Hi There,

One of our customers has set their acceptable DPPM level at 5K, for all their suppliers.... across the board.

The problem is we send them about 20 different p/n's totaling 130 pcs per month. The parts are extremely high tolerance machinings and inevitably one (or more) will be found out of tolerance (by .0001") in a month’s time.

I've been attempting to persuade their Global Quality Director that this DPPM requirement does not fit the highly complex low volume product type we manufacture for them. Not going to far with this.

Is there a standard minimum qty that must be measured before DPPM can be accurately calculated? Something I can throw back that might assist me in my effort?

Thanks - Brookway

A difficult ones.

There is no minimum as far as I know :frust:but....

The chi square tables to determine "normality" (Gausian distribution) goes as far as 60 (more or less), so it could be a good point to start (60 minimum)

The traditional control charting take in account 20 samples of 5 data points, it means 100 data points to determine SPC control limits.

Don Wheeler said that it's better to obtain the % of defective by counting insteed of using fourmulae and statistic indicators, b'cause the asumptions on the used statistics. As you said it spected to be a % defective more than a DPPM if you find 1 or more in 130 samples

As I put in another post about cp/cpk, the index has it's uncertainly so an index cpk=1.33 could be between 0.7 and 1.9 (to say something without any math) depending on the sample size. So a % defective it self has it's own uncertainly.:nopity:

Hi There,

One of our customers has set their acceptable DPPM level at 5K, for all their suppliers.... across the board.

The problem is we send them about 20 different p/n's totaling 130 pcs per month. The parts are extremely high tolerance machinings and inevitably one (or more) will be found out of tolerance (by .0001") in a month’s time.

I've been attempting to persuade their Global Quality Director that this DPPM requirement does not fit the highly complex low volume product type we manufacture for them. Not going to far with this.

Is there a standard minimum qty that must be measured before DPPM can be accurately calculated? Something I can throw back that might assist me in my effort?

Thanks - Brookway


So, if I understand correctly, you ship them about 1560 parts per year. Since they have a 5K DPPM requirement, it would allow for approximately 8 defective parts per year. One defective in 130 means a rate of 7700 DPPM. There is no minimum quantity to "estimate" DPPM. The better question is not the estimate, but the variability in the estimate. Here is a table that shows regardless of aggregation by month or year gives the same estimate, but you would feel more confident with a years data over a months data

Defects/Month Monthly Estimate (Based on 130 units per month)
0 0
1 7692
2 15385
3 23077

Defects/Year Yearly Estimate (Based on 1500 units per year)
12 8000
18 12000
24 16000
30 20000
36 24000

In addition to Steven's good advice, the dilemma you're facing is that PPM makes no sense as a performance measure when small numbers are involved. Good luck trying to get your customer to understand this. If we follow the PPM concept logically, we have to accept that the results from a very small sample will hold true after a million parts have been produced, and without knowing a lot more about the process, there's just no good reason to make that leap.

Thank you for your replies. Yes, it will be a battle changing the minds of others when the outcome to them is more work. I had a similiar challenge with a different customer where we started counting features instead of actual parts, this worked out ok. I'll put together a speadsheet that calcualtes this and see if I can sell it.

We'll keep pluggin away, thanks again.

Thank you for your replies. Yes, it will be a battle changing the minds of others when the outcome to them is more work. I had a similiar challenge with a different customer where we started counting features instead of actual parts, this worked out ok. I'll put together a speadsheet that calcualtes this and see if I can sell it.

We'll keep pluggin away, thanks again.

As a joke, you could tell them if you improve your process such that you can acheive better than 5K DPPM, you would be happy to ship them some defects to meet the target. I would highly recommend you tell them that due to the small sample size, any estimate of DPPM would be unreliable. Just to add to my comments from before, if 5K DPPM allows for approximately 8 defects per year of production, that if you reach 7 defects you would need to stop shipping because another defect would but the process out of their acceptable region. Another point to be made, DPPM is an estimate for a stable process. You did not say if all the defects came from a single p/n or randomly from all 20 p/n you ship. Are the processes similar enough to pool for DPPM calculations? Just some thoughts.

