FMEA Example for Manufacturing: How to Run One Without a Black Belt
A plain-language FMEA walkthrough for small manufacturers. Pick one process, rate severity, occurrence, and detection on simple anchors, find your RPN, and fix the top risks. No Six Sigma required.
The first time someone handed me an FMEA template, I closed it within about ninety seconds. It had forty columns, three different rating scales pulled from a reference manual I did not own, and a tab labeled “Severity Classification Cross-Reference.” I was a shift supervisor at a shop with eleven people on the floor. I did not have a black belt. I did not have a quality department. I had a part that kept coming back from our biggest customer, and I had a Tuesday.
That gap, between what FMEA actually is and how it gets dressed up, is the reason most small shops never run one. The acronym (Failure Mode and Effects Analysis) sounds like something that requires a certification and a week-long class. The page-one search results do not help. They are mostly consultant sites and standards bodies quoting AIAG handbooks, and they are written for an audience that already speaks the language. If you run a 40-person plant, that material reads like it was built to keep you out.
So let me strip the ceremony off. Here is the whole idea in one sentence.
FMEA Is Just Four Questions Written Down
For every step in a process, you ask: what can go wrong here, how bad is it if it does, how often does it happen, and would we catch it before it reaches the customer? That is it. That is the entire method. Everything else (the RPN math, the rating scales, the action columns) is just a structured way to write those answers down so you can compare them and decide where to spend your limited time.
The reason it works is not the form. It is the act of forcing a team to slow down and walk a process step by step, out loud, asking “and then what breaks here” at every station. Half the value shows up in that conversation before you have scored a single number. I have watched an operator stop a meeting cold with “well, it only fails when the morning crew loads the wrong fixture,” and that one sentence was worth more than the whole document.
You do not need Six Sigma for this. You need a process, a few people who actually run it, and a couple of hours.
Pick One Process, Not the Whole Plant
The biggest mistake I see is scope. A plant manager decides to “do FMEA,” and the team tries to map the entire facility. Three meetings in, everyone is exhausted, the spreadsheet has 200 rows, and nothing has been fixed. The project dies and FMEA gets filed under “things big companies do.”
Do the opposite. Pick one process. The one that keeps you up at night. The part that gets returned, the operation with the scrap problem, the line that the customer complained about last quarter. A process FMEA (sometimes written PFMEA) looks at a single process and its steps, not your whole operation.
Narrow it until it fits on one page. “How we machine and inspect the bracket on cell 3” is a process. “Quality” is not.
Define Your Own Rating Anchors in Plain Words
Here is where the standard guides lose people. They hand you a 10-point scale with definitions like “Severity 7: High, product operable but at reduced level of performance, customer dissatisfied.” That phrasing is fine for an automotive tier-one supplier. For your shop, write your own anchors in words your team uses on the floor.
You rate three things, each from 1 to 10:
- Severity (S): how bad is it if this failure reaches the customer?
- Occurrence (O): how often does this failure actually happen?
- Detection (D): if it happens, how likely are we to catch it before it ships? (This one is backwards: 10 means we would almost never catch it, 1 means we catch it every time.)
Then anchor each scale in plain language. Here is a starter set I have handed to shops. Edit the wording to match your reality.
| Score | Severity (how bad) | Occurrence (how often) | Detection (would we catch it) |
|---|---|---|---|
| 1-2 | Customer never notices | Almost never (years apart) | We catch it every time, automatically |
| 3-4 | Minor annoyance, no function loss | Rare (a few times a year) | A check exists and usually catches it |
| 5-6 | Customer complains, rework needed | Occasional (monthly) | We might catch it on a manual check |
| 7-8 | Function impaired, possible return | Frequent (weekly) | Easy to miss, relies on one person noticing |
| 9-10 | Safety issue or line-down at customer | Constant (most shifts) | No real way to catch it before it ships |
The exact numbers matter less than the team agreeing on what they mean. Spend ten minutes up front getting consensus on the anchors, and every score after that goes faster and means something.
