Measuring Customer Complaints in PPM - How does your company define PPM?

[Icy Mountain]

In the door or out in the field

Let's say I ship 1M items per year to an automotive assembly line. Since I 100% test the electronics, I get 5 rejects from the line per year at plug in, turn on(5 PPM, yeah?)

So, the vehicles go out and get beat up by the end user. Over the two year warranty period, I get 5,000 warranty claims for replacement. (5,000 PPM, right?). But wait, during the 2 year period, I've fielded 2M parts, so 2,500PPM, ok? Maybe not.

I tend to view PPM as an assembly line reject measure. Fielded returns (warranty) fall into a reliability calculation for me, mean time to failure or mean time between failures, failure rate, etc. Obviously, line side rejects will be much lower than defective product in the field after 2 years of wear and tear.

Where do I measure PPM?

 

[Jim Wynne]

Since this thread was resurrected and I passed on it the first time around, I'll take the opportunity to make a blanket complaint about the concept of PPM in general.

It seems to be that unless you're actually making a million of something, the method flies in the face of good statistical sense. Let's say that we have reliable data to indicate that a given process will likely yield a PPM value of 50, and that number is acceptable to everyone involved. Now let's say that in my first shipment of 1000 parts there are 20 defectives. Given those numbers, the customer might complain that our PPM level is an unacceptable 20,000. But I say that I've verified that my process is in control (i.e., it's statistically stable) and that the idea of randomness in probability tells us that if the expected value is 50 defectives in 1mm parts, we have no basis for assuming that the 50 defectives--if in fact that prediction comes true--will be evenly distributed over the chronology of the manufacture of 1mm parts. We could, at least in theory (and that's what PPM is all about) have 20 defectives in our first 1000 and then not have another one in the ensuing 1,999,000 pieces, which would mean that what the customer has defined as a 20,000 PPM level is really 20 in practical terms.

 

[Morcs74]

I personnally the most effective way to measure PPMs internally and externally is in the form of DPMO (defects per million opportunities) which then can be used to calculate the process sigma.

For example, internally a DPMO will effectively show you how good your process are:

Take an injection Moulding that goes through an assembly line that has say, 10 value added process being put into it, If you make a Million parts and 100 were scrapped off at the first step then the DPMO rating would be 10,000. On the other hand if you had ten parts out of a million being scrapped off at the 10th step then the DPMO rating would be 100,000 Hence highlighting where you should focus.

 

[Jim Wynne]

Take an injection Moulding that goes through an assembly line that has say, 10 value added process being put into it, If you make a Million parts and 100 were scrapped off at the first step then the DPMO rating would be 10,000. On the other hand if you had ten parts out of a million being scrapped off at the 10th step then the DPMO rating would be 100,000 Hence highlighting where you should focus.

I admit that I'm not a six-sigma expert (my pants keep falling down because I have no belt) but this doesn't sound right to me. If there were 100 scrapped in the first operation, the DPMO (or also PPM, in this case would be 10,000, but only for that operation, no? By the same token, how does the DPMO grow to 100,000 when there have been nine prior operations with no defects? What am I missing?

 

[Bill Pflanz]

The problem with DPMO is that it tells you nothing about your process. As the examples show, it is a point in time calculation of the defects that can be determined a number of ways depending on how you define an "opportunity". Even if you settled on a definition, the defects could be part of the process as common cause requiring corrective action or it could be from special causes that may require no action.

It would be better if the Six Sigma black belts returned to the basics of statistical analysis (including control charts) of the process data rather than calculating DPMO.

Just my opinion.

Bill Pflanz
(ASQ CSSBB)

 

[Morcs74]

One way of saying how DPMO Works:

What would you rather have, Parts scrapped off at the first stage of the process, or Parts scrapped off after they have had say 9 value added processes go into it which would mean rework or even disposal.

 

[jawatts - 2009]

Ppm

Are you using your PPM for anything else than a comparative reference to what your customer reports your PPM as?
Do it the same as your customer.

 

[Tybee - 2011]

Customer Complaints

I was asked by the staff to define what a customer complaint would be as recordable.

My initial responce was anything that would cause additional cost to the customer.

Any comments on this topic?

 

[vanputten]

What type of "Parts?"

PPM stands for Parts Per Million. What kind of parts? Good? Bad? With one defect? With multiple defects? DPPM before inspection? After inspection? Before test? After test? Internal? At one customer? At all customers?

PPM only defines a ratio of how many occurrences there have been if there were a million opportunities. But occurences of what?

I think most on this thread are speaking of DPPM. D = a part with at least one defect. PPM by itself is simply a ratio. We need to define what it is that we want to count per million.

If people believe there is only one way to count "PPM", and that everyone counts it the same, you are kidding yourself.

Defective parts per million? Defects per million opportunities? Defects and defective parts are different things. And at what stage in the process?

Where one measures this can make a huge difference. If someone is trying to convince you of there product performance with an internal "PPM" measure, we need to know from where in the process they have taken the measure. For example, a DPPM measure might be dramtically different before or after a 100% electrical test for electical parts.

When discussing particles suspended in a fluid, we could measure those "parts" in different ways - size, mass, density, weight, or any other characteristic.

What is it that is being counted and where in the process?

PPM by itself only indicates a ratio. It doesn't tell you anything more than that. What per a million, where, how measured, how was the data collected?

Regards,

Dirk

 

[Helmut Jilling]

I was asked by the staff to define what a customer complaint would be as recordable.

My initial responce was anything that would cause additional cost to the customer.

Any comments on this topic?

I would think that if the customer takes time to complain, it should be a "recordable event." You might have categories to address those compaints that end up being inappropriate. But if they took time to complain, you should record it as a starting point.