As a joke, you could tell them if you improve your process such that you can acheive better than 5K DPPM, you would be happy to ship them some defects to meet the target.

Excellent. :agree1: There was a recent discussion (http://elsmar.com/Forums/showthread.php?t=22803) here regarding PPM that further demonstrates that many customers don't understand what they ask for. Someone had a customer who was asking for CA in a situation where the supplier's PPM level was <1.5 and and there were millions of parts involved. Although it wasn't stated by the person raising the question, my guess is that the customer's required PPM level was something considerably greater than 1.5. Regardless of the PPM level, the customer can always reserve the right to debit for NC product, but my guess is that the most economical and sensible course of action would be to throw out the one-in-a-million and get back to work.

If a customer has a requirement for x PPM, and the supplier's performance is some number less than x, and at the same time has a CA system that requires action for any defective received, it's clear that the customer doesn't understand what's being asked for.

Yes, they are from different parts here or there. We have had a couple months with no defects, then a bad month with (4). The parts are complicated aluminum castings that are machined, chromated, helicoils installed, then masked and painted.

Hi There,

One of our customers has set their acceptable DPPM level at 5K, for all their suppliers.... across the board.

The problem is we send them about 20 different p/n's totaling 130 pcs per month. The parts are extremely high tolerance machinings and inevitably one (or more) will be found out of tolerance (by .0001") in a month’s time.

I've been attempting to persuade their Global Quality Director that this DPPM requirement does not fit the highly complex low volume product type we manufacture for them. Not going to far with this.

Is there a standard minimum qty that must be measured before DPPM can be accurately calculated? Something I can throw back that might assist me in my effort?

Thanks - Brookway
Brookway,

If I understand your issue, you have 20 different parts, you make a total of 130 parts in a month, accordingly, you only make 6 – 7 of each part per month.

First of all, I question the validity of adding all the parts together for your DPM calculation unless the parts are identical or nearly identical.

Applying a thumbrule very loosely, to accurately determine a failure rate the sample should be large enough such that 5 or more defects would be detected in the sample.

For your case, if using all the parts as a single sample is valid, 5000 DPM equates to 5 failures in 1000 parts (8 months of production).

If the parts are different enough that combining them into single sampling group is not valid, your required number gets even worse.

My opinion is the only way to satisfy your customer’s requirement is to perform 100% inspection and rework/scrap any failed parts and replace them with good parts.

I would not go to the customer with a story that effectively says “I cannot prove I can meet your requirements so I request you change your requirements to ones I can meet”. If I was your customer, I would start the search for somebody else that can make the parts.

Sincere regards,

Jim Shelor

Hi There,

One of our customers has set their acceptable DPPM level at 5K, for all their suppliers.... across the board.

The problem is we send them about 20 different p/n's totaling 130 pcs per month. The parts are extremely high tolerance machinings and inevitably one (or more) will be found out of tolerance (by .0001") in a month’s time.

I've been attempting to persuade their Global Quality Director that this DPPM requirement does not fit the highly complex low volume product type we manufacture for them. Not going to far with this.

Is there a standard minimum qty that must be measured before DPPM can be accurately calculated? Something I can throw back that might assist me in my effort?

Thanks - Brookway

Hello Brookway,

Your situation is a classic example of customers imposing unrealistic requirements that they have not (even statistically) evaluated themselves.

There are examples here in The Cove of suppliers with exemplary quality and delivery performance that were required to do a full blast root cause analysis on a one-time nonconformity where the root cause was very obvious.

I believe that it's OK to (professionally) push back a little. If the suppliers don't, that wil set a precedent and the situation perpetuates itself.

It's a matter of customer education. And the time to do this is during review of customer requirements (see 7.2.2). Amazing what this, and an open dialogue with the customer can do. You may not always "win", but it's always worth a try.

