The audit ends with the closing meeting. The value starts afterwards, and most of it is lost in the two weeks that follow, when nobody owns the list.
In its first year the valve plant raised 144 findings. Of those, 49 had already been raised in an earlier audit, and only 58 percent of the actions were done by the agreed date. The audits were not the problem. The twenty days after them were.
Six steps with owners
| Step | Who | When |
|---|---|---|
| Report out with grade, evidence and place | Lead auditor | 3 working days |
| Cause and action, with the evidence that will prove it | Process owner with the team | 10 working days |
| Action in place and visible | Owner of the action | As agreed |
| Effectiveness check: a new sample | Lead auditor | 30 to 60 days later |
| Closure with the check result | Program owner | After the check |
The step people drop is the second half of the second row: naming, at the moment the action is agreed, what evidence will later show that it worked. Without it the check becomes a conversation.
Four levels of one finding
The nonconformity from audit A-16 was that three of eight assembly orders at station 3 had been built against an old revision of the work instruction. There are four different questions behind it, and only the last two change anything.
- The case: what do we do with the three orders? They were checked against revision 6, two were reworked and one released.
- The place: why did this station have an old copy? Station folders were replaced by a screen showing the released revision.
- The process: why can an old copy survive a change at all? The change process got a release step with a signature: the old revision is withdrawn before the new one becomes valid.
- The pattern: where else can this happen? The same check was run at machining and final test, and one more old copy was found and removed.
An action that stops only the case is the reason findings repeat. In the second year, 21 of the 24 nonconformities got an action at process level, and the repeat share fell from 34 percent to 9 percent in the fourth quarter.
The effectiveness check
Four things make a check real.
A date, set together with the action, thirty to sixty days after the action is due and in the plan from the start. Evidence rather than a report: a new sample, drawn by the same rule as the finding, at the same place. A result with two values only, passed or not passed. And two lines of record saying what was sampled and what was found.
For the revision finding, the check came 60 days after the action was due. Ten orders were sampled at the three assembly stations, all ten were built against the current revision, and the finding was closed 71 days after the audit.
Not every check passes. One action in the same batch, a lock that was supposed to stop the same person entering a result and releasing it, was done a week late and failed its check in December. The finding reopened with a new action and a new date. That is the system working, not the system failing.
What the plant can now say
Days from finding to proven effectiveness fell from 96 to 38 in the fourth quarter. Actions closed by the agreed date rose from 58 to 76 percent over the year and to 92 percent in the last quarter.
Those two numbers are worth more than the number of findings, because they describe the only part of an audit that changes a process. A plant that closes findings with evidence stops rediscovering the same gap every year.