Articles · Level 1 · Observe · No. 1
Why the same nonconformity keeps coming back
Recurrence after a closed and verified corrective action is normally read as evidence that the root-cause analysis was shallow. Usually that reading is correct. This article is about the residue it leaves behind — the recurrences where the analysis was sound, the action was implemented, the action worked, and the problem came back anyway.
The observation §
A nonconformity is raised. It is investigated properly — five whys, or a fishbone, or a fault tree, whatever the organization uses and whatever the customer expects to see. A cause is identified. A corrective action is defined, assigned, implemented and recorded. Some months later, effectiveness is verified: the problem has not recurred. The record is closed.
Eighteen months after that, the same failure mode appears again. Not a similar one — the same one. Same characteristic, same operation, same mechanism, often the same part number.
Somebody re-opens the file. The corrective action is still in place. The procedure that was revised is still at the revision it was raised to. The training was delivered and the records exist. The added inspection step is still on the traveller and is still being signed. Nothing was withdrawn, nothing was skipped, nothing expired in the document-control sense. Every element of the response is exactly where it was left.
And the problem is back.
The standard explanation, in its strongest form §
The received answer is that the root cause was never found. What was found was a cause — a proximate one, a convenient one, one that suggested an action the organization was already willing to take. The why chain stopped early, usually at the point where the next question would have implicated a decision rather than a person or a machine. Treat a symptom and the symptom returns.
This explanation deserves to be taken seriously, because most of the time it is right. Analyses do stop early. Corrective actions are written to be closeable rather than to be effective. Retraining is used as a universal remedy because it is cheap, fast, auditable and requires nobody to change anything they own. Anyone who has read a hundred CAPA records has read most of them as this.
The literature, the training courses and the audit checklists all converge on the same instruction: push the chain further, ask why one more time, look for the systemic cause behind the local one. It is good advice and it fixes a large fraction of recurrence.
It does not fix all of it, and the part it does not fix has a specific shape.
Where the standard explanation does not fit §
Some recurrences carry four features together. Individually each is unremarkable. Together they are not compatible with shallow analysis.
The action demonstrably worked. Not appeared to work in the sense that nothing was reported. The defect rate fell and stayed down; the escape count went to zero and stayed there for a period long enough that chance is not a comfortable explanation. Whatever the analysis found, it found something real, because acting on it changed the process.
The gap is long. A recurrence three weeks after closure is consistent with an action that never worked. A recurrence after two clean years is not. The interval itself is evidence that the response was, for a time, adequate.
Nothing was undone. This is the feature that ought to be strange and rarely is. If the recurrence were caused by the response decaying — the step quietly dropped, the check quietly skipped, the revised procedure quietly ignored — that decay would be visible. Procedure compliance is the one thing a quality system measures continuously and well. In these cases the compliance evidence is clean, and it is clean because the compliance is real.
Re-analysis produces the same answer. This is the decisive one. If the first analysis had been too shallow, then a second analysis of the recurrence — done by different people, with more time, under the embarrassment of a repeat — should reach deeper and find something different. Very often it does not. It finds the same cause, proposes the same action, and observes that the action is already in place.
An investigation that arrives at the same destination twice is not describing a defective investigation. It is describing an answer that was correct and is no longer sufficient.
What the residue implies §
A corrective action is not a fact about the process. It is a solution to a problem as the problem stood, in the conditions that held on the day it was written. The solution was sized, scoped and justified against those conditions. Then the solution was written down and preserved with great care, and the conditions were not written down at all.
The examples are ordinary:
- Training as the action. It worked because the people who were trained stayed. It assumed a turnover rate. Nobody recorded the assumption, and nobody re-tests it when the rate changes.
- Hundred-per-cent inspection as containment, later made permanent. It worked at eleven parts a day. It assumed a volume. At sixty parts a day the same instruction produces a different result, and the instruction has not changed.
- A fixture or a mistake-proofing device. It worked for the part family it was built around. It assumed a product mix. The mix moved; the device is still installed, still verified, still catching the errors it was designed to catch, and no longer covering most of what runs through the cell.
- An added review step. It worked because the reviewer had the time to do it properly and the experience to know what to look for. It assumed a workload and a person. Both were replaced without a decision that felt like a change.
- A supplier notification and a tightened receiving check. It worked against the failure that supplier was producing then. It assumed a supplier process. The supplier has since changed foundry, alloy source, or subcontractor, correctly and with notification to a mailbox that is not connected to the corrective action.
In each case the action is still being executed exactly as specified. What has stopped being true is something the action depended on and never mentioned. And because it was never mentioned, there is nothing to review, nothing to re-approve, nothing to fall out of date, and nothing to fail an audit.
