The audit closes on a Thursday afternoon. The team is relieved: no Level 1, a handful of Level 2 findings, a few observations. Two weeks later the formal notification arrives, and the quality lead starts drafting responses.
Root cause: individual oversight. Corrective action: staff reminded of the procedure; procedure clarified. Status: closed.
Twelve months later, a different auditor samples a different project and raises the same finding.
That pattern is common enough that many offices treat it as normal. It is not. A repeat finding tells the authority that the organisation can respond to non-compliance but cannot learn from it — and that is a far more uncomfortable conversation than the original finding ever was.
This article explains what is actually being asked when a finding is raised, why the usual closures fail, and how to run findings as controlled work that changes the system instead of the paperwork.
What a Finding Actually Asks For
EASA's Part 21 framework distinguishes three kinds of output from oversight:
- Level 1 findings — non-compliances that may lead to uncontrolled non-compliance and a potential unsafe condition. The corrective action period is short: no more than 21 working days.
- Level 2 findings — any other non-compliance with the regulation, the organisation's own procedures and manuals, or its terms of approval. The initial corrective action period is up to three months, which the authority may extend once a corrective action plan has been agreed.
- Observations — not non-compliances, but things the authority wants the organisation to consider. The organisation is expected to give them due consideration and keep a record of what it decided.
For findings, the organisation's obligations are specific: identify the root cause and contributing factors, establish a corrective action plan, and demonstrate its implementation within the agreed period. The plan is expected to cover correction of the issue, corrective and preventive action, and the planning to implement them — and to be accepted before it is carried out.
Failing to submit an acceptable plan, or to implement it in time, is how a Level 2 finding becomes a Level 1 — with consequences that can reach the approval itself.
Read that list again and notice what it does not say. It does not ask the office to fix the sampled document. Fixing the sampled document is the easy part. The obligation is to find out why the system allowed it, and to show that the system no longer does.
Why Closures Fail
Most failed closures share the same shape: they fix the instance and describe the system.
"Human Error" as a Root Cause
If the root cause is that someone forgot, the corrective action is always a reminder. But the authority already knows people forget. The real question is why the process depended on memory at that point — why the form allowed an approval without a classification, why the folder accepted an unsigned revision, why the authority matrix was not checked when the approval was given.
Retraining as the Default Action
Training is appropriate when people did not know the rule. It is the wrong answer when they knew it and the workflow made skipping it easy. Offices that answer every finding with training build a long training log and a short list of actual changes.
Correcting Only the Sample
The auditor sampled one document. The fix is applied to that one document. Nobody checks whether the same gap exists in the other twelve projects running the same procedure — so the next sample finds it.
Procedure Edits Nobody Follows
The procedure is reworded to be clearer. The form, the tool, and the habit stay the same. Implementation evidence is a new revision of a document that describes a process the office still does not run — the gap that turns into findings against the office's own procedures.
No Effectiveness Check
The plan is implemented, evidence is sent, and the finding is closed. Nobody returns three months later to check whether the change actually stopped the problem.
Findings Live in a Spreadsheet
Findings, observations, and internal audit results are tracked separately, often by one person. Recurrence across audits and projects is invisible because nothing links a new finding to the old ones.
Closure Types Compared
| Closure type | What it fixes | What it misses |
|---|---|---|
| Correct the sampled record | The one artefact the auditor saw | Every other instance of the same gap |
| Remind or retrain staff | Awareness, briefly | Why the workflow allowed the error |
| Reword the procedure | The description of the process | The process itself |
| Change the workflow or control | The mechanism that allowed the gap | Nothing — if the change is verified |
| Verify effectiveness by sampling | Confidence the fix holds | Only what was not sampled |
Durable closures almost always sit in the bottom two rows. The top three are often necessary, but they are corrections, not corrective action.
A Finding Workflow That Changes the System
Treat each finding as a controlled item with a lifecycle, owners, and evidence — not as a letter to answer.
