DPMO (Defects Per Million Opportunities) vs. RTY (Rolled-Throughput-Yield)

[Sushil]

I need some help in understanding DPMO Vs RTY. See attached Presentation slide which I downloaded from Web. Can I relate DPMO to RTY ?. What's the rationale behind this. Is this correct ?. I did some manual calculations myself in one of the areas we are working on, this does'nt seem to be true. For my case I did the RTY calculation based on Poisson's equation & probability of zero defects as e ^(-DPU). Can someone guide me on this and how can I explain relation between RTY and DPMO..? How can I calculate RTY based on DPMO data ?.

[wmarhel]

Originally Posted by Sushil

Can I relate DPMO to RTY ?. What's the rationale behind this. Is this correct ?

For RTY (Rolled-Throughput-Yield) it doesn't matter whether there are 10 defects or just a single defect. It failed to meet to the necessary quality requirements to move onto the next operation/work center/etc.

I'm not a huge fan of DPMO as it relates to the manufacturing process, it's more of a tool to used during the design and re-design of new products/processes. A defective product is a defective product, irregardless of whether three things don't work on it, or just one. Sometimes it feels that there are engineers out there who believe that if it isn't broken.....there just aren't enough features on it.

Here's a thread on DPMO which may or not help: DPMO calculation for Testing stations- ideal way? High volume PCB manufacture

Wayne

[James Oldham]

Sushil,

You can convert RTY into DPMO and back again - they are just different ways of showing the same number. RTY as a % is # good/total, therefore 1-RTY is the % defective. Since % defective is technically "defects per hundred", just multiply this number times 1,000,000 to get DPMO.

Is the glass half full or half empty? This is another one of those cases why lots in the forum shun Six Sigma. Just focus on the # of defects and fix the problems - the form in which you report the number doesn't matter.

[Sushil]

Originally Posted by James Oldham

Sushil,

You can convert RTY into DPMO and back again - they are just different ways of showing the same number.

James.. pardon me

I am still confused. I do not understand why DPMO % is being related and termed as RTY. I did some study from a station in floor and following were my findings, which confuses me more..
total Defects =500
Total Opps =7787260
DPU = 0.41
RTY =e^-DPU=66.4%
DPMO = 64.21

as per table attached in my first post
DPMO %(RTY) for 64.21DPMO will be 99.99353%

why is this difference..? and is there any issue terming DPMO% as RTY ? can it be done..

[gvalenti9]

Hopefully I can explain this clearly.

total Defects =500
Total Opps =7787260
DPU = 0.41

m = number of opportunities/unit = 6383. I calculated this using the total # of defects (500) divided by the DPU to derive that the data was based on 1220 units then divided the total opps by the # of units.

RTY =e^-DPU=66.4%
DPMO = 64.21

Yna = Normalized yield = RTY ^ (1/m) = 0.664^(1/6383) =.99993585 which matches up to 99.99353% that you calculated

DPMO can be calculated from the Yna or from actual yield data
DPMO = (1-Yna)*10^6 = (1-0.9999353)*10^6 = 64.15
DPMO = # defects/# opportunities *10^6 = 500/7787260*10^6 =64.21

So then, why does RTY not equal DPMO and DPMO not equal DPMO?

DPMO = # defects/# opportunities *10^6 is an actual calculation based are units processed, number of defects found, and number of real opportunities per unit.

DPMO = (1-Yna)*10^6 is based on the Poisson distribution used to calculate RTY. RTY = e^-DPU assumes defects are distributed randomly and are assumed to follow a Poisson distrubtion. Yna, therefore, reflects the assumed Poisson distribution used to calculate RTY . As the defect rate falls below 10%, Yna and DPMO converge. The data that you have shows them converging quite nicely. It appears that you were only missing the one step DPMO = (1-Yna)*10^6

The benefit of the DPMO is that it normalizes the data so that assemblies of different complexities/defect opportunities can be compared objectively. The benefit of RTY is that if there is an issue with yields, it becomes immediately obvious.

[QualityKangaroo]

Sushil or anyone else out there,

I am responding to why DPMO can/can not be related to RTY.

I have found that one major difficult (as pointed out above) is the normal distribution assumption... as you may know, even if an overall process exhibits normality certain process steps do not. It is actually helpful to compare CpK (capability) to FTY (in RTY sense) to check this assumption... though not foolproof it may show you some interesting things happening in the process. However, the main reason I have added to this thread: No one seems to have brought up the point that DPMO looks at defects which, let us assume, are properly recorded and assigned a cause and process step. We can usually pull this data with ease. However RTY also pulls in "rework". It is sometimes dangerous to assume that all "out-of-spec"/"out-of tolerance" are either recorded as waste or reworked. Sometimes in-spec product is re-worked for an number of reasons (good or bad) and is therefore not captured in the DPMO umbrella. This is actually why I prefer RTY when looking at a process - it can direct you to little hidden rework zones that DPMO often misses. So, in conclusion DPMO misses out on some of the rework and therefore gives an incomplete picture of the reality on the shop floor.

Hope this helps.

Q. Kangaroo

[Jim Wynne]

Originally Posted by QualityKangaroo

Sushil or anyone else out there,

I am responding to why DPMO can/can not be related to RTY.

I have found that one major difficult (as pointed out above) is the normal distribution assumption... as you may know, even if an overall process exhibits normality certain process steps do not. It is actually helpful to compare CpK (capability) to FTY (in RTY sense) to check this assumption... though not foolproof it may show you some interesting things happening in the process. However, the main reason I have added to this thread: No one seems to have brought up the point that DPMO looks at defects which, let us assume, are properly recorded and assigned a cause and process step. We can usually pull this data with ease. However RTY also pulls in "rework". It is sometimes dangerous to assume that all "out-of-spec"/"out-of tolerance" are either recorded as waste or reworked. Sometimes in-spec product is re-worked for an number of reasons (good or bad) and is therefore not captured in the DPMO umbrella. This is actually why I prefer RTY when looking at a process - it can direct you to little hidden rework zones that DPMO often misses. So, in conclusion DPMO misses out on some of the rework and therefore gives an incomplete picture of the reality on the shop floor.

Hope this helps.

Q. Kangaroo

Why would conforming product need to be reworked?

[QualityKangaroo]

One reason (though not a good one) that I often run into is that an operator attempts to react to an existing condition communicated by another team member (downstream)... it may not even be something we currently measure but which makes the job easier down the process. Secondly, and perhaps more troubling, is that an operator establishes their own reaction limits. Most of the time this can be addressed by proper coaching and follow-up training, but I have visited many plants with the operator is prideful to a detriment - they chase a target and cause greater variation AND often rework acceptable product. One simple example: I'm making cookies and I don't like diameter (even though its in spec)... I take the dough and replace it in to a mixer. The mentality expressed is often, "I want to better, what's wrong with that?"... suddenly you have a less efficient process (and likely one with more variation considering the operator attempt to make the diameter "better"). I hope this clarifies