OperationalPlaybook Guide

How to Do an FMEA, Step by Step

An FMEA is one of the highest-leverage hours your team will spend all quarter, and most teams run it wrong. They book a two-hour meeting, fill a spreadsheet with guesses, calculate a pile of RPN numbers nobody acts on, and file it. Audit season comes, somebody dusts it off, and it has not prevented a single failure.

That is not an FMEA. That is a spreadsheet wearing a lab coat.

Here is how to actually do one. Failure Mode and Effects Analysis is a structured way to find where a process or product can fail before it fails in the field, rank those failures by risk, and fix the worst ones first. Done right, it is the cheapest insurance you will ever buy. This guide walks the full method, scoring scales included, so your team can run one start to finish.

What You Need Before You Start

Do not open the template yet. Get these four things in the room first.

A defined scope. One process or one product. “Our whole assembly line” is too big. “The torque station on Line 3” is a session you can actually finish.
The right people. Whoever runs the process, whoever inspects it, and whoever hears the customer complaints. Three perspectives minimum. An FMEA built by one engineer alone is just that engineer’s blind spots, formatted.
The process steps written down. A process FMEA marches step by step. List the steps first, in order, before you start hunting failures.
Your scoring scales agreed up front. Severity, Occurrence, and Detection each ride a 1 to 10 scale. Decide what the numbers mean before the meeting, or every score becomes a debate.

The Core FMEA Steps

Here is the loop. Run it for every process step, one row at a time.

Name the failure mode. For this step, how could it go wrong? The bolt is under-torqued. The label prints blank. The weld has porosity. Be specific. “Defective part” is not a failure mode, it is a shrug.
State the effect. If that failure reaches the customer, what happens? Loosens in the field, fails inspection, safety recall. The effect drives your Severity score.
Score Severity (1 to 10). How bad is the effect? A 1 is a cosmetic nuisance nobody notices. A 9 or 10 is a safety hazard or a regulatory failure. Score the worst realistic outcome, not the average day.
List the causes. Why would this failure happen? Worn fixture, wrong setting, operator skips a step. Each cause can get its own row, because each cause has its own likelihood.
Score Occurrence (1 to 10). How often does this cause actually happen? A 1 is “basically never seen it.” A 10 is “every shift.” Use real data if you have it. Scrap reports beat memory.
Score Detection (1 to 10). If the failure happens, how likely are you to catch it before it ships? Careful: this scale is backwards from the others. A 1 means you will almost certainly catch it. A 10 means it sails straight to the customer. Strong detection earns a low number.
Calculate the RPN. Risk Priority Number = Severity x Occurrence x Detection. Range is 1 to 1000. This is your triage list, not your scoreboard.
Sort by risk and act on the top of the list. This is the step everyone skips and the only step that matters. The RPN is worthless until it drives an action.

How to Read the Scores Without Fooling Yourself

The RPN is a sorting tool, not a verdict. Three rules keep it honest.

First, watch Severity on its own. A failure with Severity 9 or 10 deserves action even if Occurrence and Detection are low and the RPN looks tame. A rare, hard-to-catch safety failure is exactly the thing FMEA exists to surface. Do not let a low RPN talk you out of fixing something that can hurt someone.

Second, attack the right lever. A high RPN driven by Detection means your inspection is weak, so add a check or a poka-yoke. A high RPN driven by Occurrence means the failure happens too often, so fix the process, not the inspection. The three numbers tell you what kind of fix to reach for. Read them, do not just multiply them.

Third, re-score after you act. The whole point is the second number. You add a torque verification, re-run the row, and Detection drops from 8 to 2. The RPN falls. That delta is the proof your action worked. An FMEA with no re-scored rows is an FMEA nobody finished.

Close the Loop or Skip the Meeting

Take your sorted list and set a threshold. Everything above an RPN you choose, or anything with Severity 9 or higher, gets an owner and a due date. That is the entire output of the session: a short list of the highest-risk failures with a name and a date next to each.

Then you go fix them and come back and re-score. An FMEA is not a document, it is a habit. The spreadsheet is just where the habit leaves its tracks.

A few hard rules:

Score the worst realistic case, not the convenient one.
Detection is reverse-scored. Low number is good. Get this wrong and your whole sheet lies to you.
Every high RPN row leaves the meeting with an owner and a date, or the meeting was theater.
Re-score after every action. No re-score, no proof.

That is how you do an FMEA. Scope it tight, score it honestly, sort by risk, fix the top of the list, and re-score to prove it. The first one your team runs will feel slow. By the third, you will catch failures in a planning meeting that used to cost you a customer.

Run Your First FMEA This Week

We built the FMEA Template Package so you do not have to format any of this from scratch. It is a ready-made Excel workbook with the Severity, Occurrence, and Detection scales already written in, automatic RPN calculation, pre-populated sample rows so you can see a finished example, and a dashboard that sorts your risks for you. White-label placeholders throughout, so it drops straight into your shop. It will not run the analysis for you and it promises no audit outcome, but it removes every bit of setup friction between you and a finished FMEA. Find it in the OperationalPlaybook Etsy shop.