AQL (Acceptable Quality Level) vs. LTPD (Lot Tolerance Percent Defective)

[Lyndon Diong]

I favour LTPD rather than AQL. But some people disagree. Could someone let me know the practical advantage of AQL?

Thanks
Lyndon

 

[Marc]

AQL is Acceptable Quality Level.

LTPD - Lot Tolerance Percent Defective: the value of incoming quality where it is desirable to reject most lots. The quality level is unacceptable. This is the RQL expressed as a percent defective.

 

[Don Winton]

Lyndon,

Acceptable Quality Level (AQL) is the maximum percent defective that is considered satisfactory as a process average by the producer and consumer. In other words, if, on average, 4% (AQL=4.0) nonconforming product is acceptable to BOTH the producer and consumer, then the producer agrees to produce, on average, 4% nonconforming product.

Lot Tolerance Percent Defective (LTPD), expressed in percent defective, is the poorest quality in an individual lot that should be accepted. The LTPD has a low probability of acceptance. In many sampling plans, the LTPD is the percent defective having a 10% probability of acceptance using an accepted sampling plan. With this plan, the producer agrees to produce just enough nonconforming product such the consumer will accept the lot using the agreed to sampling plan and AQL level.

The primary difference is AQL is the average quality level over a series of lots where LTPD is lot-to-lot based. My suggestion: If you a producing a series of lots, use AQL to produce the nonconforming product. If you are producing single isolated lots of product, use LTPD to produce the nonconforming product.

Best Regards,
Don

 

[Marc]

Don,

Now that I know you're a statistician, I know who to direct e-mail to!

 

[Don Winton]

Not statistician. Not sure I would like to wear that hat. However, I was a practicing quality engineer before I started this ISO stuff and like to keep those skills sharp. Feel free to direct any e-mail my direction at your discretion. I will try to help.

Best Regards,

 

[Marc]

Yeah - I was trying to remember back some years from when I worked as a QE. I dealt with AQL's but not LTPD. For over 10 years I have been operating in the PPM mind set. In my mind the difference was semantics as both were part of scheme where an agreement on a certain number of defectives seemed silly.

In retrospect, I can see more of a reality now that I'm 48. 20 to 30 years ago processes and technology (including information gathering / analysis technology) were nowhere near what they are today. Expectations have grown with technology. Look at cars today. Even 10 years ago this was evident.

Today we use poke-yoke and such to ensure no bad parts are shipped. Inspections are typically built in-line and are unobtrusive. Making bad parts are part of reality. All theory asides, you have to have some threshold. "Today's" measureable is PPM.

 

[Marc]

This is from a user group message base:

-------snippo------

Whenever people speak of Acceptable Quality Level (AQL’s) and Lot Tolerance Percent Defective (LTPD’s) and anything else regarding sampling, I am reminded of the following. This is reprinted from Roger Slater’s book “Integrated Process Management” published by McGraw Hill (I would suggest this book to anyone that is involved in quality management). Used without permission.

On August 13, 1865, Dr. Ignaz Semmelweis died…a victim of learned ignorance. Official records of the time listed the cause of death as puerperal infection; a disease Semmelweis spent most of his life fighting. But the real culprits were medical colleagues, so entrenched and contemptuous in their resistance to his methods that, after 18 years of fighting conventional wisdom and professional ridicule, Semmelweis suffered a complete mental collapse. Some say his death was suicide, brought on by a self-inflicted surgical wound, which he plunged into the abdominal cavity of a cadaver, just succumbed to childbed fever.

Dr. Semmelweis suffered the fate of those who try to change things. Although he persevered longer than most and paid a more ultimate price, Semmelweis came to know the same hostility and rejection that have come to others, before and since, who challenge prevailing wisdom and dare to show that things might be done differently and better.

In 1848-at roughly the same time as Louis Pasteur was busy establishing the validity of his “germ” theory of disease in Paris and Joseph Lister was developing antiseptic procedures in London; Semmelweis was working in Vienna on asepsis as a preventative of childbed fever. In those days, throughout Europe, of every 100 women who came to an obstetric clinic to have their babies, 25 to 30 would not leave there alive. They could be expected to contract and die of puerperal infection. Most authorities were convinced that the disease was not preventable, induced by overcrowding, poor ventilation, the onset of lactation, or “miasma.” Semmelweis not only disagreed; he was blasphemous enough to suggest that physicians themselves carried the infection.

In his earliest test of this idea, he had medical students wash their hands in a solution of chlorinated lime before examining patients. Mortality rates fell immediately in his division from 10 per 100 mothers to 1 per 100. His message was simple: if doctors would but wash their hands before examining patients, infection could be minimized and lives could be saved. Semmelweis was dismissed from the clinic a year later.

