Short term vs Long term Capability - We ship millions of parts in a year

[beaser3]

I'm hoping someone can help me understand the definition between short term and long term capability. We have a customer that is grading us by the number of 8-D's that we receive from them. They issue an 8-D for a single defective part. We ship millions of parts in a year.

Although I agree that we should strive for zero defects I would like to show them what our dppm rate is in comparison to our 8-D number. The numerous 8-D count is showing that we are doing poorly but our DPPM is low (not zero....) and I know that we are overall a good supplier.

Here's one example: We have shipped 1191964 pieces since January. 110 have been defective. The DPPM is 92. The short term process sigma is 5.3 and long term process sigma is 3.8. Again, I would prefer the defective number to be zero and that is what we are striving so I am not trying to make excuses for our issues. We are to the point where we would need to spend millions of dollars to ensure zero defects and it does not seem realistic. So back to my question: since we are run our product in smaller batches would I use the short term process sigma? Would I use long term process sigma since the product is run in smaller batches but has run multiple times?

Is there a better way all together to show that the number of incidents do not necessarily mirror overall performance?

Any thoughts, ideas would be appreciated.

 

[optomist1]

How are the 8Ds categorized or are they? Are there similarities or repeat failures among the 8Ds or are they all different failures/failure modes? First step for me would be to stratify the failures as per 8Ds (or other QA data input), i.e. Pareto Chart, then address what is usually an 80/20 rule if this holds true for this case then a plan to address the 20% of causes....hope this helps.

 

[bobdoering]

How are the 8Ds categorized or are they? Are there similarities or repeat failures among the 8Ds or are they all different failures/failure modes? First step for me would be to stratify the failures as per 8Ds (or other QA data input), i.e. Pareto Chart, then address what is usually an 80/20 rule if this holds true for this case then a plan to address the 20% of causes....hope this helps.

That may be an internal analysis to sort through the problems, but the customer's metric is the what they are being graded to.

My customer does the same thing. They don't care about quantity, they care about occurrences - as that is what destroys their process flow. So, one can calculate dppm or dppb, but it is academic and does not relate to the customer's problem.

Changing your process flow to accommodate a dubious statistic - capability - is not a good idea.

 

[Bev D]

Very good point Bob
I am switching our supplier metrics from ppm to events as well. It the occurrence of an excursion that tanks us. It's not the defect rate.

Again we come back to homogeneity and more importantly the lack of it.
If all defects were random events ppm would be ok. But defects are too frequently not random. They cluster becaus they have a common assignable cause. This results in line stoppages and purges to capture all defects. It's like a mouse in Your house. Once you see one you know you have to go after the others then find the entry point and close it. If you calculated the mouse per day rate it would be less 1 per day.

 

[optomist1]

Bob, although I "agree to a degree" with the response.....for certain in the Automotive....errr I mean Mobility arena, the OEMs and likely the Tier 1s maintain and share with suppliers their respective "Scorecards", most are updated in pseudo-real-time, and they do in fact care about the quantity as well as type/classification of defects, and when the defects are noted during manufacture/assembly or during the critical 90 days or 3MIS.

The scorecard carries follow on business implications as well as warranty/cost implications.....this recent experience comes from years with an OEM as well as Tier 1s......my two cents

On the PPM vs Events front, in parts of the same field those are referred to as "tail of the scorpion", very low frequency but nonetheless "occurrences" and part of the drive to "zero". Your thoughts?

 

[Golfman25]

In our case, occurrences actually helps us because they are few. But because we are lower volume, our PPMs skyrocket. Don't you have to ship a million parts for PPM to be relevant? So we are stuck with an ugly PPM measure because we sent a mislabeled box of 100 pcs every few years or so.

 

[bobdoering]

Bob, although I "agree to a degree" with the response.....for certain in the Automotive....errr I mean Mobility arena, the OEMs and likely the Tier 1s maintain and share with suppliers their respective "Scorecards", most are updated in pseudo-real-time, and they do in fact care about the quantity as well as type/classification of defects, and when the defects are noted during manufacture/assembly or during the critical 90 days or 3MIS.

Just so you know, my customer is automotive. And, yes, number of defective parts as indicators of failed containment prior to shipping is also key - especially during the critical 90 days (although actually every day is critical.) That tends to be directed towards finding the detection and control inadequacies in the system that failed to protect the customer rather than ongoing performance. PPM, and now PPB, tend to be 10,000 foot indicators of overall health that are less interesting to them than the number of times you cause line disturbances - no matter what the quantity.

But, as we know, the correct answer to any question is "It Depends" - and if a customer still uses old school indicators, then you have to deal with them in their terms.

 

[Statistical Steven]

What is the question you want to answer? If you want to know the likelihood of a given lot having a defective unit then the short term capability is important. If you want to know what is the overall defective rate then the long term capability is the correct metric. Neither of the metrics help you with your customer issue. Think about Bev's point, defects are not random events. Capability metrics assume a random variable. For example, a lot with 10 defectives followed by 9 lots with 0 defectives gives an "average" defective rate of 1 per lot. Would your customer prefer 10 lots each with 1 defective? It's not your defective rate that matters, it is the number of non-events between events that is probably a more meaningful metric.

I'm hoping someone can help me understand the definition between short term and long term capability. We have a customer that is grading us by the number of 8-D's that we receive from them. They issue an 8-D for a single defective part. We ship millions of parts in a year.

Although I agree that we should strive for zero defects I would like to show them what our dppm rate is in comparison to our 8-D number. The numerous 8-D count is showing that we are doing poorly but our DPPM is low (not zero....) and I know that we are overall a good supplier.

Here's one example: We have shipped 1191964 pieces since January. 110 have been defective. The DPPM is 92. The short term process sigma is 5.3 and long term process sigma is 3.8. Again, I would prefer the defective number to be zero and that is what we are striving so I am not trying to make excuses for our issues. We are to the point where we would need to spend millions of dollars to ensure zero defects and it does not seem realistic. So back to my question: since we are run our product in smaller batches would I use the short term process sigma? Would I use long term process sigma since the product is run in smaller batches but has run multiple times?

Is there a better way all together to show that the number of incidents do not necessarily mirror overall performance?

Any thoughts, ideas would be appreciated.