Stijloor.

I would not go to the customer with a story that effectively says “I cannot prove I can meet your requirements so I request you change your requirements to ones I can meet”. If I was your customer, I would start the search for somebody else that can make the parts.

Sincere regards,

Jim Shelor

There is nothing wrong with (diplomatically) saying to the customer, "I can't meet your requirements because the requirements don't make sense." Illogical requirements proliferate, and untold amounts of time and money are wasted, when we take the position that the customer is never wrong.

There is nothing wrong with (diplomatically) saying to the customer, "I can't meet your requirements because the requirements don't make sense." Illogical requirements proliferate, and untold amounts of time and money are wasted, when we take the position that the customer is never wrong.



Right on Jim! Amen to that. :applause:

Stijloor

There is nothing wrong with (diplomatically) saying to the customer, "I can't meet your requirements because the requirements don't make sense." Illogical requirements proliferate, and untold amounts of time and money are wasted, when we take the position that the customer is never wrong.
On the other hand, if Brookway is conducting 100% inspection and replacing the defective parts with good ones before shipment, Brookway's DPM is 0.

There is an internal DPM to deal with because it is reducing the profit margin, but the external rate is 0.

Brookway, what is your current sampling plan?

Jim Shelor

On the other hand, if Brookway is conducting 100% inspection and replacing the defective parts with good ones before shipment, Brookway's DPM is 0.

There is an internal DPM to deal with because it is reducing the profit margin, but the external rate is 0.

Brookway, what is your current sampling plan?

Jim Shelor

Do you think that scrap rates (or predictable rework costs) are not built into piece-part pricing? If PPM (or DPM, or whatever) is to make any sense at all, it has to be applied to the process that produces the parts. If the as-produced defect rate is significantly different from the as-delivered rate, no problem has been solved by satisfying the customer's misbegotten requirement.

Do you think that scrap rates (or predictable rework costs) are not built into piece-part pricing? If PPM (or DPM, or whatever) is to make any sense at all, it has to be applied to the process that produces the parts. If the as-produced defect rate is significantly different from the as-delivered rate, no problem has been solved by satisfying the customer's misbegotten requirement.
I would really like to know what the current sampling plan is.

Of course I know what goes into the pricing model, but that does not invalidate the statement that lowering the internal DPM improves the profit.

I think having an external DPM of 0 does solve a problem, meeting the customer requirement.

I do not believe the customers requirement is unreasonable or illogical. It can be met and should be. The pricing model had to include the cost of inspecting the parts before shipment. If the model did not include these costs or included costs for statistical rather than 100% sampling, then in my opinion the cost model was wrong. We should have known we would need 100% sampling to meet the requirement.

Best regards,

Jim Shelor

I would really like to know what the current sampling plan is.

Of course I know what goes into the pricing model, but that does not invalidate the statement that lowering the internal DPM improves the profit.

I think having an external DPM of 0 does solve a problem, meeting the customer requirement.

I do not believe the customers requirement is unreasonable or illogical. It can be met and should be. The pricing model had to include the cost of inspecting the parts before shipment. If the model did not include these costs or included costs for statistical rather than 100% sampling, then in my opinion the cost model was wrong. We should have known we would need 100% sampling to meet the requirement.

Best regards,

Jim Shelor
Why does the sampling plan matter? Assume he took a sample size of 10 units and found no defects, how does that help to establish a DPM? The customer finds a defect and reports it back to the supplier. On average they receive about 1 complaint a month for a defective part. Like said, some months are zero others are four (HERE IS THE PROBLEM!).

DPM is an average over the process....if he had 1 month with 6 defects and 11 months with none, the average is 0.5 defectives per month, but the 6 defects in one month is a special cause.

Why does the sampling plan matter? Assume he took a sample size of 10 units and found no defects, how does that help to establish a DPM? The customer finds a defect and reports it back to the supplier. On average they receive about 1 complaint a month for a defective part. Like said, some months are zero others are four (HERE IS THE PROBLEM!).

