The Sampling Plan That Outran the Process
You are the lead auditor on an internal audit of a precision-machined bracket line. The audit plan specifies a random sample of 25 of the 4,200 units produced in the last quarter, drawn from the final-inspection log. During fieldwork you discover the log is a spreadsheet maintained by the shift supervisor, and the last 60 rows were entered in a single batch on the final day of the quarter, all with identical timestamps and inspector initials. The process specification requires 100% dimensional check on a key characteristic, but the log shows only 25 checks for the quarter. The supervisor says the batch entry was 'catch-up paperwork' and the checks were done on the floor. Your sample of 25 shows 2 nonconformances on the key characteristic. The client's audit objective is to determine whether the 100% inspection requirement is being met.
1. Evaluate whether your planned sample is fit for the audit objective and describe how you would revise the approach.
Read the worked answer
The planned sample is not fit for purpose for two reasons:
- Population integrity is compromised. The sampling frame (the final-inspection log) does not reliably represent the 4,200 units. A 60-row batch entry with identical timestamps and initials is a strong indicator of retrospective documentation, so the frame may over- or under-represent the true inspection population. Sampling from a corrupt frame produces conclusions of unknown validity regardless of sample size.
- The objective is compliance, not rate estimation. The objective is whether 100% inspection is met, which is a binary conformance question answered by evidence of inspection on every unit, not by a defect rate from 25 units.
Revision: (a) Re-establish the population by reconciling production records to inspection records unit-by-unit for a defined window; (b) switch to an attribute/zero-defect oriented approach - test the completeness of inspection evidence across the full quarter, not a rate; (c) expand to a stratified sample by shift and week if a rate estimate is still needed; (d) document the frame deficiency as an audit finding in its own right and notify the audit program manager, since the scope may need to expand to other lines using the same log.
2. You have 2 nonconformances in 25. A colleague proposes reporting '8% nonconforming rate.' What is wrong with that, and what would you report instead?
Read the worked answer
What is wrong:
- The point estimate (2/25 = 8%) has an enormous confidence interval at n=25 - the exact 95% binomial interval is roughly 1% to 26%. Reporting '8%' implies precision the data cannot support.
- The sample was drawn from a suspect frame, so even the interval is conditional on the frame being valid.
- The audit objective was compliance with 100% inspection, not estimation of a defect rate. Reporting a rate answers a question the audit did not ask.
What to report instead:
- A compliance conclusion: inspection evidence was not demonstrated for the majority of the quarter's production; the 100% requirement is not evidenced as met.
- The two observed nonconformances as corroborating evidence of a control breakdown, described qualitatively.
- The frame/documentation integrity issue as a separate finding with its own risk statement.
- A clear statement of limitations so the reader understands what the sample can and cannot support.