ASA Connect

I would like to share ideas, experiences, and developments with anyone who applies, teaches, or otherwise is interested in acceptance sampling quality control.  AQL, AOQL, and continuous sampling plans are key words for that field.  The origin of the field is nearly a century ago and the surge of development appears to me to have peaked in the 1960s.  Industry and the American Society for Quality have continued interest in quality control and its development.  Government statistical agencies (at least the Census Bureau) have made good use of such techniques in data processing, particularly data capture, coding, and other desk review tasks.  The practical challenges of that work beg developments beyond the content of common textbooks or other documented sources of which I am aware.  

In response to the questions and challenges of coworkers I have explored and extended theoretical derivations, devised data-computed metrics, conducted research, notably with simulations, to confirm and expand a framework for interrelating, understanding, and optimizing QC inputs and outputs, even from different contexts like AQL, AOQL, and continuous sampling.  I presented papers to ASA Joint Statistical Meetings in 2006 and 2011 that dealt in reserved way with such developments.  Recent emphasis on "near real-time" and cost management have highlighted fresh challenges for interpreting and applying this framework.  Much of this recent growth has not yet been documented beyond the draft stage, partly because continuing developments lead to reorganizing or revising the content.

Another reason for the delay in documentation is that the field of quality management is under much pressure to conduct applications in more and more operations, while funding is more and more constricted.  Ironically, there appears to be low priority on developing ways to optimize timeliness and cost with quality.  Yet the context of practical QC applications is what stimulates real advances.

My ASA papers have not inspired any response outside my workplace.  I have not seen related recent publications.  It may be that I am not looking or putting my ideas out there in the right places.  Now is the right time to see if there is any interest abroad.  Please respond if you wish to engage in this topic. 

------------------------------
Glenn Wolfgang
U.S. Census Bureau
------------------------------

I do a lot of work in acceptance sampling for manufacturers, especially in medical devices. There was a time when quality engineers were well versed in sampling methods, especially attribute sampling, in the inspection context. More frequently today, I find an opinion that inspection is a non-value added task and should be eliminated. Consequently, there is little interest in sampling plans.

Unfortunately, many companies still use acceptance sampling, but the depth of knowledge is lacking. It is more common to find an application of a standard, such as Z1.4, in which the activity is confined to the tables, while nobody in the company understands the OC curve, the meaning of AQL & RQL, etc.

I don’t envision the demise of inspection or acceptance sampling. In a contrary thread, I see a lot of discussion of risk-based xxx, where xxx means any activity in the company. The importance of supplier management, especially in medical devices, coupled with lack of general understanding of risk and sampling, seems to be an area ripe for exploration.

------------------------------
Dan O'Leary
------------------------------