# Variable AQL In a C=0 Sampling Plan

#### tbb001

##### Registered
I am working on developing a Total quality management plan for a currently unregulated industry dealing with consumable products. Due to the nature of this product we have 3 different risks, A. Manufacturing Contamination, B. Shipping Contamination and C. Cosmetic and Packaging/Product Labeling Errors. Due to the nature of this product all sampling for Risk A is destructive, while the risk of contamination is very low due to the controlled inputs from this supplier. I am looking for feed back regarding this testing philosophy. The below example is for Risk A. Risk A has a large chance to affect downstream customer acceptance but failure would not result in a immediate medical emergency for downstream customers. This is comparable to heavy metal tolerances in food products for example.
General Question:
When using a C=0 sampling plan, @ target AQL, would you follow the normal sample plan but instead of utilizing the tables acceptance/rejection criteria to just use accept on failure at that sample? For example, for a sample letter G @ AQL 1.5 for single normal inspection the table would return a sample size of 32, with AC/RE of 1/2 respectively. Would we sample 32 and just use a AC/RE of 0/1?
The Plan is outlined below:
Inspection Level: General 1 for Existing Suppliers, General 2 for New Suppliers
Table 2A (Z1.4) is used for each where C=0 with the switching rules affecting the AQL used to determine sample size.
AQL for this RISK A is set as .65 for tightened, 1 for normal, and 1.5 for reduced. Where this would be changed based on the switching rules for that particular supplier.
For Example, @ Sample Letter G the sample sizes would be as follows;
.65 = 32, 1 = 32, 1.5 = 20
Given that these case samples are then sent to a lab as a mixed sample as it would not be possible to test each case individually due to cost.

Risk B and Risk C are non-destructive testing and will follow normal inspection plans without switching rules as both can be tested when samples are pulled for Risk A.

#### japayson

##### Involved In Discussions
If you want to use a C=0 plan you should obtain a copy of the booklet by Squeglia (from ASQ). Also understand that if you want to go that route you should prove it out by a series of lots inspected and show that your supplier quality is so good that you can safely use the C=0 plan. That's my take on it.

#### Bev D

##### Heretical Statistician
Super Moderator
can you clarify what you mean by: “Given that these case samples are then sent to a lab as a mixed sample as it would not be possible to test each case individually due to cost.” I interpret this to mean that you will combine all samples somehow into a homogenous group and then the test lab will extract 1 small subset of the super sample to test?

#### Bev D

##### Heretical Statistician
Super Moderator
A few things to consider:
remember that the AQL is teh acceptable quality level. 95% of the time the AQL defect rate will be accepted and higher defect rates have a slowly decreasing probability of acceptance. Most people really want a RQL - Rejectable quality level (aka the LTPD - lot tolerance percent defective - in a standard table).

The chosen AQL (Or better, the RQL) is always chosen based on the severity of the defect.

Inspection is a bandaid. Process improvements, mistake proofing, and process control monitoring such as SPC is always better. And is much more ‘modern’ than inspection. In fact the Japanese founders of the so called Total Quality Management approach was in response to gettign away from inspection and relying instead on reducing variation thru experimentation and controlling that reduced variation through SPC…

#### tbb001

##### Registered
If you want to use a C=0 plan you should obtain a copy of the booklet by Squeglia (from ASQ). Also understand that if you want to go that route you should prove it out by a series of lots inspected and show that your supplier quality is so good that you can safely use the C=0 plan. That's my take on it.
Awesome, I will check this out. Can't believe how hard it was to find this as a resource.

#### tbb001

##### Registered
can you clarify what you mean by: “Given that these case samples are then sent to a lab as a mixed sample as it would not be possible to test each case individually due to cost.” I interpret this to mean that you will combine all samples somehow into a homogenous group and then the test lab will extract 1 small subset of the super sample to test?
Yes, the lab is taking the sample sent and homogenizing the products to then extract for testing.

#### Mike S.

##### Happy to be Alive
Trusted Information Resource
Awesome, I will check this out. Can't believe how hard it was to find this as a resource.
Not if you do a search on this forum, or google, for c=0 sampling plans!

Just don't use ANSI Z1.4 / MIL-STD-105 plans and make C=0.