Home › Forums › Old Forums › General › How to calculate DPMO and Sigma

This topic contains 29 replies, has 13 voices, and was last updated by Slimcane 9 years, 11 months ago.

Viewing 30 posts - 1 through 30 (of 30 total)

- AuthorPosts
- July 22, 2004 at 6:52 pm #64173
If I have 50 units failed among 1,000 production units, what is the DPMO(defect per Miillion opportunity)? How to calculate it? and what is the sigma level?

Thank you.

J.J.July 22, 2004 at 7:02 pm #64174J.J.,

How many opportunities for a defect per unit are there??July 22, 2004 at 7:10 pm #64175Probably it has 3 failure modes. So, I will say 3.

July 22, 2004 at 7:14 pm #64176With 3 Opp, 50 defects (not defectives) and 1000 units, go here http://www.isixsigma.com/sixsigma/six_sigma_calculator.asp?m=advanced and plug in the numbers. Then you can look to the right pane and look at how the calculations are done. Good luck

July 22, 2004 at 9:26 pm #64181First you have to define the opportunities for defects,based on your experience ,without this figure it is not possible to calculate DPMO.Regards

July 23, 2004 at 5:28 am #64198Dear JJ,

As everyone has said you need to know number of opportunity per unit before you calculate DPMO.

Still you can take down the formula to calculate it,

DPMO=(defects/opprtunities) * 1000000

defect = total defects in defective units

opportunities = total unit * number of opportunity per unit

For sigma calculation you can use following formula in MSExcel,

=Normsinv(1-defects/opportunities)+1.5

Cheers!

AjitJuly 23, 2004 at 5:34 am #64199Dear Ajit:

Very nicely explained. Just one last thing.

For sigma calculation you can use following formula in MSExcel,

=Normsinv(1-defects/opportunities)+1.5

Why calculate DPMO when it is not used at all in the sigma calculation? And, why add 1.5? Why not subtract it? Or, some other number other than 1.5 plus or minus? Cheers!

July 23, 2004 at 5:45 am #64200Dear JJ:

In the first post, you said 50 units failed out of 1000 units. Then you mentioned three failure modes. Let’s call the three modes A, B, and C. Out of the 50 failed units, may be 20 are due to mode A, 15 due to mode B and 15 due to mode C. The total failed units are 20+15+15 = 50. Does this mean there are three opportunities? I am not so sure. Or, a unit may have failed due to one or more modes prevalent at the same time. Is this possible in your application? For example, is there anything like AB failure mode, or BC failure, or AC failure, or ABC failure mode? May be we need to think about this some more instead of just rushing to some kind of DPMO and Process Sigma calculation. The math part is easy. It is the rest that is difficult. That’s where we must invest some more time and effort.

Charmed I am by all this.July 23, 2004 at 5:53 am #64201Dear JJ:

Just more clarification. Matt had pointed out the need to distinguish between defective and defects. A defective part can have more than one defect. I am assuming that what you call failed “units” is the same as “defectives” that Matt is thinking about. The modes A, B, C (since you mentioned three failure modes) that I refer to are the ways in which a “unit” becomes “defective”, or has one or more “defects”.July 23, 2004 at 6:30 am #64204Dear Charmed,

Answers to your questions,

1. DPMO is representative of capability as sigma is, the reason I mentioned it because it was part of JJ’s question.

2. 1.5 explains shift which difference in long term and short term capability, calculated over period of time and established emperically (Motorola was involved in the study). It is added to get the short term sigma. You can read about it in any six sigma book or site.

Also, just in case you are wondering why not subtract it, since you have just one sample for defect opportunity it can not give both ST and LT. The number you get out of normsinv gets treated as LT and 1.5 added to get LT.

(If you have more questions write to me on asingh1017@hotmail.com )

Cheers!

AjitJuly 23, 2004 at 7:11 am #64208Identical with my question,you hit the point,thanks and regards.

July 23, 2004 at 10:31 am #64212Dear Ajit:

Let me refer again to the equation for Process Sigma calculation, kindly mentioned in your post.

For sigma calculation you can use following formula in MSExcel,

=Normsinv(1-defects/opportunities)+1.5

I asked, why calculate DPMO when it is not used at all in the sigma calculation? And, why add 1.5? Why not subtract it? Or, some other number other than 1.5 plus or minus?

