Process failure mode and effects analysis is the structured method a manufacturer uses to decide where to spend control effort. It lists every step of the build, for each step the ways it can go wrong, the effect of each failure on the customer, the likelihood and detectability of each failure, and the controls that will keep it from reaching the next step. It is a working document that determines where inspection stations are placed, which parameters are monitored, which steps are gated, and which problems are considered acceptable residual risk.
That last point is why it belongs in a procurement conversation rather than in a filing cabinet. A PFMEA is a formal, written record of what the supplier has decided it cannot or will not control. Buyers routinely receive assurance that "we follow PFMEA" without ever seeing what risk was accepted on their product, and the failures that reach a customer in volume are overwhelmingly the ones that sat on a risk register without an effective action attached.
What a PFMEA Actually Contains
The core of the document is a table. Each line describes one failure mode at one process step, scored on three dimensions, with the resulting risk number and the resulting action. Understanding the columns is what lets a buyer read one in fifteen minutes rather than accept it as a formality.
| Column | What It Records | Why a Buyer Cares |
|---|---|---|
| Process step | The specific operation — paste printing, placement, reflow, cleaning, depanelisation | Reveals the process granularity: does the analysis treat "assembly" as one step or twenty |
| Failure mode | The specific way the step fails — insufficient paste volume, tombstoning, voiding, thermal damage | Named failure modes mean the team knows its own defect history; generic entries mean the document was copied |
| Effect | What the customer sees — intermittent open, latent short, field failure, cosmetic reject | Connects internal process to your product's reliability, not just to a visual accept/reject call |
| Severity (S) | 1-10, worst-case consequence | Driven by your product class; a medical or automotive build should carry higher severity than consumer goods |
| Occurrence (O) | 1-10, how often the mode arises | Should be supported by internal defect data, not intuition |
| Detection (D) | 1-10, how likely the control catches it before shipment | The column most easily flattered: claiming good detection without a control that performs it |
| RPN | Severity x Occurrence x Detection | The prioritisation number — and the number most often managed downward instead of the risk |
| Action and owner | What will be changed, by whom, by when | An action with no owner and no date is not an action |
The column set is standard across the industry and derivable from any quality manual, which is exactly why the presence of a PFMEA proves very little. The information is all in the entries. A credible document names specific failure modes that match the defects the process actually produces, cites real occurrence data, and pairs each high-risk line with a control that physically exists on the floor.
The RPN Trap — and Why Buyers Should Look Past It
The risk priority number is a useful sort key and a well-known weakness. Because it is a product of three subjective scores, it can be reduced by adjusting any one of them without changing anything on the shop floor. Lowering an occurrence score from 6 to 3 because a process was "improved" lowers the RPN by half while leaving the process untouched. Lowering a detection score has the same effect. The number moves; the risk does not.
Two practices make the figure actively misleading. The first is treating a single RPN threshold as the trigger for action, so lines that score below the threshold receive no attention regardless of how severe their effect is. A failure mode with severity 10, occurrence 2 and detection 5 scores 100, identical to one with severity 2, occurrence 5 and detection 10 — and the two deserve completely different responses. A latent open on a safety-related automotive board and a cosmetic blemish on a consumer device are not equivalent risks, and no multiplication makes them equivalent.
The second is scoring detection optimistically. Detection should reflect the probability that the control in place catches the failure, not the probability that it would be caught if someone were looking. A PFMEA that credits automatic optical inspection with a low detection score because an AOI station exists, without accounting for the escape rate the station actually demonstrates, is describing an intention rather than a control. The measurement of what inspection actually catches is covered in our guide to PCB testing methods, and the diagnosis of yield losses that escape is set out in first-pass yield diagnosis.
Key Takeaway: Read a PFMEA by severity first, RPN second. Sort the table by severity descending and look at what is being done about the highest-consequence failure modes, regardless of their RPN. A document that manages one composite number rather than a severity-ordered risk list is optimised for the audit rather than for the product.
How the Analysis Should Differ for Your Product
A generic PFMEA applied unchanged to every customer is a warning sign, because the correct analysis depends on the product. Severity is assigned by consequence to the end user, and consequence differs by application. The same manufacturing defect carries different severity in a consumer accessory, an industrial controller and an automotive safety module.
