Failure Mode and Effects Analysis (FMEA): How to Identify and Prevent Failures Before They Happen

Why prevention is worth more than detection

The cost of a failure increases dramatically the later it is discovered in the product lifecycle. A design error caught during concept review costs far less to correct than the same error discovered during prototype testing, which costs far less than an error found during production validation, which costs far less than a warranty return or a field recall. FMEA’s core value is that it systematically identifies potential failure modes at the earliest possible stage, before tooling is cut, before production begins, and before products reach customers. Even when FMEA does not prevent every failure, it creates a documented record of which risks were considered and what controls were in place, which is important for liability, regulatory compliance, and continuous improvement.

The distinction between FMEA and traditional quality control inspection is important. Inspection detects failures that have already occurred. FMEA prevents failures by identifying their root causes and removing them from the design or process before production begins. A process that relies primarily on end-of-line inspection to catch defects is reacting to failures; a process that incorporates FMEA into its design and launch phases is designed to prevent them.

The two main types: DFMEA and PFMEA

FMEA is applied at two distinct stages of product development:

Design FMEA (DFMEA) analyzes the product design to identify failure modes caused by design decisions: incorrect material selection, insufficient structural strength, inadequate tolerances, sensitivity to temperature or vibration, or failure of safety-critical functions. DFMEA is conducted by the design engineering team before the design is released for tooling or manufacturing. The analysis focuses on what can go wrong with the product’s function, and what design changes can eliminate or reduce the risk. The outputs of DFMEA directly feed into the product’s verification and validation test plan: high-severity design failure modes should be specifically tested to confirm they have been addressed.

Process FMEA (PFMEA) analyzes the manufacturing or assembly process to identify failure modes caused by process steps: incorrect machine settings, operator errors, tool wear, inadequate fixtures, contamination, or out-of-sequence operations. PFMEA is conducted by the manufacturing engineering team before production launch. The analysis focuses on what process steps can introduce defects into an otherwise acceptable design. The outputs of PFMEA directly feed into the process Control Plan, which specifies which process parameters must be monitored, at what frequency, and with what reaction plan when a deviation is detected.

DFMEA and PFMEA are sequential: DFMEA is completed first (as it defines the product), and PFMEA follows (as it defines the process for making the product). The two analyses are linked: a design characteristic identified as critical in DFMEA (because a failure would have high severity) must be reflected as a control item in the PFMEA and the Control Plan.

The FMEA worksheet: severity, occurrence, and detection

The standard FMEA worksheet captures the following for each failure mode:

Function: What the product component or process step is supposed to do. A function is stated as a verb and a noun: “transmit torque,” “seal against fluid ingress,” “position the workpiece within 0.05 mm.”

Failure Mode: How the component or process step could fail to perform its function. A single function may have multiple failure modes: “transmit torque” could fail as “no torque transmitted,” “insufficient torque transmitted,” or “torque transmitted intermittently.” Each failure mode is analyzed separately.

Effect of Failure: What the customer (internal or external) experiences when this failure mode occurs. Effects are described from the customer’s perspective: “vehicle inoperative,” “assembly cannot be completed,” “dimensional nonconformance detected at inspection.” Multiple effects may result from a single failure mode.

Severity (S): A score from 1 to 10 rating the seriousness of the failure’s effect on the customer. Score 9 to 10 is reserved for failures that affect safety or regulatory compliance without warning. Score 7 to 8 represents high product dysfunctionality. Score 5 to 6 represents moderate loss of performance. Score 1 to 2 represents a barely noticeable effect. Severity scores are assigned based on the worst potential effect and cannot be reduced by adding detection controls; only a design change that removes the root cause can reduce a severity score.

Cause of Failure: The specific mechanism or root cause that would allow this failure mode to occur. Causes must be stated in terms that can lead to corrective action: “insufficient weld penetration due to incorrect wire feed rate” rather than “bad weld.” A single failure mode may have multiple potential causes, and each cause is analyzed separately with its own Occurrence and Detection scores.

Occurrence (O): A score from 1 to 10 estimating the frequency with which this cause will produce this failure mode, given the current design or process. Score 10 indicates near-certain occurrence; score 1 indicates the cause is extremely unlikely. Occurrence scores are calibrated to historical defect rates, warranty data, or engineering judgment about the robustness of the current design relative to the identified cause.

Current Controls: The prevention controls (design features or process controls that reduce the likelihood of the cause occurring) and detection controls (inspection, testing, or monitoring activities that would detect the failure mode before the product reaches the next customer) currently in place.

