Two Overdue Gauges Out of Thirty
During a supplier audit at a machining supplier you sample 30 of 600 gauge calibration records using a random attribute plan. Two gauges are past due - one by 45 days, one by 60. Both are micrometers used at final inspection of parts shipped to you. The supplier's procedure QP-8.4 requires recall within 30 days of the due date and a monthly overdue-gauge report; no overdue report exists for the last six months. The quality manager argues the two affected lots were re-checked on a calibrated CMM anyway, and asks you to write a single minor nonconformity because 'it is only 2 out of 30'.
1. What can you legitimately conclude from 2 of 30, and should you expand the sample before writing the finding?
Read the worked answer
The sample is weak evidence about the population; the documentation gap is strong evidence about the control.
- Point estimate 2/30 = 6.7%. The exact 95% confidence interval is roughly 0.8% to 22% - that spans 'one gauge in a hundred' to 'one in five'. It supports some failure rate but cannot support a statement such as 'approximately 7% of gauges are overdue'.
- The missing monthly overdue reports for six months are not a sampling result. That is a direct observation of a required monitoring control that did not operate. One documented control failure establishes a systemic condition without any extrapolation.
Sample actions, in order of value:
- Targeted, not larger-random. Stratify by gauge type, area and last-calibration date; screen 100% of gauges used for final inspection and for customer-specific characteristics if the population is manageable. A bigger random sample from 600 buys precision you do not need once the control failure is established.
- Keep a record of the original random result and the expanded screening separately, so the report does not blur what was sampled and why.
- Verify the 600-record population is complete - an inaccurate population list would undermine both the sample and the control.
2. Major or minor? Address the 'we re-checked on the CMM' argument explicitly.
Read the worked answer
Major, on these grounds:
- Systemic, not isolated: a required control (monthly overdue reporting) is absent for six months, so the organisation has no mechanism to detect overdue gauges.
- Directly affects product acceptance: the gauges are used at final inspection of product shipped to the customer. Measurement traceability failure at acceptance is a high-risk condition.
- Repeatable and undetected by the supplier until the audit found it - the supplier's own process did not surface the condition.
On the CMM argument: re-measuring the two lots may be valid containment for those lots, and you should verify it actually happened (records, which characteristics, calibrated status at the time). It is not a correction of the system that allowed overdue gauges into use, and it says nothing about the lots measured with other overdue gauges you have not found. Containment reduces the immediate product risk; it does not reduce the severity of the management-system failure.
Also test the supplier's claim that only two gauges are affected - that claim is itself unverified until the monthly-report control or a full screen exists.
3. Write the nonconformity statement as it should appear in the report.
Read the worked answer
Requirement: QP-8.4 requires gauges to be recalled within 30 days of the calibration due date and requires a monthly overdue-gauge report.
Condition: At the time of audit, 2 of 30 sampled calibration records were overdue (45 and 60 days); both gauges are micrometers used at final inspection. No monthly overdue-gauge report had been produced for the preceding six months.
Evidence: calibration record IDs and due dates for the two gauges; review of the QP-8.4 requirement; confirmation from the quality manager that no reports exist for the six-month period.
Why it matters: measurement traceability at final inspection is not assured and the supplier's detection control did not operate.
What to keep out: the root cause ('the clerk was overloaded'), the fix ('we will buy software'), and unsupported generalisations ('most gauges are overdue'). Those belong to the corrective action response. Keep the statement factual, verifiable and free of remedy language, so the response can be assessed against it.