DPM is an average over the process....if he had 1 month with 6 defects and 11 months with none, the average is 0.5 defectives per month, but the 6 defects in one month is a special cause.
Statistical Seven,

The point I was trying to make is short of 100% inspection I do not believe there is a statistical sampling plan that will tell us the defect rate is <= 5000 DPM.

For instance, for your example of 10 randomly inspected and 0 defects found. Using a single-mean single-sided confidence test, we are 95% confident that no more than 27% of the population is defective (37 units assuming a population of 132 units).

If we do 100% sampling and find 1 unit defective of 132, that is 7500 DPM. However, we can either rework the unit or make a new unit and when we ship the lot it contains 0 defects.

Whether or not I would go back to the client and renegotiate the 5000 DPM requirement depends on when I knew about the 5000 DPM requirement. If I knew before the contract was signed, no way do I go back to the client and try to renegotiate the requirement. If on the other hand this is a new requirement after production has already started, then I go to the client to negotiate the increased cost for meeting the new requirement.

There is a lot more we need to know about the circumstances before rendering a final opinion on it.

Best regards,

Jim Shelor

Of course I know what goes into the pricing model, but that does not invalidate the statement that lowering the internal DPM improves the profit.

That's not what you said:

There is an internal DPM to deal with because it is reducing the profit margin, but the external rate is 0.


You said x reduces the profit, which is not the same thing as not-x improving the profit. It's possible for it to have no effect, if the scrap rate is built into the pricing (which was the point I was trying to make).


I think having an external DPM of 0 does solve a problem, meeting the customer requirement.
"Solving" one problem while failing to address its root cause (process capability) almost surely means that no solution has been found, unless you've found a way to make 100% inspection 100% effective.


I do not believe the customers requirement is unreasonable or illogical. It can be met and should be. The pricing model had to include the cost of inspecting the parts before shipment. If the model did not include these costs or included costs for statistical rather than 100% sampling, then in my opinion the cost model was wrong. We should have known we would need 100% sampling to meet the requirement.


"Be careful what you agree to" is certainly a valid point. This does sound like another good example of why contract review is important. Nonetheless, if it's found that what's been agreed to is impossible (economically or otherwise), what are the alternatives?

Jim,

Yes, if we find the requirement is impossible to meet, we have few alternatives. In fact, we have one. Go to the customer, with hat in hand, and negotiate a change that will be possible for us and still meet the customer's needs.

Regards

Jim Shelor

Jim,

Yes, if we find the requirement is impossible to meet, we have few alternatives. In fact, we have one. Go to the customer, with hat in hand, and negotiate a change that will be possible for us and still meet the customer's needs.

Regards

Jim Shelor

Exactly. :agree1:

Brookway,

1) What is your enternal rejection rate?

2) What is your customer doing with the parts they find out of spec? Are they are using them because they are out only .0001?

If you are finding a lot of parts out before shipping, you have manufacturing issues to solve. If your customer can use the parts they find out of spec, try negotiating a new spec that you can both live with and is representative of your process capabilities. If your process can't meet the required tolerances then initiate 100% inspection.

Shipping nonconforming product to a customer should never be acceptable and trying to negotiate that it is won't win you business. Your customer goal is 5000 ppm this year, it could be 2500 ppm next year - continually improvement at work.

Hi Brookway

I am new on this forum. I have similar problem but my qty produced per month is 3-5 million pcs and I have to check for dimensions (though not so critical as yours) and appearance. As for dimension, I could used 100% go/no gauge as it is tolerance is +/-0.5 only. For appearance, I used camera for 100% visual inspection but this is really dependable on the operator eyes condition ie fatigue, judgement etc. My customer is asking for 20 PPM because they are in automotive industry.

Back to your post ;), for small batch qty, is it possible for you to do 100% measurement instead of statistics to ensure that your parts are all acceptable to the customer ?