DPM (Defects per Million) vs. DPMO (Defects per Million Opportunities)

[optomist1]

Hi To Everyone,

A generic question for the agile minds of Elsmar.....if you have a part say with 10 or 11 different dimensions/part, and whether one dimension or six dimensions are beyond the specification say +/- 0.01mm, the part is unacceptable.

Would you use DPM or DPMO? Seems like an obvious question yet in my experience; you use DPMO when you have a part or item that has a number of different attributes/part such a nicks, cracks, or surface not parallel etc. Or are different dimensions on the same part treated as different measures?

The DPM vs DPMO selection would certainly make a difference in assessing current performance.

Any insight would be welcome...

Thank You...

Marty

 

[Bev D]

Re: DPM vs DPMO

In *my* opinion, DPM vs DPMO is a false issue that was brought up by the 'silly' part of Six Sigma.

The real question is: are you concerned with defects (Poisson) or defectives (Binomial) and what question are you trying to answer to with the data...

Use of the Poisson for defects obviates the need for accounting for the number of opportunities which is usually a non-value add task.

The critical understanding is that if you have a high enough defect rate, you can improve the defect rate without substantially improving the defective rate. SO, the choice is situational, based on the question and what action you are concerned with.

DISCLAIMER: I don't typically use ppm (or is dpm/dpmo cousins) unless I actually have millions of things AND the defect rate is so low that the number of leading zeros in the proportion is difficult to track. I also eschew percentages...

 

[db]

Re: DPM vs DPMO

I have a client that has one part with 32 different machining operations. So, each part would have 32 opportunities. However, their customer sees it as if one dimension is out of spec, the entire part would be rejected, so they measure as DPM, not DPMO. They don't really care about the opportunities, they care only about the part.

Taking that to an extreme. I have a part with 1 million dimensions. If one is off, and I send the part to the customer, and it gets rejected, they will say I have 1,000,000 DPM (since the only part I sent was defective), yet, my internals would show 1 DPMO. Quite a difference, but the only one that matters is what my customer grades me on.

 

[optomist1]

Re: DPM vs DPMO

Hi Bev,

Poisson or Binomial it depends on the situation....which makes this foundation of analysis question all the more important. Thank you for the insight, as one's choice strongly influences the analysis.

Regards,
Marty

 

[optomist1]

Hi db,

Your insight better states my point or concern; the metric changes depending on your perspective.

With this one crucial selection, you can drastically change the process performance measure. Sometimes it is a question of "how real" you want to be.

Thanks for your input...

Marty