Calculating PPM - Standard practice to calculate PPM from parts returned?

[Dawn]

PPM


Is it standard practice to claculate PPM from parts returned as a total returned or can they be tracked by how many were nonconforming rather than the total sent back? If they were sorted and good parts returned to the customer can we only count the nonconforming number?

 

[Kevin Mader]

Dawn,

It depends on your situation, but my advice, stick with the hard numbers. Soft numbers can be misleading.

Nonconforming, a strange label I think. Used often describe materials of undetermined status (since they can't be considered conforming until proven some how). I think that this often creates confusion with some folks.

For me, report the numbers as they are proven or known. So materials returned as nonconforming by the customer that is later determined to be fine (i.e. the customer didn't want it, so they said it was defective) do not count as part of the numerator in your equation. However, the example you gave (returned, sorted, good parts returned) has value as well. Here you may want to report both values (customer perception, actual off quality). Remember: customer perception often goes further than actual figures (the 'unknowables' such as the cost of ill will). And regardless of status, all parts were returned at cost, which is a nuetral factor in the PPM world, but figures plenty in the Cost of Quality.

Just some thoughts. So back to the group...

 

[Don Winton]

Dawn,

I would suggest the PPM be calculated from the actual nonconforming product, rather than the number returned. This avoids the soft number Kevin refers to.

I would track the total returns as another category (perhaps 100% inspected). Kevin brings up other good points to which I have nothing to add.

Nonconforming, a strange label I think.

Yea, I chuckle as well. I believe (I may be wrong) the term stems from the Tort issues that have plagued companies. The implications of the term 'defect' are VERY far reaching in the legal community. ISO even put something to this effect in the Introduction portion of ISO 8402. I advise anyone who will listen not to use 'defect' in any documented record of any kind. I like your term 'off quality' as an acceptable substitute for nonconforming.

As an off thought, I like to separate 'nonconforming' from 'noncompliance.' I use nonconforming to apply to off-quality product and apply noncompliance to a breakdown in the Quality Management System (i.e. a noncompliance was issued against ISO 9002, Section 4.17 for...). When reviewing records, I can tell at a glance what applies to product and what applies to the system. Some companies I have visited use nonconformance to describe everything: product, system, processes, etc. This makes for a records mess, IMHO. Perhaps I am being too complicated? Perhaps not?

Regards,

Don

------------------
Just the ramblings of an Old Wizard Warrior.

 

[John C]

Don,
Consciously or not, you are following ISO 9000 to the letter. I think I have used 'conforming' and 'compliant' indidcriminately in the past, and maybe threw in a lot of 'defects' to boot. I'll take your advice about not writing down 'defect' in reports and minutes in future. I'll also follow your useage of conform and comply;
Interested in the point you made, and anxious to check out whether my hallowed ISO 9002 document (supported by me pocket Collins Dictionary) stood up to appraisal, I checked it out and found about 20 uses of conformance/nonconformance/conformity/nonconformity all specifically referring to product. I found two cases only of compliance; "compliance with reference standards, plans, documented processes" in 4.9 (Process control) and "comply with planned arrangements" in 4.17 (Int Audit).
Unfortunately, I found one exception; In 4.16 (Q. records) it says; "...demonstrate conformance to specified requirements and the effective operation of the quality system".
I think they slipped up there, and should have complied with or conformed to your method, as I will do in future. Thanks for the advice.
and rgds,
John C

Collins dictionary;
Comply - consent, yield, do as asked
Conform - comply with accepted standards, convention, rule, pattern, custom.
(which seems to indicate that 'comply' is a response to a request or order, while 'conform' is a subset of comply, relating only to some type of universal definition - JC)

 

[Don Winton]

Consciously or not, you are following ISO 9000 to the letter.

Hmmm, I guess my Lead Assessor brainwashing, er, training stuck better than I thought.

Regards,
Don

------------------
Just the ramblings of an Old Wizard Warrior.

[This message has been edited by Don Winton (edited 12 May 1999).]