Detection (D): A score from 1 to 10 rating how likely the current controls are to detect the failure mode or its cause before the product or assembly leaves the current manufacturing step or service location. Score 1 means the detection is almost certain (the failure will definitely be caught). Score 10 means there is no current control that can detect the failure. Counterintuitively, a high detection score is bad (meaning the failure is hard to detect), and a low detection score is good (meaning it is easy to detect).

Risk Priority Number (RPN): The product of the three scores: RPN = S x O x D. Maximum RPN is 1000 (10 x 10 x 10). Minimum RPN is 1 (1 x 1 x 1). High-RPN items are prioritized for corrective action. However, RPN alone is not sufficient for prioritization: a failure mode with severity 9 and low occurrence and detection scores may have a lower RPN than a less dangerous failure mode with moderate scores across all three dimensions, yet the high-severity item should always be addressed first because of its safety or regulatory implications.

The new AIAG-VDA FMEA methodology (FMEA-MSR)

In 2019, AIAG (Automotive Industry Action Group) and VDA (German Association of the Automotive Industry) jointly published a revised FMEA handbook that replaced the traditional RPN with a more nuanced approach. The new methodology introduces separate Action Priority (AP) ratings (High, Medium, Low) based on a three-dimensional combination of S, O, and D scores, rather than a single RPN product. It also adds a third FMEA type: FMEA-MSR (Monitoring and System Response), which analyzes the effectiveness of diagnostic systems in vehicles and automated equipment in detecting faults during operation and triggering appropriate responses. The revised methodology requires a seven-step process: planning and preparation, structure analysis, function analysis, failure analysis, risk analysis, optimization, and results documentation. Organizations working to IATF 16949 requirements are expected to align their FMEA practice with the AIAG-VDA handbook.

Linking FMEA to the Control Plan

The PFMEA and the Control Plan are companion documents: the PFMEA identifies which process parameters are critical (because their variation can cause high-severity failures), and the Control Plan specifies how those parameters are controlled in production. A characteristic flagged as a Special Characteristic in PFMEA (typically indicated with a specific symbol mandated by the customer) must appear in the Control Plan with a specified measurement method, sample frequency, control method (SPC chart, attribute inspection, or error-proofing), and reaction plan. If the PFMEA and Control Plan are not synchronized, critical characteristics may go unmonitored in production, which defeats the purpose of both documents.

Common errors in FMEA

• Stating effects at the wrong level of abstraction. Describing the failure effect as “part does not meet specification” instead of “vehicle inoperative” or “assembly cannot proceed” makes the severity score meaningless. Effects must be described from the customer’s perspective — what the customer experiences — not from the engineer’s measurement perspective.
• Confusing the cause with the failure mode. “Insufficient material strength” is a cause. “Bracket fractures under load” is the failure mode. Confusing the two produces analysis that cannot lead to root-cause corrective action, because the cause statement does not identify what can be changed to prevent the failure.
• Reducing severity scores by adding detection controls. Severity is a property of the failure’s effect on the customer; it does not change based on how well the failure is detected. Adding a 100-percent inspection step does not make a safety-critical failure any less dangerous if it escapes detection. Severity can only be reduced by changing the design so that the failure either cannot occur or has a less severe consequence.
• Completing FMEA after production has started. An FMEA completed after the process is already in production is primarily a documentation exercise. It can still capture knowledge and identify improvement opportunities, but it cannot influence the design decisions that have already been made. Maximum value is obtained when FMEA is completed during the design and process planning phases, while there is still time to change things.

Frequently asked questions

What is FMEA and what is it used for?

Failure Mode and Effects Analysis (FMEA) is a structured, proactive risk analysis method used to identify every way a product, process, or system can fail, assess the severity of each failure’s effects, estimate the probability of each failure occurring, and evaluate the ability of current controls to detect failures before they reach the customer. It is used during product design (Design FMEA) and manufacturing process planning (Process FMEA) to identify high-risk failure modes and drive corrective actions before production begins. FMEA is required by automotive quality standards (IATF 16949), aerospace standards (AS9100), and medical device standards (ISO 13485), and is widely used in any industry where failure prevention is preferable to failure detection. Its primary output is a prioritized list of failure modes with recommended corrective actions, linked to the product’s verification and validation plan (for DFMEA) or the process Control Plan (for PFMEA).

How is the Risk Priority Number (RPN) calculated and what are its limitations?

