An audit program that is not measured is cut in the next cost round, and the people who could defend it have nothing to show. Six figures, collected per quarter, are enough.
Six key figures
| Key figure | Year 1 | Year 2 |
|---|---|---|
| Audits done against plan | 79 % | 100 % |
| Findings per audit | 7.6 | 4.9 |
| Actions closed by the agreed date | 58 % | 76 % |
| Days to proven effectiveness | 96 | 38 in Q4 |
| Repeat findings | 34 % | 22 % |
| Measured benefit per EUR of program cost | not measured | EUR 2.4 |
The last one needs a word. The valve plant program costs EUR 33,280 a year: 264 auditor hours, 168 auditee hours and 80 hours of program management at an internal rate of EUR 65, which is EUR 1,387 per audit. Against that it counted EUR 38,400 of scrap and rework avoided, twelve fewer customer complaints at EUR 2,400 each, and 36 hours less unplanned downtime at EUR 320 per hour: EUR 78,720 in twelve months.
Only effects that were measured in the process after an effectiveness check were counted. An estimate of avoided risk does not survive the next cost round.
Read them together
No single figure means anything alone.
Findings falling while repeats rise means the audits are getting softer, not the processes better. Check the sampling and the grading, and observe one audit yourself. Many opportunities and few nonconformities means auditors are avoiding the uncomfortable grade. Audits not done against plan means the program is losing to daily work, which is a planning problem, not a motivation problem: put fewer audits in the next program and agree the dates with management.
In the second year of the plant, findings per audit fell from 7.6 to 4.9 while the repeat share fell from 34 percent to 9 percent in the last quarter. Falling findings alone would have been a warning. Falling findings with falling repeats is the result you want.
Building auditors
An internal auditor is a colleague with a second role, and the program has to make that role possible. The sequence needs no external certificate, only a record of what was done.
Training on the method, then two audits as second auditor with an experienced lead, then one audit led under observation with feedback on the same day, then qualification to lead audits in named process areas. After that, at least two audits led per year, or the qualification goes to sleep.
The valve plant runs six part-time auditors who led 24 audits between them, four each on average. Nobody audits their own area. Where the best expert for a process works inside it, that person briefs the auditor before the audit: they are a source, not the auditor.
Two routines keep a team calibrated. Twice a year all auditors grade the same five real findings and compare. Once a year each auditor is observed and gets feedback on conversation, evidence and wording.
Before an external audit
An external audit of a management system looks at whether the internal program is planned, independent, evidence-based and acted on. The program itself is the preparation.
What helps is the audit program with dates and done flags, reports that show evidence, places and samples, nonconformities with causes and effectiveness checks, and open findings with realistic dates. Open items are normal. Dates that passed without a note are not.
Do not clean up beforehand. A plant that removes its open findings removes the proof that it finds things.
Audit the audit
Once a year, somebody who does not run the program checks it with the same rules: plan, independence, evidence, grading, actions, effectiveness. It is the shortest audit in the plan, and it is the one that keeps the rest honest.