Multiply to Get RPN
Once you have your three numbers for a given failure mode, you multiply them:
RPN = Severity x Occurrence x Detection
RPN stands for Risk Priority Number. It runs from 1 (1x1x1, basically harmless) to 1000 (10x10x10, a frequent severe failure you would never catch). It is not a precise measurement. It is a sorting tool. Its only job is to float the worst stuff to the top of your list so you know what to attack first.
Let me show you what a real row looks like.
A Filled-In Row or Two on a Real Step
Say the process is machining that bracket on cell 3, and one step is “operator deburrs the mounting hole by hand.”
Failure mode: burr left in the mounting hole.
- Severity: the customer’s automated assembly arm jams on a burr and stops their line. That is a 9. A line-down call from your biggest account is about as bad as it gets short of a safety recall.
- Occurrence: it happens maybe twice a month. Call it a 5.
- Detection: today, the only check is the operator eyeballing it. Burrs in a recessed hole are easy to miss. That is a 7.
RPN = 9 x 5 x 7 = 315.
Now a second row on the same process, a different step: “operator stamps the lot number on the part.”
Failure mode: lot number stamped illegibly.
- Severity: customer cannot read it, calls to ask, mild annoyance, no function loss. Call it a 3.
- Occurrence: rare, a few times a year. Call it a 2.
- Detection: it is visible, the next operator usually notices. Call it a 4.
RPN = 3 x 2 x 4 = 24.
Look at those two side by side. 315 versus 24. The math just told you something your gut already knew but the team might have argued about: the burr is the fire, the stamp is not. You attack the burr.
Attack the Top Few, Not All of Them
This is the discipline that separates an FMEA that changes something from one that just sits in a binder. Once you have scored every failure mode and you have a column of RPNs, you sort that column high to low and you draw a line under the top three to five. Those are the only ones you work on right now.
Do not try to fix everything. A shop with twenty failure modes that tries to action all twenty actions zero of them. Pick the worst few. For the burr at RPN 315, the fix might be a simple go/no-go pin gauge the operator pushes through every hole, which is a fifteen-dollar tool. You are not trying to make the part perfect. You are trying to knock the top risk down.
A note on chasing the number: a high RPN driven by severity is different from one driven by occurrence. You usually cannot lower severity (a burr jamming a customer’s line is always bad), so you go after occurrence (why does the burr happen) or detection (how do we catch it before it ships). When detection is the weak link, that is where your five-dollar gauges and your 5 Why analysis earn their keep, because finding the real cause is often what lets you stop the failure at the source instead of just inspecting harder.
Re-Rate After the Fix
The FMEA is not done when you make the change. Put in the go/no-go pin gauge for the burr, run it for a few weeks, then come back and re-score that row. Severity is still 9 (the burr would still jam their line if it got through). But occurrence might drop to a 2 once the gauge catches the root cause earlier, and detection drops to a 2 because the gauge catches it every time.
New RPN = 9 x 2 x 2 = 36.
You went from 315 to 36 with one cheap tool, and now you have it written down. That before-and-after is the proof you show the customer, and it is the thing that turns the next return into a corrective action instead of a crisis. If a failure does slip through after your fix, that is when your right-sized CAPA process takes over and feeds the lesson back into the FMEA.
Where This Fits
An FMEA done this way is a living document, not a one-time event. You pull it back out when a process changes, when a new failure shows up, or when you onboard a new part. It becomes the memory of your process: every way it has bitten you, ranked, with the fix and the proof attached.
This is the third piece I have walked through in my Quality Systems series, alongside root cause work and corrective action. They are the same idea wearing different hats: look at what actually happens on your floor, write it down in plain language, and put your limited hours against the thing that hurts most. None of it requires a belt of any color.
If it helps to start from something already built, I put together a one-page process FMEA worksheet with the rating anchors above already filled in, the same one I have handed to shops on a Tuesday afternoon. It is a starting point you can mark up and make your own, and you can find it with my other Quality Systems tools.
Liked this? Get the free SOP Writing Cheat Sheet.
The 6-step SOP format on one page. Free.