Now, let me ask another question. (Thanks for the email ID.) Why use all this Normsinv and try to calculate Process Sigma and wonder about 1.5 sigma shift which has been beaten to death in our threads here. I have the value of DPMO, or DPO = defects/opportunities, or DPU = defects/units and then, I say, leave me alone and let my boss and I deal with this. Forget all the rest. How about that for a new Smart Business Strategy (SBS), no SS, and no BS. Cheers.

July 23, 2004 at 10:40 am #64213Dear Ajit:

Just one more quick point. You said,

…..since you have just one sample for defect opportunity it can not give both ST and LT.

I didn’t interpret 50 failed units for 1000 units made in that fashion. The number 50 here could be some kind of an “average” developed over a period of time. We don’t know. It is not stated. Or, it could be just one data point, as you have interpreted. But, I doubt that. Or, imagine this. I break up 1000 units produced into smaller batches of 100 units or 50 units and then count the number of failed units. The total is still 50. Granted, I don’t have the details of what would happen. What if JJ has such info and can post it for us. (JJ do you have it and can you post it if you do?). Now what? We have to think about all these different possibilities before using the equation to find DPU, DPO, DPMO, or that Normsinv things with 1.5.

Sorry JJ. I know you asked a simple question. Hope all of this is still on topic for you. Have a great weekend. Cheers.July 23, 2004 at 11:20 am #64215Dear Charmed,

The formulas were facts about how to calculate (straight answers to straight questions ). Now that since you are talking about should one calculate sigma after DPMO ?

My two penny – Ofcourse not! It does not make any sense to calculate sigma. Just DPMO and period.

As far as data is concerned I am not sure if 50 is defects or defectives. Worse situation will be if the opportunities are not even defined. So cant comment until I know what to use DPU, DPO or DPMO.

Hope this will suffice. Let me know.

Thnx

Ajit

July 23, 2004 at 11:28 am #64216Dear Ajit:

The formulas were facts about how to calculate (straight answers to straight questions ). We are in agreement. Some need help in this area, and that is what you have provided and I commend you for that.

Now that since you are talking about should one calculate sigma after DPMO ? ……

My humble opinion is that beyond just blindly using mathematical formulae, we should think about what we are doing and why. That is what I would do, if I was a teacher in a class and had students who were learning about Process Sigma and how to calculate it. JJ did not provide any details. Neither you or I know anything more. But, we can certainly discuss the situation. I hope we soon hear from JJ. May be this helps clarify where I am coming from.July 23, 2004 at 11:41 am #64217Dear Charmed,

I am not surprised by your post, infact I was expecting it.

Your point is valid. I think it differently and I dont think so it is a problem as long as I make it clear what is being said.

If JJ has further clarifying question, would be happy to help.

Thnx

AjitJuly 23, 2004 at 12:28 pm #64223Ajit, I agree fully. And on that happy note we have to end this, unless we learn some more. Regards.

July 26, 2004 at 1:25 pm #64429Thank you for all of your responses. In my case, we rejected the components if we found one defect, such like one demention out of specs, or leaking. either one could cause reject. Say, one kind of parts, 4 units rejected in April because of out of specs, and 5 units rejected in May because of leaking, and so on. How can I get DPMP for last 12 month? In our company, we calculate the DPMP for each month and use (4/total sample)*1000000. but I do not have the total number for a year to evaluate a certain supplier.

July 26, 2004 at 2:45 pm #64439Cont.

If 4 units rejected in April out of 3000 samples, DPMO is (4/3000)*1000000

If 23 units rejected in May out of 1200 samples, DPMO is (23/1200)*1000000

Those are current practice. I am not sure it is correct. most of time the parts are rejected just because of one defect, and they are not going to check more if found one defect in a part and the part will be returned.July 26, 2004 at 2:51 pm #64441Dear JJ:

Thanks for coming back. Do provide more details, if it is not considered confidential and your bosees are not going to come after you. What does data for one year look like? May be you can make a small table which includes: Month, number made, number inspected, number rejected, reason for rejects (out of spec dimension, leaking, etc.) Please note I have suggested number made and number inspected, assuming less than 100% inspection. Since you mention “leaking” how is this determined? What is the component? What are the dimensions that are out of spec? If you can, tell us. Then later we can start talking about minimum loss to society. Wouldn’t that we wonderful if each one of us is focused on something so important! With my warmest regards.

July 26, 2004 at 2:59 pm #64443Sorry, I cannot answer your questions. Those questions are not related to my original questions and I think latest post is clear enough. Why are you worry about what dimentions? Dimentions are out of specs is one kind of defects. no more information.