What Semmelweis overlooked in his attempt to convince others of the simple expediency of clean hands was the symbolism of the time. Physicians did not want to use a chlorine hand wash because it robbed them of an important sign of status. An accepted practice among physicians in those days was simply to wipe one’s surgical smock after treating a patient. The often-smeared surgical smock came to be a visible sign of one’s professional importance. Doctors were not about to give up so handy an index of personal worth just because some crackpot had data to show that they could save lives by washing their hands. They consciously and clearly decided to ignore-indeed, reject-factual information in order to protect and retain intact a personally cherished practice. They chose to be ignorant.

What can be learned from Semmelweis’ experience? First, we can learn that intentional ignorance can be a problem, even among the brightest of people. And, secondly, we can perhaps learn something about management for, like those 19th century physicians, many modern managers are responsible for the spread of problems in their organizations. Managers, too, practice learned ignorance.

We are faced with a paradox. We usually equate ignorance with a lack of knowledge, with too few opportunities to learn and broaden our repertoires. But, as we have seen, ignorance itself may be learned when remaining ignorant suits our personal purposes.

Why is it that people refuse to accept the obvious. It always amazes me when people use AQL and LTPD as an acceptable method (excuse?) to produce nonconforming product. By definition, AQL suggests that a certain percentage of produced product can be nonconforming, and THAT IS ACCEPTABLE! Granted, this AQL is usually agreed to between the customer and producer, but to what end. Is the consumer then producing an upper level product using an AQL from vendors to pass this level of quality to its customers. How absurd!

This is what happens when producers forget what the “P” in SPC stands for. Under the assumption that a process is operating at a capability of 4% (AQL=4.0) nonconforming, then it is accepted that 4%, on average, will be nonconforming. Would it not be more feasible to increase capability to a point where there are less than 10 PPM nonconforming and provide SPC data to verify this? Would this not decrease costs to both the producer and consumer? Would this not decrease the costs associated with sampling at all? The only sampling required would be the samples required for the SPC (or equivalent methodology) chart(s). The producer increases capability to produce more acceptable product, reducing its costs. The consumer reduces costs by reducing the number of complaints, returns, etc. Acceptance sampling is an antiquated practice that should be reviewed carefully before implementing. Do not be fooled by quick fixes and do not fall into the trap of “learned ignorance.”

 

[Don Winton]

It is nice that a QE (former or otherwise) is here and willing to discuss these things.

I concede that an agreement to “produce a certain number of defectives is silly.” Not only that, it is absurd and, IMHO, stupid (see post above). It seems a shame that sampling and AQL’s are still a part of the ASQ’s portion of the CQE certification process (which, among others, is why I did not choose to become certified. But that is another story).

AQL’s are outdated. Process capability and PPM are the measure of the quality characteristics of today’s standards.

IMHO, I have become disillusioned at the QE profession, in general (Altho, there are still a a lot of very good ones still out there). It seems that Deming, Juran and others are ignored in order to implement a Quality Management System that is “easy.”

“Do It Easy, But Do It My Way.” Anon.

“How Difficult It Is to Save the Bark of Reputation From the Rocks of Ignorance.” Fransesco Petrarch

The opinions expressed here are those of the author and do not reflect those of the host.

Best Regards,

 

[barb butrym]

Well..well..a bit of a philosophiocal offering today, eh?

I like that hand washing story....I'll adopt it in my training...I always like to add a bit of story telling to bring home a point

 

[Roger Eastin]

Having worked with Ed Schilling a number of years at GE (he wrote a book on Acceptance Sampling), I'm not so convinced that A.S. should be a favorite whipping post. Dr. Schilling favored using AQL's as a benchmark of a process that should be re-negotiated periodically, much in the same vein that we re-negotiate Cpk's. In fact, some processes are more aligned with AQL-style sampling and I think that, in this case, Dr. Schilling's idea is a good one. Why not agree to an AQL of 3.0% and in a year, re-negotiate the AQL down to 2.0% (based on process improvements,etc.)? I don't think that sampling plans are the right target. I think that being satisfied with the level (Cpk, AQL, etc.) we are at presently is the real target. I know Deming liked to whip the AQL subject, but I think it was because managers were satisfied with 4.0% defective! Can we not do the same thing with any other indicator? Personally, I get discontent with quality management whenever we as a group become satisfied at a certain level of quality - whether that is attaining a Cpk, attainging an AQL or compliance with ISO/QS9000.