1. Intake
Record the finding as received: reference, level, requirement or procedure cited, sampled artefacts, the date of written notification, and the resulting deadline. Observations are recorded the same way, with a decision field instead of a corrective action.
2. Containment
If the finding points at something that could affect a product or an active package, deal with that first — correct the affected records, hold the affected approval, notify where required. Containment is about the effect, not the cause.
3. Extent of Condition
Before writing the plan, check how far the problem reaches. Sample the same procedure across other projects, other teams, and the last few months of records. The extent check tells you whether this is one slip or a system behaviour — and it is the single step most closures skip.
4. Root Cause and Contributing Factors
Ask why until the answer is about the system, not a person: the approval form does not require a classification; the authority matrix is not consulted at approval time; partner files enter the package without intake. Record contributing factors too — workload, unclear ownership, tool limitations — because they often explain why the gap appeared now.
5. Corrective Action Plan
Write actions that address the root cause, with owners, dates, and the evidence each action will produce. Separate three things clearly: the correction (fix the instances found), the corrective action (remove the cause), and the preventive action (extend the fix to similar processes that share the weakness).
6. Acceptance and Implementation Evidence
Record when the plan was submitted, when it was accepted, and what evidence demonstrates implementation. Evidence should show the changed control operating — a record produced by the new workflow — not just a revised procedure document.
7. Effectiveness Check
Schedule a check, typically a sample of new records two to three months after implementation. If the sample is clean, close the finding on the record. If not, reopen it before the authority does.
Make Findings Data, Not Correspondence
The workflow above is hard to run on email and a spreadsheet, for the same reason email approvals are hard to audit: the decisions are scattered and the links are implicit.
When findings are structured records, three things become possible:
- Recurrence is visible. A new finding can be linked to earlier findings, observations, and internal audit results that cited the same procedure or the same cause.
- Findings point at the system. Each finding links to the procedure, the controlled documents, the approvals, and the roles involved — so the corrective action can be targeted at the right control, such as the authority matrix or a missing approval gate.
- Internal monitoring feeds the same register. Findings raised by the office's own compliance monitoring follow the same lifecycle as authority findings, so the office finds its own gaps first — the posture the audit readiness checklist is built around.
The First 30 Days After an Audit
- Register every finding and observation within a week of notification, with level, deadline, cited requirement, and sampled artefacts.
- Contain product-affecting issues immediately. Do not wait for the root cause to correct records that matter now.
- Run an extent-of-condition sample for each finding before drafting a plan.
- Hold one root cause session per finding with the people who run the process, not only the quality office. Stop when the cause is a control, not a person.
- Draft plans that change a control — a required field, a gate, a role check, an intake step — and say what evidence will show it working.
- Book the effectiveness checks in the calendar the same day the plan is accepted.
- Record a decision for every observation, including "considered, no action, because…".
Signals Your Findings Process Is Not Working
- The same finding text appears in more than one audit report
- "Human error" or "oversight" is the most common recorded root cause
- Most corrective actions are training records or procedure rewording
- Corrective action evidence is a document revision, never a record produced by the changed process
- Nobody can list open findings, their deadlines, and their owners without asking the quality lead
- Observations are acknowledged in correspondence but have no recorded decision
- Effectiveness checks are not scheduled, or are skipped when the finding is already closed
If two or more of these are true, the office is closing findings without reducing the risk behind them — and the next audit is likely to find the same gaps in a different place.
Final Takeaway
A finding is not a request to fix a document. It is a request to show that the organisation understands why its system allowed a non-compliance, and that the system has changed.
Closures built on reminders, training, and reworded procedures answer the letter but leave the mechanism intact. Closures built on extent checks, system-level root causes, control changes, and scheduled effectiveness checks are slower to write, but they are the only kind that stop the same finding coming back.
The practical test is simple: for any finding closed in the last year, can you show — without archaeology — how far the problem reached, the root cause in terms of a control, the change made to that control, evidence of the changed control operating, and a clean effectiveness sample afterwards? If yes, the office is learning from oversight. If not, the next audit already has its first finding.