The recurrence, on this reading, is not the return of the original problem. It is the first visible consequence of a condition that ceased to hold at some earlier, unrecorded moment. The nonconformity is the alarm — a slow, expensive, downstream alarm that fires long after the thing it is reporting actually happened.
Which is why re-running the analysis produces the same answer. The analysis is not looking for the change. It is looking for a cause, and the cause it finds is the same cause it found before, because the cause has not changed. What changed is the world the answer was sized for.
What this is not §
This reading is only useful if it is narrow. Four things that look like it and are not:
- Shallow root-cause analysis. The ordinary case, and by volume the dominant one. It is identified by a second analysis that reaches further than the first. If re-analysis finds something deeper, the diagnosis is shallow analysis and the remedy is better analysis. Nothing in this article applies.
- An action that was never implemented. A revision issued but not deployed, training scheduled but not delivered, a step written into the procedure and never onto the traveller. This is nonconformance to the organization's own corrective-action process, it is detectable by an ordinary audit, and it is a different failure with a different remedy.
- Erosion of practice. The step that gradually stopped being done, the check signed without being performed, the workaround that became the method. Real, common, and the opposite of the case described here — erosion shows up as a gap between record and practice, and these cases have no gap.
- A different problem that resembles the old one. Same symptom, different mechanism. The test is whether the failure mode is genuinely identical, not whether the complaint text reads the same.
And one thing this article is not claiming: that most recurrence works this way. It does not. Most recurrence is one of the four above. The claim is narrower — that a residue exists, that it is not rare, and that it is currently explained by a category that does not fit it.
What would settle it §
The distinction is decidable from records that already exist. Four questions, in order, on a recurrence you already have:
- Did the action measurably work? Not was it closed. Is there data showing the rate moved and stayed moved for a defined period?
- Is it still being executed as written? Records, not assurances — signed travellers, training matrices, inspection results, the device still in place and still verified.
- Does re-analysis reach further? Have someone who did not do the first analysis do the second one, and compare destinations rather than methods.
- What changed in between? Volume, staffing, product mix, supplier, equipment, schedule, shift pattern, customer requirements. Anything the action's adequacy depended on, whether or not the action mentions it.
If 1 is yes, 2 is yes, 3 arrives at the same place and 4 identifies a change, then treating the recurrence as an analysis failure will produce a third analysis, a third action, and a third recurrence. If 3 reaches further, the ordinary explanation is the right one and it should be used.
The fourth question is the one that no standard corrective-action form asks, and the only one whose answer would have been free to record at the time.
A name for it §
What this article has described is a corrective action that was correct, that worked, that is still being executed exactly as written, and that no longer prevents what it was built to prevent. That state has a name.
Silent Expiry is the state of a control that remains documented, in force and verified as compliant while at least one condition its protective effect depends on has ceased to be true — and whose failure produces no signal. A control in this state continues to behave exactly as specified. It no longer protects anything.
The name is not doing any explanatory work above. Nothing in the four questions depends on it, and naming the state adds nothing to what the article has already shown: it is what the condition of the control is called once those four questions have resolved the way this article describes. The term is a label for an observation, not an account of why the observation occurred, and it is offered here only because an observation is easier to carry when it has a name.
The full test — the three conditions that must hold together, and the cases they exclude — is in the canonical definition, Section 1. It costs nothing to read and nothing to cite.
Why this matters §
A quality system is built on the assumption that a control, once verified, keeps protecting until something visibly breaks it. That assumption is sound for controls that fail loudly, and it is the reason long-lived controls are rarely revisited. For a control that fails by staying correct, the assumption holds only as long as the world it was sized for does. Accepting that changes the question worth asking about every long-lived control — not is this being followed, which is already known, but is the reason it exists still true, which nothing asks.
Where this goes §
The next six articles reach the same state from six other observations: the chosen number nobody can justify, the sentence “we have always done it that way” read literally, the audit that passed three weeks before the failure, controls followed perfectly that protect nothing, the limits of effectiveness verification, and the difference between a system that works and a system that used to. Six independent routes to one destination is what establishes that these are one thing and not seven — and each of them ends the way this one does, with the observation first and the name after it.
That is the whole of Level 1 · Observe. The mechanism underneath it, stated precisely enough to be argued with, is Level 2 · Understand, and the diagnostic you can run on a control of your own arrives at Level 4 · Diagnose. The reading sequence shows where each level sits.
Permanent URL §
https://silent-expiry.pages.dev/articles/01/
Cite as: Articles, No. 1 — “Why the same nonconformity keeps coming back.” Silent Expiry, silent-expiry.pages.dev, 2026. This article is derived from the canonical definition, Rev3, and does not modify it. Where the two differ, the definition governs.