 

[John C]

Dawn,
In a given lot of returned material, the relationship between the number of items in that returned lot and the number of nonconforming items found in it by 100% inspection (or sample of the same) tells us nothing, and we cannot calculate any valid ppm from it.
With extra, reliable information - for example, that the returned lot is a random sample of your delivered material - then you can calculate the ppm from the number of nonconformities per the number you inspect. But, if the customer has sent back a portion of your delivery, mixed good and not good, it is much more likely that they have been isolated from the bulk of the material by some sampling plan which ensures that the lot returned will not be a random sample.
To find the actual (and useful) ppm number, you would need to gain the customer's cooperation and be sent valid information as to what the returned material actually represents.
PPM can be misleading and dangerous data if not used carefully. For example, a case I observed where ppm was applied on a daily basis where deliveries were made every day of maybe 200 items. Target is 1000 ppm; Day one zero n/cs from sample of 20 from 200, so zero ppm - good. Day two, the same. Days three, four, five, etc, the same - still good, zero ppm. But day 65 one n/c from sample of 20, which is 50,000 dpm! 50 times greater than the target ppm. PANIC! DISASTER! Find a new supplier! A valid calculation would be 769 ppm, but try expaining that to the 'Find someone to go over and kick that supplier's ass' brigade.
rgds, John C

 

[Kevin Mader]

JC,

Good points on arbitrarily selected targets, they may be huge pitfalls. Good advice on getting together with the supplier. After all, it is a partnership you are after.

Management by intimidation wreeks disaster. Just awful. To the 'ass kicker brigade', be gone!

Dawn, if you have the past data, you may be able to develop a picture of where you have been and where you are today. They may indicate where you are likely to be in the future and help to set a goal (mutual as John suggested) that you and the supplier will work towards. Reach that, revise again. Continuous Improvement!

Regards,

Kevin

 

[George Trybulski]

Dawn

I whole heartedly believe in counting just the parts that were found "nonconformant" from the customer. One customer of ours does not use "FIFO" and they took a sample of 10 out of our last lot sent of 100 and found 3 nonconformant parts. They then panicked and sent us back the 100 parts along with the last 4 unopened lots of 100 each ( 500 in all). We sorted the 500 and found only 7 of the 500 nonconformant. Should we use the 500 or the 7 for our dppm ? Our dppm is calculated on a monthly basis by parts found nonconformant against the number of items shipped for that month. We then take the dppm totals from each month and calculate a three month rolling average to track whether there is improvement or not. Works for us and our registrar. Happy number crunching !

 

[George Trybulski]

Dawn

I whole heartedly believe in counting just the parts that were found "nonconformant" from the customer. One customer of ours does not use "FIFO" and they took a sample of 10 out of our last lot sent of 100 and found 3 nonconformant parts. They then panicked and sent us back the 100 parts along with the last 4 unopened lots of 100 each ( 500 in all). We sorted the 500 and found only 7 of the 500 nonconformant. Should we use the 500 or the 7 for our dppm ? Our dppm is calculated on a monthly basis by parts found nonconformant against the number of items shipped for that month. We then take the dppm totals from each month and calculate a three month rolling average to track whether there is improvement or not. Works for us and our registrar. Happy number crunching !

 

[John C]

George,

What are the odds against 3 out of 7 n/cs finding their way into a random sample of 10? We could use the binomial theorem (we could if we could remember how) to work it out, but I don't think we should bother. The most likey explanation is that someone has blundered. In fact, two people have blundered; Your material control has broken down and someone has put three rejects into a good box. In such a situation, Murphy's law directs that their inspector (who has rationalised random inspection and come to the conclusion that, if they went in at random, then they'll come out at random) will pick one box for his sample and find the only three n/cs in the lot.
It might not have been quite that way, but something equally stupid, no doubt. Check on your handling and ask them to review their sampling.
Also, if my explanation were true, it would be more reasonable to assume, than not to assume, that this sort of mix-up happens quite often but only gets found in one case out of ten, on average.
Also; Note that this is an extreme case, where the data is so odd that it arouses suspicion. Misleading data is probably more common than good data, but we go on pumping it out and making decisions based on it.
Better to look at the process. Don't measure mistakes, stop making them.
rgds, John C