The Risk Priority Number (RPN) is calculated by multiplying three scores, each on a scale of 1 to 10: Severity (S, the seriousness of the failure’s effect on the customer), Occurrence (O, the frequency with which the failure cause will produce the failure mode), and Detection (D, the difficulty of detecting the failure before it reaches the customer). RPN = S x O x D. Maximum RPN is 1000; minimum is 1. The main limitation of RPN is that it treats the three scores as equally weighted, which can produce misleading prioritization. A failure mode with Severity 9, Occurrence 1, Detection 1 has an RPN of 9, while a failure mode with Severity 3, Occurrence 3, Detection 3 has an RPN of 27 — but the first failure mode is far more dangerous. For this reason, the AIAG-VDA FMEA handbook (2019) replaced RPN with Action Priority (AP) ratings, which weight high-severity scores more heavily. Regardless of RPN, any failure mode with a severity score of 9 or 10 should be addressed regardless of the occurrence and detection scores.

What is the difference between Design FMEA and Process FMEA?

Design FMEA (DFMEA) analyzes the product design to identify failure modes that result from design decisions: incorrect material selection, insufficient structural strength, inadequate tolerances, or failure of safety-critical functions under extreme conditions. DFMEA is conducted by the design engineering team before the design is released for manufacturing, and its outputs feed into the product verification and validation test plan. Process FMEA (PFMEA) analyzes the manufacturing or assembly process to identify failure modes introduced by process steps: incorrect machine settings, operator errors, tool wear, inadequate fixtures, or out-of-sequence operations. PFMEA is conducted by manufacturing engineering before production launch, and its outputs feed directly into the process Control Plan. The two types are sequential and linked: DFMEA identifies critical product characteristics, which PFMEA then specifies how to control in the manufacturing process. Both documents must be synchronized so that characteristics flagged as critical in DFMEA appear as controlled characteristics in PFMEA and the Control Plan.

How should severity, occurrence, and detection scores be assigned?

Severity (S) scores rate the effect of the failure on the customer, where 10 is a safety or regulatory non-compliance issue without warning and 1 is a barely noticeable effect. Severity is assigned based on the worst possible consequence of the failure and cannot be changed by adding detection controls — only a design change that removes the root cause can lower a severity score. Occurrence (O) scores rate the frequency with which the identified cause will produce the failure mode, where 10 means near-certain occurrence and 1 means extremely unlikely. Occurrence scores are best calibrated to historical defect rates, warranty data, or reliability data for similar designs or processes. Detection (D) scores rate how likely the current controls are to detect the failure before it reaches the customer, where 1 means detection is almost certain and 10 means there is no detection control. A common mistake is assigning a low (favorable) detection score to 100 percent inspection: even 100 percent manual inspection typically has a detection efficiency below 90 percent and should not receive a score of 1 or 2 unless the detection method is automated and verified. The AIAG-VDA handbook provides detailed scoring tables for each rating that should be used as the basis for team discussion.

What is the AIAG-VDA FMEA methodology and how does it differ from the traditional approach?

The AIAG-VDA FMEA handbook, published jointly by the Automotive Industry Action Group and the German Association of the Automotive Industry in 2019, replaced the traditional FMEA approach used in the AIAG 4th edition handbook. The key differences are: the new methodology replaces the Risk Priority Number (RPN) with an Action Priority (AP) system that gives more weight to high-severity scores and reduces the ability of low detection scores to offset high severity scores; it introduces a formal seven-step process (planning and preparation, structure analysis, function analysis, failure analysis, risk analysis, optimization, results documentation) that requires explicit linking between the product or process structure, the intended functions, the failure modes, and the effects; it adds FMEA-MSR (Monitoring and System Response) as a third FMEA type for analyzing diagnostic systems in vehicles and automated equipment; and it requires documentation of the basis for each score assignment, making the analysis more auditable. Organizations certified to IATF 16949 are expected to align their FMEA practice with the AIAG-VDA handbook, though the specific implementation timeline is agreed between the customer and supplier.

How does FMEA connect to the Control Plan in manufacturing?

The Process FMEA and the Control Plan are companion documents in the Advanced Product Quality Planning (APQP) process. The PFMEA identifies which process parameters and product characteristics are critical because their variation can cause high-severity failures; these are designated as Special Characteristics (typically indicated with a diamond, star, or other symbol mandated by the customer). Every Special Characteristic identified in the PFMEA must appear in the Control Plan with a specified measurement method (what is measured and how), a sample plan (sample size and frequency), the type of control (SPC chart, attribute inspection, poka-yoke, or parameter monitoring), the specification or control limit, and the reaction plan that operators must follow when a deviation is detected. If the PFMEA and Control Plan are not synchronized — for example, if a characteristic is added to the PFMEA in a later revision but not carried through to the Control Plan — the characteristic will go unmonitored in production, defeating the purpose of the analysis. Customer-specific requirements (from automotive OEMs like Ford, GM, or Stellantis) often specify additional requirements for the number of Special Characteristics, the required control method, and the required documentation format.