Shifting the Paradigm: Assessing The Effectiveness of AQL In An Organization

In previous blogs, I described the challenges of acceptable quality level (AQL) testing and why the approach does not necessarily provide us with the best decision-making process relative to lot compliance.

If AQL testing is commonplace in an organization, consider viewing the AQL’s department activity as a process where input to the process is various types of batches that enter this process. The output to this process is whether the lot failed or not. Next, calculate the proportion of lot failures (p) for some period of time, e.g., weekly, by dividing the number of lots that failed by the number of lots that were tested. When determining the subgroup time-period interval, we need to have a large number of lots tested during each time period so that several lot failures occur.

If we have a similar number of lots tested during each subgroup, we can plot these p values on an individuals control chart and assess the frequency of lots. I understand that traditional control charting would suggest a p-chart; however, p-charts can create false signals. I will address this issue in a future blog.

For regions of stability in the control chart, a Pareto chart could then be created to summarize overall the types of failures captured within the AQL department during these periods of time. This insight could give the organization direction on where to focus their process improvement efforts.

A previous blog described how AQL sample-size uncertainties translated to customer-quality-delivery uncertainties. An organization could attempt to quantify how many unacceptable lots are missed by the AQL department because of a low sample size through a re-testing effort of some lots that failed and did not fail. If this were done, I would suspect that many of the failed lots would pass the second time. This information collectively could be very valuable in obtaining a better overall picture of the value of the department and in determining where improvements could be made.

In my next blog, I will describe a methodology that can often be used in lieu of AQL testing.

The content of this blog was taken from Section 21. 12 of IEE Volume III .