I just want to know, in my case, what correct practice would be.

Thank you!

July 26, 2004 at 3:14 pm #64448Dear JJ: When you posted the second message, I was typing my message. There is a time delay between events. May be you did not realize that. In my humble opinion, your calculation of DPMO is inconsistent. Let’s see what others feel. No more questions. Cheers!

August 9, 2004 at 5:12 am #65545I appreciate greatly your input:guiding us to the right location .kind regards.

August 9, 2004 at 6:26 am #65550My request to all . Use this forum to get value out of the discussions and not to challenge each other. It is surprising that you are talking about helping some one and end up not doing that in the end and in addition confuse all others as well.

Now to JJ :

Are you calculating this to compile your in process yield OR to calculate the Outgoing Quality of the product ( as received by your end customer).

The formulae that Ajit gave you in one of the mails is the Right one to calculate the in process DPMO. However you need to be clear on the opportunities for Defects. It depends on the number of parts that goes into the assembly of the final part say in a PC board your opportunities for defects are the number of components used, no of solder joints and no of process steps that go into the product. Any of these that can go wrong and end up as a defect is an OFD. This is the trickiest part of Six Sigma. Once you have come up with a number for OFD stick to it . Don’t change. This number might impact your absolute value of the DPMO. But if you looking at calculating the improvement in DPMO it does’t matter. This is what Motorola taught us.

Don’t get hassled by the Sixma calculations.

In case you are calculating the outgoing defect of the factory, it is a Binominal (Bad/Good) and just calculate the % defective (50/1000) and that is truly what the customer sees and cares about.

Hope this gives you some relief.

regardsAugust 9, 2004 at 4:35 pm #65585Defects Per million opportunities (DPMO):

Process standard deviation (*) can be estimated to be s or can be figured using the following formulas:

DPMO= d/u X 1,000,000

Where d= number defectives units and u= number of units in total.

So for your problem the answer willbe

50/1000 X 1,000,000 = 50,000

August 9, 2004 at 4:55 pm #65592Defects Per million opportunities (DPMO):

Process standard deviation (*) can be estimated to be s or can be figured using the following formulas:

DPMO= d/u X 1,000,000

Where d= number defectives units and u= number of units in total.

So for your problem the answer willbe

50/1000 X 1,000,000 = 50,000

Long-Run Process Sigma:

0.8406+ Square root of (29.37-2.221* (In (DPMO))

* This formula embeds the assumption that in the long run, there will be a shift in the process mean of up to 1.5 sigmas, due to speacial cause variation. This 1.5 sigma shift is not a statistiocally prescribed value,but is instead based on long-term experience from Motorola’s manufacturing operations, which in gave rise to its inclusion in the process sigma caliculations throught out six sigma quality practice.August 10, 2004 at 3:53 am #65631I respectuflly disagree with that. Opportunities are based on what it takes to do the work, not on how it can fail. For instance, Motorola used to count each component on the bill of materials as one opportunity, and each electrical connection (solder joint, crimp, etc.) as one opportunity.

It didn’t matter that an electronic component could be placed in the wrong location, the wrong orientation, could be overlooked, could be damaged, etc. – installing the component was a single, countable assembly task. Likewise, a solder jonit can be defective for a variety opf reasons, but it still is just one joint. One opportunity per joint.

If you cannot establish an objective way to count the opportunities necessary to do a job, then each unit is a single opportunity.August 10, 2004 at 11:26 am #65646DPMO=(50/1000)*1000000 = 50000

sigma level = 3.14 (according to conversion chart of sigma levels to DPMO)

regards, VAugust 10, 2004 at 1:35 pm #65667Thank you for your information. I think my case would be Binominal. we are evaluating supplier quality level.

Thank you for everyone. I am clear of my questions now.June 5, 2008 at 3:40 pm #146640I couldn’t get access to call the Excel function properly. If you are using Office 2003, the function/dll file Excel calls is not even available! Try this for a complete VB solution:

http://www.source-code.biz/snippets/vbasic/9.htmThis worked flawlessly for me. - AuthorPosts

Viewing 30 posts - 1 through 30 (of 30 total)

The forum ‘General’ is closed to new topics and replies.

© Copyright iSixSigma 2000-2018. User Agreement. Any reproduction or other use of content without the express written consent of iSixSigma is prohibited. More »