iSixSigma

Counting Opportunities and Taguchi Loss Function

Six Sigma – iSixSigma Forums Old Forums General Counting Opportunities and Taguchi Loss Function

Viewing 41 posts - 1 through 41 (of 41 total)
  • Author
    Posts
  • #36287

    Charmed
    Participant

    Dear Forum Members:
    Before we can estimate the Process Sigma level, we must be able to determine the DPMO. This is defined as the ratio of number of defects to number of opportunities (DPO). The ratio is multiplied by one million to convert  DPO to DPMO. The more difficult part of this ratio calculation is the denominator – the number of opportunities.  How do we count the opportunities? 
    I found two very interesting articles by Gary Cone on this subject (Use keryword search, type Gary Cone opportunties to get the articles). 

    Six Sigma Opportunity Counting
    Correct Method to Count Six Sigma Opportunities
    The following is taken from the first article.  Please note the bolded sentences. The question at the end was asked by Gary Cone himself. In the second article,
    Some caveats for opportunity counting (the first three could actually be caveats for Policy Deployment):

    Whatever method is chosen, it must not conflict with other company initiatives.
    Whatever method is chosen, it must be simple and uniform throughout the organization.
    Companies pursuing defect reduction are also pursuing time reduction (responsiveness is the real competitive weapon, defects just get in the way of that).
    ‘Complexity’ is a measure of how complicated a particular product or service is. It’s doubtful that we will ever be able to quantify complexity in an exacting manner.
    If we assume that all characteristics are independent and mutually exclusive, we may say that ‘complexity’ can be reasonably estimated by a simple count. This count is referred to as an ‘Opportunity Count.’
    In terms of quality, each product manufacturing or service delivery step represents a unique ‘opportunity’ to either add or subtract value.
    Remember, we only need to count opportunities if we want to estimate a sigma level for comparisons of goods and services that are not necessarily the same.

    (Please note the bolded sentence.) That said, how do we get good guidance on how to do this? After all, the sigma metric in Six Sigma is based on an accurate counting of opportunities. If there are to be any rational discussions and comparisons of your organization’s capabilities, this must be seriously considered.
    Let’s turn to that touchstone of clarity, the Holy Grail for all Six Sigma professionals, Mikel Harry’s The Vision of Six Sigma: A Roadmap for Breakthrough. Chapter 12, Nature of Opportunities, discusses opportunities. While not exactly straightforward, there is a discussion of ‘active’ vs. ‘passive’ opportunities, where the conclusion on page 12.16 is that if we do not measure (test or inspect) the opportunity, it is not really an opportunity. So opportunities must be the things we test or inspect. Right?
    Well, Motorola really touts the ‘bandit’ pager line in Boynton Beach, Florida, as an example of Six Sigma thinking, and one of the strongest points they make is that the product receives no test or inspection prior to being shipped to the customer. But by Mikel Harry’s definition, no test, no inspection, no active opportunities, sigma level does not exist, therefore this process cannot be compared or learned from.
    Anybody seeing anything wrong with this picture? This definition cannot be correct, but what is? Post your thoughts on the iSixSigma discussion forum. (Question asked by Gary Cone.)
    The following is from the second article.  Here Gary Cone talks about the focus on counting opportunities, which mean, by definition means testing and inspection – not necessarily value-added activities. 
    The best manufacturers in the world do very little test and inspection. Why? They don’t need to. The best manufacturers in the world also have test or inspection time limited by the line flow. So now in order to embrace Six Sigma, these companies would have to add test and add inspection? I don’t think so.
    What is wrong with the focus on test and inspection for counting opportunities? (Question asked by Gary Cone.)

    The focus is on non-value added activities.

    The more we are sure of quality, the less we will test and inspect potentially worsening the sigma level. At its extreme, the denominator of the DPMO equation is 0, which means we have an undefined sigma level.
    I have seen other threads where Green Belts are discussing how to pick “good” SS projects and the discussion here often seems to ignore the “value-added” component of all activities that has been emphasized by Gary Cone. In one thread, where a green belt wanted to start a project, when some one raised the question of cost savings to the company, another green belt said, “That should not be the issue to be considered in the choice of a good SS project.  Cost savings will follow automatically from any good SS project. “
     
    Why am I digging myself into this big hole?  Quite simple. 
     
    Gabriel started talking about minimum loss to society and I started thinking about the Taguchi Loss Function.  These ideas are intertwined. From Gary Cone’s discussion it is clear that we cannot overlook “costs” even in the very first step of picking a SS project.  I do see the problem though – most greenbelts are not high up in the management chain to have any meaningful idea of “costs” or have any access to financial data.  The solution may lie in including some “green belts” from the finance department as well, who might be able to provide this much needed insight.  Engineers alone cannot make it.  Neither can bean counters working alone make it. I look forward to your thoughts on this subject – from all levels – GB, BB, MBB, Champions and you name the Almighty CEO. My CEO is already on my tail for engaging in – guess what? Have a great week. With my warmest regards.   

    0
    #104305

    Charmed
    Participant

    Dear All:
    Sorry about the varying font sizes.  That’s how it turned out when I finished copying and pasting.  All the stuff starting with “I have seen.. .” are my thoughts. Rest is from Gary Cone’s articles. Have a great week.

