Acceptance sampling

Acceptance sampling is a major field of statistical quality control; a key reference on acceptance sampling is [a1].

As a typical application, a company receives a shipment (lot) from a producer. After a sample is taken, a decision must be made regarding disposition (lot sentencing), i.e., whether or not to accept the entire lot. It is not the goal of acceptance sampling to estimate the quality of the lot, or to provide a direct form of quality control, but just to ensure that the output conforms to requirements. Acceptance sampling is useful when 100% or 0% inspection is infeasible due to cost considerations, the destructibility of the product, the reputation of the producer, etc. Acceptance sampling provides for a less expensive alternative to exhaustive inspection, but there is always the potential for error in the decision process.

An acceptance sampling plan is a statement of the sample size to be used together with the acceptance/rejection criteria. Acceptance sampling plans are classified by attributes or variables (cf. also Acceptance sampling plan for attributes; Acceptance sampling plan for variables), and single, double, multiple, or sequential sampling plans are possible.

In a single sampling plan, one sample is taken to base the decision on, whereas decisions may be postponed under the latter schemes if not enough information is forthcoming from previous samples. In general, large, homogeneous lots are of most use to the acceptance sampling process. The units for inspection should be chosen randomly; at times, stratification is of use.

Various plans may be used, depending on the objectives of the organization and the background and reputation of the producer.

References

[a1]	D. Montgomery, "Introduction to statistical quality control" , Wiley (1991) (Edition: Second)