Sampling, Test Procedures, and Data Recording
Sampling determines how much you inspect and how you decide. Distinguish 100% inspection from sampling. Know random vs representative sampling, stratified sampling (by lot/batch), and the danger of biased sampling (e.g., only easy-to-reach items). Understand AQL (acceptable quality level) and LTPD (lot tolerance percent defective), and that sampling plans carry producer's risk (α) and consumer's risk (β). A sampling plan does not guarantee zero defects; it controls the probability of acceptance at a given quality level. Traps: assuming a sample that passes proves the whole lot is good; confusing AQL with a defect allowance per unit; ignoring lot homogeneity.
Test procedures must be written, controlled, and followed exactly: scope, equipment, calibration status, environmental conditions, step sequence, acceptance criteria, and records. Data recording demands legibility, contemporaneous entry, traceability to the item and procedure, and correction by strike-through-with-initials (never erasure or overwrite). Electronic records need audit trails and access control.
Rule of thumb: record enough that an independent inspector could reconstruct exactly what was done, with what equipment, when, and by whom. Worked example: a lot of 500 units sampled per a plan with sample size 50 and acceptance number 2; if 3 defectives are found, the lot is rejected under that plan - you do not "average" or re-sample to pass it. Re-sampling to obtain a passing result is a data-integrity violation.