The quality of supplied components is often critical to achieving the desired quality of an end-product. At the manufacturing stage, the quality of the components shipped by the supplier is usually checked with an incoming inspection. But how well do traditional inspection procedures actually protect you from poor supplier quality?
Acceptance Sampling is a technique to determine whether or not you should accept a batch. It is based on sampling, which is useful since you often don’t want to check everything, especially when the batch is large or measuring expensive.
The most widely used method of Acceptance Sampling is the so-called AQL system, described in the ISO norms 2859 and 3951 (and Military Standards 105 and 414). It includes two elements: a sample size and a criterion (also known as Acceptance Quality Limit or AQL). To determine whether or not to accept a batch, the quality of the sample is compared to the AQL.
Two opposing risks
An important feature of the AQL system is that risks are bounded. But what are those risks? Actually, there are two: first, that you accept a batch which has poor quality. Secondly, that you reject a batch of perfectly good quality. The first risk, on the buyer’s side, is often called the ‘consumers risk’. The second, on the seller’s side, is the ‘producer’s risk’.
Statistically speaking, the consumer’s risk is the probability that, given that the total batch is poor quality, you draw a sample of better quality than the AQL. The producer’ risk is the exact opposite: the probability that, given that the total batch is good quality, you draw a sample of poorer quality than the AQL. These risks, or probabilities, are very different.
Don’t overlook a risk
When an AQL system is used, the focus is only on the producer’s risk. Whereas the quality manager’s is interested in the consumer’s risk. It is not widely known that the real risk is two-fold, and that these two risks are different in nature and magnitude.
So when using Acceptance Sampling, make sure you don’t blindly trust the guarantee of a bounded risk without questioning exactly whose risk is bounded. The widely-used traditional AQL systems only protect the supplier, and not the customer. Alternative Acceptance Sampling procedures, which control both the supplier’s and the customer’s risk, are therefore preferable.
Would you like to know more about Acceptance Sampling? Please contact Heleen Muijlwijk.