FMEA (Failure Mode and Effects Analysis)

FMEA is a prioritisation method in which a risk-assessment team scores every answer ("failure mode") to the question "how could this part or step break?" from three angles. Three questions are asked for every failure mode: how often does it occur (Occurrence, O), how severe is the outcome when it does (Severity, S), and how low is the chance of it being detected before it reaches the customer (Detection, D). All three are scored from 1 to 10 and multiplied; this product is called the Risk Priority Number (RPN). The output is an RPN for every failure mode and a rank it produces; the highest RPN shows the failure to address first. Stamatis's 1995 book compiled the classical method and turned it into an industry standard; it is a mandatory part of quality-management systems in automotive, aerospace and medical-device manufacturing.

FMEA's underlying thinking differs from other ranking methods: it is not about finding the best alternative but about catching the most dangerous one in advance. Think of a doctor performing triage. The doctor does not rank patients by "who is healthier"; the doctor prioritises by "whose condition is more urgent right now." FMEA likewise does not praise failure modes, it finds which one will do the most harm if it slips through unnoticed. Multiplying the three factors stops one very high factor from masking the others. A fair comparison between a frequent but minor failure and a rare but fatal one becomes possible this way. This is a risk philosophy within engineering discipline that converts expert judgement directly into a number; it does not use weights.

The method proceeds through three steps.

First, listing failure modes and scoring them. The team lists every possible failure mode in the product or process under review. Three scores are given for each: occurrence, severity and detectability. All are between 1 and 10; 10 shows the worst case (occurs very often, produces a very severe outcome, is very hard to detect).

Second, calculating the risk priority number. FMEA multiplies the three scores: occurrence times severity times detectability. The figure ranges from 1 to 1000; a large figure means high risk. No normalisation or weighting occurs at this step, because the three scores already sit on the same 1-10 scale.

Third, ranking and correcting. Failure modes are ranked by risk priority number from highest to lowest; the highest figure shows the risk to be tackled first. After the team takes corrective action, the three scores are reassessed and the figure recalculated; a fall shows the action worked.

The formulas behind each step, the intermediate tables and citation formats are given on the DecisionMind method page; this card carries no formulas.

The risk priority number tells you how urgently a failure mode should be addressed in terms of these three factors; it says nothing more. A figure of 192 does not mean "19.2 per cent risk" or "a 19.2 per cent chance of failing"; it is the product of three separate 1-10 scores. It cannot be compared with another assessment's RPN, because the scale depends on each team's own scoring habits. A high RPN does not mean "an accident is bound to happen"; it means "taking the three factors together, this failure demands the most attention."

A small RPN gap is not a decisive priority order but one sensitive to the scoring. Because the RPN is a product, a single one-unit change in any of the three scores makes it grow or shrink disproportionately. The report should show which score the ranking is most sensitive to.

Thus instead of writing:

"FMEA found the riskiest failure"

the report should read:

"With these O, S and D scores, this is the failure with the highest risk priority number; the priority may change if one of the scores changes"

Classical FMEA works with crisp data: three whole numbers, from 1 to 10, for every failure mode. As of this writing, DecisionMind has no registered FMEA extension alongside the base method. For situations where scores are given in linguistic terms ("frequent," "moderate," "rare") or disagree between experts, the literature has defined fuzzy and intuitionistic risk-analysis approaches, but these are not yet registered in DM3 as a separate FMEA family member.

You need: an occurrence, severity and detectability score for every failure mode, on the same 1-10 scale and given consistently by the same team. Classical FMEA assigns no weight to the criteria; the three factors carry equal say. A minimum of two failure modes is required; three to twelve failure modes are comfortably tracked.

If you want to see a product or process's possible failures in advance and know which to address first, FMEA is a suitable choice. Its typical fields are manufacturing reliability, process quality management and risk assessment.

The situation where it should not be used arises when the question asked is "which alternative is generally better" among a set of options. FMEA is not a selection method, it is a prioritisation method; it does not answer questions like "the best supplier." If scores are linguistic or uncertain and hard to reduce to a precise 1-10 number, fuzzy risk-assessment approaches should be considered instead of crisp FMEA.

Prioritising failure modes, three risk factors scored precisely → FMEA

Same goal, scores are linguistic or uncertain → fuzzy or intuitionistic risk-analysis approaches

A general ranking or choice is needed among alternatives → ranking methods such as TOPSIS, SAW

A difference in weight between criteria matters → a weighted ranking method; FMEA uses no weights

FMEA's greatest strength is its simplicity and disciplined structure. It builds a common language among team members; the concepts of "occurrence," "severity" and "detectability" can be discussed within the same frame even by people with different technical backgrounds. It can be calculated by hand, requires no software, its result reduces to a single figure and translates directly into an action list. Because it can be recalculated after corrective action, it shows progress in concrete terms. These features explain why the method sits as a mandatory step in quality-management standards.

Its limitations also come from its structure. First, treating the three factors as equally weighted is an assumption; in some applications severity is considered far more important than occurrence, but classical FMEA does not distinguish this (Liu, 2016). Second, the same RPN can arise from different score combinations: 2x5x8=80 and 4x4x5=80 give the same figure while concealing different risk profiles. Third, the 1-10 scale is ordinal in nature; multiplication requires a ratio scale, and this inconsistency has been criticised in the literature (Liu, 2016). Fourth, the scores rest on expert opinion and can vary from team to team.

The most common mistake is reading the RPN as a precise probability or percentage; the RPN is the product of three separate scales. A second mistake is comparing different teams' RPNs directly; scoring scales can shift from team to team. A third is ignoring a failure entirely because its RPN is low; a failure with very high severity but low occurrence can receive a low RPN and still needs to be tracked separately with its own threshold. A fourth is assuming risk has fallen after corrective action without reassessing the scores.

The governing principle is this:

An FMEA result is the product of the occurrence, severity and detectability scores the team supplied; if one of the scores is contested, the risk order is contested too, and highly severe failures must be tracked separately regardless of their RPN.