The table below shows how a single failure mode — an insufficient solder joint on a fine-pitch component — should shift across applications. The process control response changes with it.
| Application | Typical Severity Rationale | Control Expected |
|---|---|---|
| Consumer electronics | Function loss in a non-safety device, replaceable at modest cost | AOI on fine-pitch positions; sample X-ray |
| Industrial control | Downtime in a production line, possible consequential damage | AOI plus defined X-ray sampling on critical nets; functional test |
| Medical device | Patient-safety relevant; recall exposure | Full X-ray on critical components; documented process capability; traceability to lot |
| Automotive | Safety-relevant, warranty and regulatory exposure | 100% X-ray on safety-critical joints; capability studies; PPAP-grade evidence |
When the control expectations are matched against a supplier's actual process, the PFMEA becomes a document with commercial consequences rather than a compliance checkbox. For instances where the analysis must feed into a formal submission, the evidence package is described in our guide to PCB supplier PPAP. Where the requirement class itself drives severity assignment, the framework is set out in IPC Class 2 versus Class 3.
Questions That Reveal Whether the Document Is Alive
A PFMEA maintained as a working tool is updated when a defect occurs, when a process changes, or when a new product is introduced. One that was written for a certification audit and never touched again shows it immediately under questioning. The following questions are answerable in minutes by a team that uses the document.
What is the highest-severity failure mode on my product, and what control addresses it?
A live document has a named answer. A dormant one produces a general statement about quality systems. The purpose of asking is not to challenge the answer but to see whether the analysis was ever specialised to the product in front of them.
When was it last updated, and what triggered the revision?
The correct answer names an event — a defect trend, a process change, a new material, a customer complaint. A revision history showing only annual reviews with no content changes indicates the document is not connected to production data.
Where do the occurrence scores come from?
Credible occurrence scoring is supported by internal defect data: defect rates per process step, PPM, or a defect Pareto. Answers that cite experience or general industry figures indicate the scores were estimated rather than measured.
Which detection controls have a demonstrated escape rate?
Detection credit should correspond to what a control actually catches, verifiable through inspection escape data or validation records. A control with no measured effectiveness is an assumption recorded in a column.
Which actions are open, and who owns them?
An open-action list with named owners and dates shows the analysis drives work. A table in which every line is closed indicates either a fully controlled process, which is rare, or a document that was formally signed off and set aside.
How does the output connect to the control plan?
The control plan is the operational counterpart of the PFMEA: it states which parameters are monitored, at what frequency, and what the response is when a limit is exceeded. The two should agree line for line. Where they diverge, the control plan — which governs what operators actually do — is the operative document.
Where FMEA Fits in the Quality Chain
The PFMEA is one component of a connected set. Design FMEA examines failure modes arising from the design itself and belongs with the development function. The process FMEA examines the build. The control plan operationalises the process analysis. Measurement system analysis establishes that the data feeding occurrence scores is trustworthy. And when a failure does occur, the disciplined response runs through a structured problem-solving report rather than an informal adjustment.
For a buyer, the practical value of understanding this chain is that it shows where the leverage is. Reviewing a PFMEA before award is far cheaper than diagnosing a field failure after ramp, because the analysis is the one quality document that formally states what the supplier intends to control and what it intends to accept. Measurement reliability, which determines whether the occurrence scores mean anything, is covered in our guide to Gauge R&R and measurement system analysis, and the post-production response structure is set out in our guide to 8D reports and corrective action.
The Bottom Line
A PFMEA is a decision record, not a certificate. It states, in writing and in advance, which failure modes the supplier considers material and which controls it will maintain against them. The value to a buyer is not in its existence but in its content: whether the failure modes named match the product, whether severity drives the priority order rather than a single composite score, whether occurrence and detection scores rest on measured data, and whether the actions are closed by engineering change or by editing a number.
At Huaxing PCBA, process FMEA is maintained per product family and updated against internal defect data rather than reviewed on a calendar. Severity is assigned by application class, high-severity modes are controlled with named inspection and process gates, and the control plan is kept aligned with the analysis line for line. Read our supplier audit guide or contact our engineering team to review the risk analysis behind your product before you commit a build.
