Does C=0 strictly mean 1 bad, all bad, all the time?

[Mortalis]

As a contract medical device component manufacturer we have been using the C=0 sampling plan matrix for determining what the AQL will be at various stages of our manufacturing. We have been using this for over 20 years. We recently implemented a variation on that theme. Our Final Inspection department is our gate keeper to the Customer and our Final inspectors are very critical, which isn't a bad thing but, with all the Nonconforming Material Reports generated due to finding 1 NC pc in a 4.0 sample was stifling our Quality Engineering department and overburdening our MRB process. To ease this burden we implemented a 'find 1 pc in sample, take another sample and check for previous found NC. If no more found, scrap the 1 sample and move on. If 1 or more found for that same NC in the second sample, write up the NC' process. This has drastically reduced our time in MRB and has not had an appreciable effect on Customer Complaints.

I do understand that the processes of Nonconforming material and Material Review is there to help implement process improvements and what I have stated above tends to speak against this but we have not eliminated the NC and MRB all together.

Is this still considered C=0? When asked by our Customers during audits we proclaim to use the C=0 sampling plan. I guess, technically, we are using that sampling plan.

Your thoughts?

 

[somashekar]

Is this still considered C=0? When asked by our Customers during audits we proclaim to use the C=0 sampling plan. I guess, technically, we are using that sampling plan.

THIS thread and the attachment in one of the post can be helpful to you.
Also see the threads below in this page for similar threads....

 

[Jim Wynne]

As a contract medical device component manufacturer we have been using the C=0 sampling plan matrix for determining what the AQL will be at various stages of our manufacturing.

You can't use a sampling plan to determine the AQL--you determine the AQL in advance, based on risk, and then you use the chart to determine n, the number of samples to inspect.

We recently implemented a variation on that theme. Our Final Inspection department is our gate keeper to the Customer and our Final inspectors are very critical, which isn't a bad thing but, with all the Nonconforming Material Reports generated due to finding 1 NC pc in a 4.0 sample was stifling our Quality Engineering department and overburdening our MRB process.

If your sampling is creating a burden with NC parts, there's something wrong in production that sampling isn't going to help. If you repeatedly find defective parts, the sampling plan you've selected is doing its job, and should be causing you to implement corrective action to eliminate causes.

To ease this burden we implemented a 'find 1 pc in sample, take another sample and check for previous found NC. If no more found, scrap the 1 sample and move on. If 1 or more found for that same NC in the second sample, write up the NC' process. This has drastically reduced our time in MRB and has not had an appreciable effect on Customer Complaints.

Without knowing more about the product(s) and what's being inspected, but given your dependence on inspection, it sounds like just looking for the same defect in the second sample is looking for trouble. Do something to move away from inspection as much as possible rather than doing more of it when your sampling plan is successful.

I do understand that the processes of Nonconforming material and Material Review is there to help implement process improvements and what I have stated above tends to speak against this but we have not eliminated the NC and MRB all together.

There isn't necessarily a reason to do away with your MRB, but if you have an excessive amount of "Use as is" dispositions it almost always means that there's something wrong with the specifications.

Is this still considered C=0? When asked by our Customers during audits we proclaim to use the C=0 sampling plan. I guess, technically, we are using that sampling plan.

What if you were to tell your customers what actually happens?

In addition to the foregoing comments, it appears that you might be using acceptance sampling inappropriately to begin with, but again, without more information it's hard to tell. What about SPC?

 

[Mike S.]

Lots of good points in Jim's reply.

To directly answer your question, no you are not doing C=0 sampling when you approve lots where a defect was found. I think you know this. You are trying to have the best of both worlds -- lower sample sizes of C=0 plans and the C>0 acceptance numbers of conventional MIL-STD-105 type plans. It doesn't work that way.

Do you really think you keep finding that one defect in the lot because of bad luck?

 

[indubioush]

Is this still considered C=0? When asked by our Customers during audits we proclaim to use the C=0 sampling plan. I guess, technically, we are using that sampling plan.

No, you are not using that sampling plan. I agree with everything the previous posters stated, but I'll add in a little emotion. I don't know what type of component you are manufacturing or whether your customer is relying on your inspection data at all, but it is very possible that you are putting patients at risk through your actions. In addition, you are misleading your customers. If I found that one of my suppliers was doing this, it would be a very big deal!

Okay, but lets say you make only low risk components. You should be able to follow the sampling plans in ISO 2859-1 and ISO 3951-1. There are AQLs that do allow a certain amount of rejects. But you should ensure your customer is aware of this and is okay with it.

 

[Mike S.]

Seeing as it appears that the OP was using a 4% AQL or AAQL in the example he gave I would hope he is making low-risk products, as the probability of accepting a lot with defects is relatively high with that plan even when used correctly, which they are not.

 

[David-D]

I would defjnately agree that it is not a C=0 sampling plan. Two points which may not have been discussed:

The intent of C=0 plans is that there is no evidence of nonconformances in the lot; when there is, corrective action is to remedy the nonconformances in the product (100% screen, etc) and the process. The intent is to focus on the continuous improvement and not knowingly providing the customers defects. This is what underlies MIL-STD-1916, ISO 24247, and other c=0 (Accept on Zero (AoZ)) plans. By definition your aproach is diverging from this philosophy (even if it was satisfactory to all parties from a statistical perspective).

Another point is that you are in reality converting the origional sampling plan to a double samping plan with increased customer risk. It 3os unclear of you are just checking 1 more part or re-pulling the entire sample quantity, but either way, if you construct the OC curves for the sampling plans you can see how the customer risk increases statistically (not just philosophically as above).

To reemphasize another's point, if you are regularly failing a 4.0 AQL plan, then you would seem to have processes which are regularly creating nonconformances. The recomended focus should be on improving the process rather than making the sampling plan less sensitive.

Best regards,

David