This site is great by the way...I have found a lot of good information so far...

Now to my issue...I am the quality manager of a small contract manufacturing company (23 employees). We make a heat pack for a medical device company.

I am NOT a statistical guru by any means and I actually inherited the current AQL inspection criteria from the engineer who is no longer here. Our testing is simple "Go/No Go" we have specs for temps and weight and the bags we pull every hour either pass or they fail.
We typically have production lots in the 35,001 to 150,000, S4, .65 so we are pulling a total of 80 samples throughout the job ( a job may run in one day or over multiple days).
In addition, a production asssoc is pulling 1 bag per hour and performing the same test.
And even though the Std allows us 2 failures before we reject the lot, we stop the line if the temp is out on even 1 of the bags.
Does this sound like a sampling plan we should be using or is there a better one someone could suggest?

Perhaps you could clarify...better in what way? So the question would need to be clarified, do you want to ensure less defects/lot? Is the reliability of your current sampling plan in question?

The sampling plan you're using now aligns with ANSI/ASQC Z1.4-1993, the S4 is used when smaller sample sizes are necessary or preferable and large risks can be taken. Typically this is decided after 10 or more lots display acceptable defect % under a normal inspection/sample plan. As long as you, and your customers, are accepting the AQL of .65 and your lots have a history of being defect free under normal sampling you're good to go.

I would recommend picking up a decent book on sampling plans, typically they will outline accepted standards and make for a great reference...ASQ has a few which come to mind. ;)

I'll let a few covers more experienced suggest improvements, but from my side of the world the sampling numbers sound right.

Oh, and congratulations on becomming a Cover, it's a great resource which keeps you thinking and learning.

I forgot to mention the guy you have pulling a bag every hour, sounds like he's doing this just to give you peace of mind as this follows no plan and really is unnecessary if your primary sampling plan is within accepted standards. If it were me I'd have him doing something more productive. :tg:

OK.."better" is the wrong word...maybe I mean "more efficient".

With the AQl of .65 in place, how many potentially defective product are we allowing into the field?

I'm not sure if this will help you but the old MIL-STD-105E and MIL-STD-414 were cancelled and replaced by a consolidated standard in 1996. They are now MIL-STD-1916 can be found here assistdocs.com/search/search_fsc.cfm or downloaded from my attachment :D

I'm not sure if this will help you but the old MIL-STD-105E and MIL-STD-414 were cancelled and replaced by a consolidated standard in 1996. They are now MIL-STD-1916 can be found here assistdocs.com/search/search_fsc.cfm or downloaded from my attachment :D

MIL-STD-105E was replaced by ANSI / ASQ Z 1.4

Hmm yes, I was going by the ANSI/ASQ Z 1.4 ....not sure when the 105E was mentioned....

Anyhow, I apologize Boerne, I had lost track of this thread and hadn't seen your question regarding the AQL of .65

AQL in this case would be defined as "the maximum percent defective (or the maximum number of defects per hundred units) that, for purposes of sampling inspection, can be considered satisfactory as a process average" The idea is to have a low probability of accepting a lot which has more percent defective than the .65 .

As suggested, it would be a decent idea to pick up a good reference book, to have on hand for things such as this as it will give you a more thorough review of samping plan information and your options. More so if you don't deal with sampling plans on a frequent basis it is easy to forget.

The important thing when it comes to this type of sampling plan is to be sure your customer approves of the AQL and the plan, else you might end up with a very unhappy customer.

OK.."better" is the wrong word...maybe I mean "more efficient".

With the AQl of .65 in place, how many potentially defective product are we allowing into the field?

This is a tough question to answer...

The easy answer is that you will ACCEPT a lot that is .65% defective 95% of the time.

to answer your question, we would need to know how your line normally is and whether or not it is stable or if the defects 'come and go'.

AQL in this case would be defined as "the maximum percent defective (or the maximum number of defects per hundred units) that, for purposes of sampling inspection, can be considered satisfactory as a process average" The idea is to have a low probability of accepting a lot which has more percent defective than the .65 .


That is not quite the right interpretation. You have a high probability (around 90%-95% typically) of accepting a lot which has a percent defective of 0.65 or less.

Generally you need to hit a rate of about 5x ( depending on the specific plan) greater than AQL before you hit a high probability (i.e. 95%) of rejecting a bad lot.

Tim F

Yes Tim is right, the true way the plan is statistically determined is by ensuring you are accepting lots with .65 or less percent defective to a 90-95% probability level, which as an effect means you probably won't accept a lot which has more percent defective than your .65 aql. Sometimes I tend to look at things backwards at an angle through a mirror and around a corner...this year is one of those times :lol:

There really are so many numbers of ways to decide on a sampling plan though, my suggestion to do some research still stands, you may find that your current plan is sufficient, or you may find a suitably efficient way of ensuring better quality, in either case you'll likely know more.