The GM asked me to explain to the board about AQL (using layman terms).

[vasilist]

Dear friends,

the time has come in the organisation i work for to use the AQL philosophy in its contracts with the suppliers. Top management gave me the chance to demonstrate to them what AQL stands for and how are we going to use it in order to minimise our expenses from non conformances caused by the suppliers.

Is it possible to show me a way of how to make such presentation in NON - technical words? I love statistics and i love technical language but in this case i am sure that i will not succeed if i use technical language.

Thank you for your time reading this question.

vasilist

[Stijloor]

Originally Posted by vasilist

Dear frinds,

the time has come in the organisation i work for to use the AQL philosophy in its contracts with the suppliers. Top management gave me the chance to demonstrate to them what AQL stands for and how are we going to use it in order to minimise our expenses from non conformances caused by the suppliers.

Is it possible to show me a way of how to make such presentation in NON - technical words? I love statistics and i love technical language but in this case i am sure that i will not succeed if i use it 100%.

Thank you for your time reading this question.

vasilist

How about zero defects instead of AQL's?

An advanced search on The Cove Forums on AQL revealed this.

Some more searching through the threads and posts is necessary.

Stijloor.

[vasilist]

Thank you very much for your answer. Iwill be honest with you. We are no such mature to proceed with Crosby's philosophy!

Yes i would love to but my understaning of "how things work here" is that we are not in this level (at least now). Please do not take tis as a bad opinion of me for the organisation i work for. Just take it as a comment of knowing "where we are". Again thanks.

[Tim Folkerts]

I used to think of Z1.4 type sampling plans as a way to determine quality and control processes, but I have been coming around to seeing them as providing a rough screening of quality. With a pass/fail test, the sample sizes needed to really determine the defect rate (and hence the quality of the lot) becomes huge.

Furthermore, AQL is more of a protection for the producer, not for the consumer.

AQL =1 mean that most of the time, a lot that is 1% defective will be accepted. Basically the producer is saying "You agreed to take lots that are up to 1% defective, so you have to come up with a sampling plan that almost always will accept the lots if they contain 1% defective parts." This means that you would often accept lots that are worse than 1% defective. How often you accept worse lots depends on the actual plan you choose.

Consider a typical plan: a lot of 1000 with AQL=1 and Type II normal inspection. You need to inspect 80 pieces. Lots with 1% defective will be accepted 95% of the time, but lots with 3.33% defective will be accepted 1/2 the time, and even 7.66% defective will be accepted 5% of hte time.

If you want to be better at catching slightly bad lots, you need a bigger sample = more cost.

So I see Z1.4 sampling plans a "insurance" against a truly horrible lot. Maybe a tool broke and 1/10 the parts are bad; maybe 2 different parts got mixed together and 1/4 off the shipment is the wrong parts; maybe something was miscalibrated and all the part are a little to big. But if you want to catch a slow drift in quality, or catch 2% defective instead of 0.5% defective, you won't have much luck if you use sampling.

Tim F

[AndyN]

Originally Posted by vasilist

Dear friends,

the time has come in the organisation i work for to use the AQL philosophy in its contracts with the suppliers. Top management gave me the chance to demonstrate to them what AQL stands for and how are we going to use it in order to minimise our expenses from non conformances caused by the suppliers.

Is it possible to show me a way of how to make such presentation in NON - technical words? I love statistics and i love technical language but in this case i am sure that i will not succeed if i use technical language.

Thank you for your time reading this question.

vasilist

The basic answer is NO! If they don't know statistics, then they will be completely lost! Trying to explain how operating curves are derived etc. is futile.

Plus, by accepting this old philosophy of allowing for defects from a process, you will be on a slippery slope to poor quality.......believe me!

[Tom Slack]

It has been my experience that behind every question there is a statement. Also, behind every statement, there is a question. I wonder why managemet is asking about AQL.

I suggest making a graph of AQL vs Cost. This would demonstrate the dimishing returns on increased sampling. I would anticipate some questions about sampling alternatives.

I have done much more SPC work than Acceptance Sampling, but I think sometimes sampling is a good strategy.

[bobdoering]

AQL (hopefully c=0) is simply a verification of product that should have been contractually verified as coming from a stable controlled process. It is sampling - and sampling always has error - in the parts you did not check, primarily. As Tim says, sampling plans are a cheap "insurance" against a truly horrible lot.

I guess one question may be what are you using in your system now, and: "How is that working for you?"

What is the risk of one bad part getting into your process? How about 5? 25? What is the cost in overhead, labor and materials when they do enter your system? At the board level (if I am interpreting that term correctly for your example) you need to talk money. How much money (overhead, labor and materials) are you losing now, and how much can you save if you implement the new overhead of sampling? Is there a payback? You need to estimate that, because there is a cost to sampling - inspectors, gages, labor, space, overhead, etc. They will get that more than nuts and bolts OC curves.

[qualitytrec]

We use AQL alot but it is not very good at keeping customers satisfied even it passes the AQL it sometimes gets rejected on the fill line.

Here is the direction I would go.

AQL is for lot acceptance. It shifts the majority of the risk to the customer as it allows for a % of defects to be found in the sample and still be deemed an acceptable lot. The % of defects is not a direct representation thus you run the risk of accepting lots with higher % defects than what is found on the acceptance plan. It should not, in my opinion, be used for critical quality decisions but for things considered minor which are acceptable at some level (ie, minor print defects, pits in some types of glass, etc...). Other quality acceptance should be determined through SPC and better data than can be gleaned from AQL.


Mark