Acceptance sampling to assess the quality of incoming products arrived in the 1920s as an efficient alternative for 100% inspection. However, nowadays more and more industries consider acceptance sampling as something from the past. Why should you bother as a manufacturer with checking the quality of your suppliers at all? Let them convince you by using methods of statistical process control of the day-by-day quality of their processes. Or, isn’t it that simple?
Inspection of 100% of incoming products is very expensive. It is expensive in terms of time, handling efforts, and labour costs. Moreover, 100% inspection sounds as if each item is individually and perfectly assessed with respect to their quality. However, in practice this is not the case. As a matter of fact, new defects may be introduced and due to the tiresome nature of the inspection job mistakes are readily made. Also, it may well be that the production team is not motivated to deliver high-quality products as the products will be inspected anyway. Finally, this method of 100% inspection is not always applicable, e.g., when inspection is destructive.
Alternatively, acceptance sampling may be used. Rather than inspecting all incoming products only a small sample from a lot (or, batch) is taken to decide on accepting or rejecting the lot as a whole. The approach yields a strong reduction in inspection time and costs, and besides allows for destructive testing of products. This sampling approach, however, induces also risks as the small sample does not provide a full picture of the lot quality . Thus, the decision to accept or reject a lot based on the limited information from the sample might be erroneous and this risk can only be alleviated at the cost of increasing the sample size. Therefore, high quality requirements cannot readily be guaranteed by acceptance sampling, as this would require very large samples to achieve an acceptable risk level.
In order to overcome these sampling issues one may check the quality one stage earlier, namely in the production line itself. The manufacturer thus requests evidence that the production processes of its supplier are in control and capable of delivering the quality as specified. The supplier will do this by applying methods of statistical process control and measuring the relevant process parameters to verify that specification targets are indeed consistently being met. In this way, the manufacturer may refrain from acceptance sampling at all.
In many production environments, statistical process control indeed appears a great solution to guarantee high quality end products. Unfortunately, it is not applicable in every environment due to a variety of reasons. It may just require too large investments (e.g., measurement tools, measurement technology, training of operators, reduced production efficiency) and efforts to measure the critical product parameters on a daily basis in production. It may also be that volumes are too low or it might just be too expensive to create a (near) defect-free production process, such that 100% inspection will always be necessary.
Hence, have 100% inspection and acceptance sampling died out and has statistical process control taken over? No, definitely not! The specific demand of the situation at hand will tell which way to go.
There is vast experience and background information available with respect to setting up a sample plan to manage the risks on wrong conclusions in a specific situation. E.g., see next references for more insight.
– A.J. Duncan – Quality Control and Industrial Statistics (4th edition, 1974)
– E.G. Schilling and D.V. Neubauer, Acceptance Sampling in Quality Control (2nd edition, 2009)
– D.J. Wheeler and D.S. Chambers, Understanding Statistical Process Control (2nd edition, 1992)
– W.H. Woodall, D.C. Montgomery – Research Issues and Ideas in Statistical Process Control (1999)