Any comments for Quality @ AQL=0.1, C=0 ???

ASQC re-numbered the once free (taxpayer sponsored development) MIL-STD-105 as ASQ-Z1.4, slapped their own cover on it, and now charge $40 or $50 for the same document! The standards are identical, so see if you can get one of the old �MIL-STD's� and then get your money back!

Bottom-line answer: It depends. If AQL=0.1, C=0 is OK with your customer, then it�s OK by us.

AQL. Acceptable Quality Level. The maximum percentage or proportion of nonconformities in a lot or batch that can be considered satisfactory.

Continuing, as the frequency of the occurrence of defects decreases, the probability of detection of any individual defect decreases, as one might expect, and that research result was reported by Malcolm Gladwell in the �New Yorker� last fall [ http://www.newyorker.com/fact/content/?011001fa_FACT ]. That's why: * Paying airport inspectors more won't make planes safer. * Training airport inspectors more won't make planes safer.

Proof of the pudding is in the eating: �Tests Show No Screening Improvements� Blake Morrison, USA TODAY, 03/25/2002 [ http://www.usatoday.com/news/attack/2002/03/25/usat-failures.htm ]. Airport inspectors are good at finding fingernail clippers, but lousy at finding guns and bombs.

Some companies use AQL=0.1, C=0 at incoming, when they have no history. But recognize that AQL is a flawed concept. It is an anathema to �zero defects� and �six sigma� quality orientation, where companies improve their process, rather than saying: * �Awww shucks, that�s not toooo bad.� OR * "Close enough for government work, eh?"

Taking a different angle, Ed Schilling at RIT wrote �Acceptance Sampling In Quality Control�. It includes a very comprehensive description of a Lot Tolerance Percent Defective (Dr. Schilling's a good guy, but he didn't spend a lot of time thinking about his choice of terminology.) [LTPD] plan he developed that: * Allows acceptance only on zero nonconformances. * Defines the largest fraction nonconforming that has a 0.1 or 0.05 (depending on the plan) probability of acceptance.

Advantages to the LTPD approach are: * Sample size is relatively low. * Plan is oriented to protect the consumer when working from a LTPD, rather than the producer when working to an AQL (as in Z1.4). * Keyed to accept on zero nonconformances, rather than demonstrating that there is ONLY one reject in the lot (Z1.4 allows acceptance, in some cases, with 1 or more nonconformances in the sample).

Returning to the original point on how best to use AQL, over the past few years, IPC added a very simple table to a number of documents [eg, 6012A, p20] that defines zero acceptance number sampling plans. Although based on an AQL, it takes steps to appease the anti-AQL zeolots: * AQL increases according to criticality of customer application. * AQL changes according to seriousness of the defect. * Has a zero acceptance number. * Is very easy to use.

Finally, QA types at the �Quality Magazine� [ http://www.qualitymag.com/ ] forum may comment.

reply »