Nobody trusts a 98% CAPA closure rate—define “closed” as three verifiable conditions

The five 'Closed' ones that the auditor casually clicked on

The second-party audit is in the afternoon session, and the auditors do not count the closure rate; they only randomly select five CAPAs that are marked as closed. First question: What specific actions were taken? Second question: Who verified it and based on which data? Third question: Was the control plan or work instructions updated? Among the five, if three respond with 'strengthened training,' 'communicated,' or 'responsible person confirmed effective,' stop there—the figure could be 98%, and the onsite atmosphere is already cold.

What clients and certification bodies truly value is not the color blocks on the board, butWhether improvements occurred, whether they were effective, and whether they were written into standards that will continue to be implementedThe quality management plan organizes CAPA/8D into problem, cause, action, effectiveness verification, and document update for linked management. The closure criteria are configurable and include verification requirements, precisely to prevent the mindset of 'filling out the form equals closure.' Inflated closure rates boost face value but bury recurrence and audit nonconformities.

"'Close'—if it can be completed with one click, the closure rate measures the speed of filling out the form."

Redefine closure as the requirement to simultaneously satisfy three conditions:

1. The measures have been implemented——There is verifiable evidence on the ground (changes to tooling, addition of inspection items, system validation launch, feedback of supplier process changes, etc.), not just 'notified' or 'planned for next week'.
2. Validity has been verified——Confirm with the observation period according to the agreed method in advance; the verifier is separated from the person responsible for the measures; the door must not be closed before the end of the observation period.
3. The relevant documents have been synchronized——FMEA, control plans, work standards, and inspection specifications have been revised, or are marked in writing as 'No revision needed after evaluation' with the evaluator noted.

The third item is the most commonly overlooked and also the most critical: the on-site practices have changed, but the guidelines are still the old ones, so the next shift continues to follow the old standards, and the problem will inevitably return. If any of the three elements is missing, the state machine can remain in 'under verification' or 'document under revision,' but it should not be counted in the 'successfully closed' denominator. Having overdue items remain stuck in 'in process' without being counted in the denominator is another form of inflation—the denominator must honestly grow over time, and the overdue list should be exposed separately.

Before writing access control into the system configuration, first use spot checks to solidify the procedures: randomly select closed orders, and if three questions are not answered completely, reopen them. If procedures cannot withstand spot checks, launching access control will only drive more 'other' attachments and empty measures. The separation of responsible persons and verifiers must also be reflected in the organizational chart—when a small team lacks personnel, at least ensure cross-shift or cross-functional verification, and it is forbidden for the same account to both write measures and approve them.

Four types of water injection, how to recognize them at a glance

The most covert is the first type. Training can be part of the measures, but simply "enhancing training" without attendance records, assessments, or evidence of behavior change cannot be rebutted or proven. An effective way to write it should be close to: "Operation cards add self-check item X, with N consecutive batches having no similar defects" or "Incoming material sampling plan upgraded to stricter standards, with zero occurrence of the same defect from the supplier within the observation window."

Overly broad root cause classification (always 'improper operation') can blind the repeated issue rate—not because of statistical incompetence, but because of failed classification management. Root causes should be detailed enough to guide measures; if the same code recurs within the observation window, it should automatically be linked to the original CAPA and escalated (increasing severity, requiring management review, etc.).

Three signals that are more honest than the close rate

It is recommended to place the main board's focus on the C position of the loop rate, and instead look at the combination:

The closure rate is only used as a supplementary reference after the access control system goes live. Additionally, attach two operational tables: a list of overdue items not being stopped, and a list of items reopened/recurring in the recent quarter. There is no need to stay up all night organizing this before the review—just open them as needed in normal times, and the review will pass. If on-time closure rates are increased through repeated delays, the average number of delays and the maximum overdue days should also be disclosed, to prevent 'on-time' from being legally inflated through loopholes in the system.

Severity-graded protected resources: Minor issues go through a simplified CAPA (cause and verification are still required), while major issues and safety-critical features follow the full 8D process. Always using the full 8D would overwhelm the site with paperwork; always simplifying would let major problems slip through. Classification standards should be written into the system and frozen to avoid temporary 'special cases.' Fast response processes (such as on-site QRQC) can shorten containment time but do not replace subsequent root cause analysis and verification—successful containment does not equal CAPA closure. Customer complaint entries and internal non-conformance entries should be integrated into the same closure requirements, otherwise a split board phenomenon occurs where 'customer complaint closure rates look good, but internal recurrence goes unacknowledged.'

Make a Pareto chart of the recently closed cases for the last quarter based on root cause coding: if the top few items remain unchanged for a long time, it indicates a failure in measure quality or document synchronization, rather than 'more training measures are needed.' The agenda of the improvement meeting should focus on whether the material changes for the top Pareto items have been implemented, rather than on the month-to-month increase of a few points in the closure rate.

How to weave nonconformance, suppliers, and documents into a main thread

CAPA is not another unrelated table. When nonconformities or customer complaints reach the threshold (significant, recurring, customer disruption, etc.), a CAPA should be created or linked; upon closure, it should be possible to trace from the nonconformity record to the actions, verification data, and document revisions. When issues with incoming materials are issued to suppliers via 8D, the external portal can report progress, but the verification and closure standards remain the primary responsibility of the OEM quality team—if the supplier portal shows 'closed' without verification data, the closure rate of the OEM is similarly inflated.

File linkage must require mandatory confirmation before closing, rather than 'fix it later when you have time' after closing. If measures change work methods but the controlled files are not updated, it is equivalent to leaving improvements in personal memory. If changes in measurement or inspection standards are part of the measures, they should be linked to the corresponding controlled objects to avoid quality events and file modules talking past each other. Isolation and flow blocking address 'defects are still circulating'; CAPA addresses 'whether similar problems will occur again'—both are needed, but assessment metrics should not be combined into a single number.

System side should align: cannot close with three missing elements; the person responsible for measures and the verifier should be separate; tasks in the observation period should pop up when due; overdue and duplicate issues are automatically exposed; tiered templates are optional. Prompting overdue and duplicate defects is valuable, but it is not appropriate to automatically re-judge or automatically close — closing is a management action with evidence.

The CAPA capability of JEDP-QMS emphasizes management along the same chain and configurable closure requirements. Overdue anomalies and CAPA statuses on the quality dashboard should originate from the same source as the documents. Metrics should be addressed in the daily service meeting and management review, rather than creating an additional monthly report for show. In Phase One, you can first integrate closure controls and overdue lists into the weekly meeting. Detailed root cause coding and repeated issue statistics should be addressed after the quality of actions begins to be questioned, to avoid overwhelming indicators right from the start and to prevent having to manage everything on-site at once.

Observation period and intermediate states: allowing 'unclosed loops,' prohibiting 'false closed loops'

Many organizations get stuck on psychological barriers: a Kanban board cannot show 'incomplete'. As a result, tasks are closed before the observation period ends, or 'measures implemented' is written as already completed. A healthier state machine allows intermediate states—measures implemented, verification in progress—and lists the expected verification date separately in the weekly meeting. Intermediate states occupy execution transparency, not shame; false closure consumes trust, and a single audit can reveal it.

The observation period length should be set according to the consequences and frequency of failure, rather than a uniform thirty days. Rare major failures may require a longer window and stricter verification; frequent minor issues can use a shorter window but must be measurable. When the verification method is written in the measures field, it must be executable: sampling plan, data sources, pass thresholds, and escalation path in case of failure. Temporarily adding verification instructions when closing should be regarded as padding and rejected.

If internal assessments only tie to the closure rate, the organization will optimize for the speed of closing cases; if they also tie to the rate of recurring issues and the rate of effectiveness, then the speed of closing cases will give way to the quality of measures. In management review materials, it is better to have one less attractive chart and two more lists of overdue and recurring issues—the latter is the evidence of whether improvements actually occur.

Don't trust the bulletin board yet, restart what needs to be restarted first.

Take the ten most recently closed CAPAs and check them item by item against the three criteria. If more than half pass, it means the framework is still intact; if less than half pass, reopen them and supplement the evidence before discussing the board revision. Do two small things simultaneously: formally freeze the closure definition and classification in writing; in the weekly meeting, regularly look at the overdue queue and recurrence list, not at 'our closure rate has risen again.'

When there is no clear person responsible for improvement, first assign an owner before assigning the module—otherwise, tasks will be automatically generated but no one will follow through. When the team is not yet capable of basic root cause analysis, first supplement the method, otherwise the system will only faithfully record the four words 'improper operation.' Creating a CAPA just to handle a sudden audit is going in the wrong direction; the indicators will only become a new performance show. If customer complaints and CAPA are only filed as forms without effectiveness verification, it also does not meet the system's common requirements for closed-loop management—archiving is not closing.

If the recurrence list continues to grow after two weeks, the problem is usually not that 'people don’t close tickets,' but that the measures are unverifiable or the documents are not synchronized. Lay out the root cause codes of the recurrences in those two weeks and ask the codes: what observable objects did the last measures actually change? If you can't answer with tangible changes, don't hold a recognition meeting again; first, reopen those few tickets. A slight drop in the closure rate is not frightening; what is frightening is having perfect numbers while the failure modes return unchanged.

• QMS quality solution
• IATF Six Vulnerable Points
• Incoming Material and Supplier Quality Closed Loop
• Nonconforming Isolation and Circulation Block