Was Deming right or wrong? he was right, but many people misinterpret what he really said and they are wrong.
This is one of too many of Deming's discussions that has been simplified to the level of slogan. In the process of attempting to explain a complex concept in simple terms we have not just clarified his dense prose, but we have eliminated the nuances that truly layout the case.
First, given that the point of these discussions is whether or not sampling detects and screens out higher than desired defect rates, teh whole argument goes off the rails when the use AQL only plans. c=0 AQL plans are even worse. Plans for this type of discussion must involve RQLs; the defect rate you want to detect and reject. An AQL is the defect rate you will accept most of the time it is present.
The following scenarios have a foundational assumption of an RQL based plan:
When the defect rate is exceptionally low sampling is useless. Given that hte sample size will be dependent on the defect rate the sample size will be very large anyway. If the defect has a severe effect we must 100% inspect until the process is mistake proofed. If the defect isn't severe then inspection is a waste.
When the defect rate is at the level you want to reject sampling becomes silly. you will randomly fail to reject some lots and randomly reject others. You may improve the over all defect rate if you do 100% inspection but you will still be sending lots that have high defect rates. Again the all or nothing scenario can be invoked.
When the defect rate is above the level you want to reject sampling results in a foregone conclusion. You will reject almost every lot. And again the all or nothing scenario can be invoked.
BUT when you have a defect rate below the level you want to reject, sampling does provide protection against increases in the defect rate.
Of course Miner is correct, in the long run the best thing to do is to improve your processes and control them through various means available: SPC, mistake proofing, controlling inputs, etc.
Two interesting articles on this issue are:
?Mood?s theorem, Deming?s kp Rule and the Death of Acceptance Sampling? Steven Rigdon, Quality Engineering, 8(1), 129-136 (1995-96
?Acceptance Sampling? The Enterprise Strikes Back! AS9100 c=0 Plans; When Slogans Supplant Science? Quality Engineering, 18:237-236, 2006
Both are available through ASQ?s knowledge Center?
We must always remember that no single QC tool was ever meant to be the only tool. They all have their time and place?a hammer should never be used to drive a screw and a screw driver makes for a pretty lousy hammer?
slogans should never replace thought.
:
