Reading the Findings Distribution
Over 12 months your internal audit programme raised 84 findings across six departments: Production 28, Warehouse 17, Purchasing 12, Maintenance 11, HR 9, Sales 7. Audit days spent per department: Production 22, Warehouse 8, Purchasing 6, Maintenance 7, HR 4, Sales 3. By requirement, the largest categories are document control 19, competence and training 14, calibration 12 and corrective action 11. Fifty-two of the 84 findings (62%) repeat findings raised in the previous 18 months; document control accounts for 14 of those repeats. The quality manager shows the Production bar to the management review as evidence that Production is 'our worst area' and proposes a three-day re-audit of Production.
1. What analysis do you present in place of the raw department bar chart, and what does it change about the proposed re-audit?
Read the worked answer
Normalise before comparing. Findings per audit day:
| Dept | Findings | Audit days | Findings/day |
|---|---|---|---|
| Production | 28 | 22 | 1.27 |
| Warehouse | 17 | 8 | 2.13 |
| Purchasing | 12 | 6 | 2.00 |
| Maintenance | 11 | 7 | 1.57 |
| HR | 9 | 4 | 2.25 |
| Sales | 7 | 3 | 2.33 |
Raw counts track audit effort, not performance. Production has the lowest finding rate of the six; Sales and HR are highest but on tiny exposure, so treat them cautiously.
Then Pareto by requirement: document control 19 (22.6%), training 14 (16.7%), calibration 12 (14.3%) - the top three are about 54% of all findings. That is where a systems fix pays.
Effect on the proposal: a three-day Production re-audit would mostly reproduce a high finding count because Production is the most-audited department - it is measuring the auditor, not the process. Redirect effort to the document-control and training systems, and to why findings repeat. Present the normalised rate and the Pareto, and say explicitly that finding counts are a detection measure, so they must be read alongside process data (complaints, scrap, on-time delivery) before any department is labelled worst.
2. What does the 62% repeat rate tell you, and what should the next audit focus on?
Read the worked answer
It is a signal about the corrective action system, not about Production.
- A repeat finding means the earlier correction did not remove the cause, or the corrective action was never verified as effective. Audit the corrective action process itself: how causes are determined, whether actions address cause or symptom, who verifies effectiveness and against what evidence, and whether verification is recorded.
- Check whether repeats are literally the same condition at the same location (a failed fix) or the same requirement failing at new locations (a system-wide weakness). These lead to different actions.
- Examine whether findings are closing on 'document submitted' evidence rather than observed effect - a common cause of a high repeat rate.
Next audit moves: a focused process audit of corrective action and document control, using the repeat findings as the sample; verify a handful of previously 'closed' actions in the field rather than in the file. Use root-cause tools (5 Whys, cause-and-effect) on the document-control cluster to find whether the issue is ownership, template control, training or system access.
Caveat to state: repeat rate is also sensitive to auditor consistency and classification drift. Check that the same requirement was cited the same way before drawing conclusions.
3. Monthly finding counts rose from 5 to 12 over the year. A colleague calls this a deteriorating trend. How do you test that claim?
Read the worked answer
First establish whether the chart is measuring quality or measuring detection. Finding counts are output of the audit process, so before any trend claim:
- Normalise by audit effort - findings per audit day per month, and per auditor. If audit days rose or a new auditor joined with a harder interviewing style, the count rise may be entirely detection.
- Check scope and criteria stability - new sites, new processes, a changed checklist or a stricter classification of observations all shift counts without any change in performance.
- Check for a real signal - plot the normalised rate on a run chart or control chart with limits. A run of six or more points on one side of the median, or a point beyond a control limit, is a signal worth investigating; a rising count with rising audit hours is not.
- Triangulate with process data - complaints, scrap, rework, on-time delivery, supplier rejects. If those are flat while findings rise, the story is about auditing, not quality.
Report the honest ambiguity. 'Finding rate per audit day rose from 0.6 to 0.9 alongside a change in auditor and checklist; we cannot separate detection from deterioration, and here is the data needed to do so.' That is more useful - and more defensible - than confirming a trend the data does not support.