks in a process. Lean uses it to decide where mistake-proofing pays off first and as the input to the control plan and standard work. Its home is quality engineering, and in automotive it is a required part of the PPAP submission.
Related guides and tools
- Reliability-centred maintenance: where FMEA-style analysis picks the maintenance task
- PFMEA template in Excel and PDF, with RPN and action priority columns and a rating guide
- DFMEA template in Excel and PDF, with a class column for special characteristics
- Control plan template: where the FMEA's controls are written down for every shift
- Poka-yoke: how to prevent the causes the FMEA ranks highest
- Root cause analysis: the after-the-fact tools, 5 Whys, fishbone and fault tree
- APQP: where the DFMEA and PFMEA sit in a launch, and how they feed the control plan and PPAP
More method guides
- The 6 Ms of production
- Poka-yoke
- Model area 90-day plans
- Moving from Excel to one system
- Reducing machine downtime: a worked month
- Andon
- Cycle time vs lead time, worked through
- Root cause analysis
- How many operators a line needs
- Reducing scrap and rework: a worked month
- Unsafe acts and unsafe conditions
- Just culture
- Stop-work authority
- Material review board (MRB)
- Nonconformance and NCRs
- Problem solving methods
- Quad chart
- OEE calculation
- Quality assurance vs quality control
- SOP meaning and format
- Job hazard analysis
- Near miss reporting
- Fishbone diagram
- Pareto chart
- Histogram
- Control chart
- What is a gemba walk
- What is lead time
- What is operational excellence
- How to improve product quality
- Preventive vs predictive maintenance
- Kanban
- PDSA cycle
- Safety moments
- Just-in-time manufacturing
- Kaizen
- 5S
- Total quality management (TQM)
- CAPA (corrective and preventive action)
- Layered process audit (LPA)
- Lean tools
- Lockout tagout (LOTO)
- A3 problem solving
- Preventive maintenance
- APQP
- High reliability organization (HRO)
- Manufacturing KPIs
- Quality metrics
- Practical problem solving
- CMMS
- Risk assessment
- Production efficiency
- Total productive maintenance (TPM)
- Reliability-centered maintenance (RCM)
Or browse the topics, the full library and the lean glossary.
Rather see it on your own problem?
Bring the problem you are stuck on, and we will show you on a call how LeanSuite handles it.