The use of actual test data or a Pass/Fail sample, that is the question. What is the right answer?

Dan Watson

Involved In Discussions
Hello,

My fastener company is registered to the ISO 9001 and IATF 16949 standards. We outsource our part hardening since we do not have this capability. The part hardness specification is typically 35 to 42 HRC, which is on the customer part drawing as well as the internal production drawing.

The quality inspectors take five samples per 50,000 pieces at incoming inspection. The dilemma that the inspectors face is that if one or more samples fail they take this information to the plant manager. The plant manager usually gets our internal heat treatment (annealing is our only capability) supervisor involved as well. Either the plant manager or heat treatment supervisor gives a verbal approval to release the parts for further processing.

Instead of wasting the quality inspectors' time, I said to stop our receiving inspection testing and use the certificate of analysis from our outsource supplier. May reasoning is that the outsourced suppliers are on pour "approved vendor list" and that the supplier can be accepted as a pass or fail criteria.

Our Product Engineer and Heat Treatment Supervisor countered with stating that only one sample needs to be tested as a "Pass/Fail" sample and to use the supplier COA for the actual data. As a side note, our customer does have a sampling plan calling for 5 samples for 50,000 pieces. If the first sample fails, the pull another until one does pass to use as the "Pass/Fail" sample.

Am I doing the sampling wrong? What would others do in this situation? Is a single "Pass/Fail" sample appropriate?
 

Bev D

Heretical Statistician
Leader
Super Moderator
OK this just sounds like “test to pass”. Think about it. This is not a sample pan. This is delusional plan. No amount of failures will deter you or your supplier from continuing to test until you get a pass….what does this tell you about the quality of the lot? Nothing. How will you know if it’s a good lot or a bad lot? Wishful testing is just a complete waste for everyone.

You are better off figuring out how to properly test the lot with a real sample plan with a real acc/rej number. Oh and have you and your supplier performed an MSA on your testers? Hardness is notoriously difficult to test due to local variations. What is the real specification that you need? If all of your past lots have not contributed to any defect or failure rate in your product then maybe you just need to increase your specification limits - but do you have any actual data? Or do you only have the data you and your supplier ‘liked’?

Sorry for the harsher critique but someone needs to wake up and face the physics instead of bowing down to the psychics…
 

Dan Watson

Involved In Discussions
OK this just sounds like “test to pass”. Think about it. This is not a sample pan. This is delusional plan. No amount of failures will deter you or your supplier from continuing to test until you get a pass….what does this tell you about the quality of the lot? Nothing. How will you know if it’s a good lot or a bad lot? Wishful testing is just a complete waste for everyone.

You are better off figuring out how to properly test the lot with a real sample plan with a real acc/rej number. Oh and have you and your supplier performed an MSA on your testers? Hardness is notoriously difficult to test due to local variations. What is the real specification that you need? If all of your past lots have not contributed to any defect or failure rate in your product then maybe you just need to increase your specification limits - but do you have any actual data? Or do you only have the data you and your supplier ‘liked’?

Sorry for the harsher critique but someone needs to wake up and face the physics instead of bowing down to the psychics…
Thank you. I thought I was losing my sanity. Thank you for being a voice of reason. I do appreciate your candor and honesty.
 

Dan Watson

Involved In Discussions
To follow-up, I have done an MSA on our hardness testers and gotten marginal results. As you, the Hardness is notoriously difficult to test due to local variations. We have never done an analysis with our suppliers since that type of study is not supported. The automotive customer has published hardness specifications on their drawing (35 to 42 HRC). They also test and historically we have a very low failure rate, so I guess one can "assume" that the parts are uniform. We have been collecting the data from the receiving inspections for the past seven years (was not collected prior to my arrival). This company really does not support data collection and testing. They have the "one sample bliss" attitude.
 

Randy

Super Moderator
Pass/Fail inspections? There's no guarantee that a lightbulb will come on a 2nd time and no guarantee the fasteners you didn't test won't fail.
 

Dan Watson

Involved In Discussions
I had revised the SOP for hardness testing. I submitted the revised SOP and here is the response from the head engineer:

Reason for Change - New Work Instruction for Receiving Inspection – Hardness Testing; Sampling; Actions



Change Order information:

---------------------------

DOCUMENT number: xxx

Old Rev: A

New Rev: B

Description: Work Instruction for Receiving Inspection – Hardness Testing Training Required: NO

Comments by engineer:

I do not agree with recording values. I do agree with a check to verify the parts have been hardened. Given the variation in a batch there is always the possibility of a few parts being a point or two above or below the spec. That is why I suggested rely
 

Ed Panek

QA RA Small Med Dev Company
Leader
Super Moderator
A good test in "What should I do here?" is what would you do if the biggest customer for that part were in the conversation? Once you start studying to the test it becomes a game.

The problem with gaming the system is that it will catch you and now you are in a web of lies. Customers are not stupid and there are other suppliers.
 
Top Bottom