Arguably, acceptance sampling is an effective and efficient means to ensure the proper surveillance of product and service quality.  Using this approach, a random sample (of product or service) is inspected or tested to determine if the observed level of quality is acceptable or unacceptable.  If the sample supports the hypotheses of unacceptability, then certain corrective actions are invoked.  On the other hand, if the sample supports the hypothesis of acceptability, then no action is taken.

Owing to the nature of acceptance sampling, it is possible to design a set of actions based on certain probabilistic alternatives (outcomes).  In this manner, the governance of a process is made more efficient and less prone to decision error.  In short, the methodology allows the quality of a process to be regulated with known degrees of risk and confidence.

One such methodology is fully described in MIL-STD-105D: Sampling Procedures and Tables for Inspection by Attributes.  Recall than an attribute is a characteristic or feature of some product or service that either conforms or fails to conform to its respective quality standard.  For example, a unit of product (or service) is either delivered on time, or it is not – there is no middle ground, so to speak.  Therefore, attribute sampling is generally useful for describing how well a job is done, in terms of defects per hundred observations, or percent defective.

Essentially, there exist five general types of sampling plans – each having its own merits and drawbacks.  The following is a brief summary of each type:

1) Single sampling plans: A random sample is selected from a lot of product (or service).  Next, the quality disposition of the lot is decided based on certain inspection criteria and various tabulated values.  If the observed case is in agreement with the theoretical case (tabulated statistics), no action is taken.  On the flip side, action is taken in the event of disagreement.  Of interest, single sampling plans are the most common type and are usually the easiest to deploy and implement, although not necessarily the most efficient in terms of sample size.

2) Double sampling plans: Once the first sample is inspected, there are three possibilities – the lot is accepted, rejected, or no decision is rendered.  In the event of indecision, a second sample is taken.  In turn, this information is combined with the first sampling so as to facilitate a final decision.

3) Multiple sampling plans: This is an extension of the double sampling plan – more than two samples are needed to reach a conclusion. The advantage of multiple sampling is smaller sample sizes.

4) Sequential sampling plans: This type of sampling plan is an extension of the multiple sampling plan – units of product (or service) are sequentially selected and then progressively inspected (or tested).  After each unit is examined, a decision is made to accept the lot, reject the lot, or select another unit for inspection.  Of course, sampling continues to a predefined point.  If the lot has not been rejected up to this point, it is accepted with a known degree of statistical confidence.

5) Skip lot sampling plans: Skip lot sampling only inspects a fraction of the submitted lots.