    0
    #104317

    Cone
    Participant

    Charmed,
    Good post. Thank you for trying to bring up the level of discussion here. I get the message.
    I have two thoughts on this.
    First on the question of opportunity counting. If you do it, it should only be for one purpose; comparisons of capability to be able to know where to learn or where to apply additional resources. It has no value to the GB, BB, MBB in terms of working projects. It also should not be used for choosing projects. Projects flow from strategy and should only be on those processes that have an effect on the gap between who we are and who we want to be. That’s all strategy is!
    Second on the question of the Taguchi loss function. Everyone should understand it and it’s complementary capability measure Cpm. There are very few cases where Cpm should not be the dominant metric instead of Cp and Ppk. If you don’t know what the target should be assume it is the center of a bilateral specification or the “desired” end of a unilateral specification.
    Clearly the passive and active opportunity discussion is wrong. Clearly Taguchi, in all of the things he taught us, is right. Look beyond loss function at using DOE in design, inner/outer arrays, etc. If we comprehend these things, questions like active/passive opportunities, the shift, using 4.5 to represent 6 on attribute data, etc. will become passe.
    Gary

    0
    #104320

    Charmed
    Participant

    Dear Gary, or should I say Gary Cone? If so, do let us know.  I would feel very honored. I am still trying to see how to tying opportuntities, defects, etc. to Taguchi’s Loss Function. Would love to hear from others as well. With my warmest regards.

    0
    #104335

    Need Direction
    Participant

    Consider four performance opportunities that defines a
    system called “X.” For our discussion we will say the
    opportunities are A, B, C and D. Suppose only A and B
    are “active.” This says they are regularly assessed in
    terms of “performance to standard.” However,
    opportunities C and D are not assessed, so they are said
    to be “passive” in nature. Now, say we observe DPU=1.0.
    What is the DPMO?

    0
    #104336

    Charmed
    Participant

    Dear Need Direction:
    Very good question. If it is not a trick question, I think you already know the answer. You don’t need direction.

    0
    #104346

    Peppe
    Participant

    DPU is defined by IPC as “average defect per unit”. Do you mean the same ?  Rgs, Peppe 

    0
    #104374

    Tim F
    Member

    Charmed asked
    “Well, Motorola really touts the ‘bandit’ pager line in Boynton Beach, Florida, as an example of Six Sigma thinking, and one of the strongest points they make is that the product receives no test or inspection prior to being shipped to the customer. But by Mikel Harry’s definition, no test, no inspection, no active opportunities, sigma level does not exist, therefore this process cannot be compared or learned from.
    Anybody seeing anything wrong with this picture?”
     
    I see room for misunderstanding, but nothing specifically wrong.  The key phrase is “prior to being shipped.”
    Motorola certainly tests during manufacture, to ensure that processes stay in control.  They know the odds of a bad solder joint or a defective battery.
    What they apparently don’t do is any final inspection.  Their previous knowledge gained by monitoring the process along the way makes them confident that there will be so few defects that it isn’t worth having someone at the end of the line pull out some or all of the units to see if they will really turn on and display a message.
     
    Tim F
     

    0
    #104376

    Charmed
    Participant

    Dar Tim F:  That is not a question that I asked.  It is a question Gary Cone asks in the original article. I thought I had made that very clear. Sorry, for the misunderstanding. With my warmest regards.

    0
    #104378

    Charmed
    Participant

    Dear Tim F:   You said, “What they apparently don’t do is any final inspection.” You are very right. I agree completely. Again, it was Gary Cone who posed the question about counting opportunities (and its relation to test/inspect) which you can find in the original article.  Your remarks actually open a potential “flaw” in the entire line of reasoning employed to arrive at DPMO, which is why I made this post. With my warmest regards.   

    0
    #104390

    Tim F
    Member

    Charmed,
     
    Sorry about attributing the question directly to you, rather than to the original source of the question.  I just didn’t read carefully enough!
     
    Could you elaborate on the potential flaw you think I have uncovered?  I can think of several reasons why DMPO calculations might be flawed (distributions aren’t truly normal, interactions between defects, not counting defects that are considered inconsequential, and perhaps more if I think about it) but I’m not sure which, if any, of these you have in mind.
     
    Tim F  

    0
    #104398

    Charmed
    Participant

    Dear Tim F:
    Thanks for asking the question.  You have given me a chance to discuss something that motivated my post. Let me start with the  definition of Opportunity, found on this website, and the direct quotes from Gary Cone’s second article, as mentioned in my first post in this thread.
     https://www.isixsigma.com/dictionary/Opportunity-102.htm
     
     
    OpportunityAny area within a product, process, service, or other system where a defect could be produced or where you fail to achieve the ideal product in the eyes of the customer. In a product, the areas where defects could be produced are the parts or connection of parts within the product. In a process, the areas are the value added process steps. If the process step is not value added, such as an inspection step, then it is not considered an opportunity.Opportunities are the things which must go right to satisfy the customer. It is not the number of things we can imagine that can go wrong .
     
    Correct Method To Count Six Sigma Opportunities
    By Gary Cone
     
    https://www.isixsigma.com/library/content/c010506a.asp
     
    What is wrong with the focus on test and inspection for counting opportunities?

    The focus is on non-value added activities.
    The more we are sure of our quality, the less we will test and inspect potentially worsening the sigma level. At its extreme the denominator of the DPMO equation is 0, which means we have an undefined sigma level.
    We could disrupt the flow of work trying to comply with this definition.None of this sounds right, does it?
    After definining opportunities as arising when we conduct tests/inspections, Gary Cone then poses these three questions and also proceed to answer them. I was puzzled by point 2 above. At its extreme the denominator of the DPMO equation is 0, which means we have an undefined sigma level. 
     
    Can the denominator (opportunities) in the ratio Defects/Opportunities actually go to zero, with the numerator (defects) having a finite value?  I was trying to tie this assertion to the remarks about Motorola’s pager which does go through any inspection (final, as you have pointed out).  Does inspection mean no opportunity?  This cannot be true. Is there any situation like X defects but zero, or tending to zero, opportunities. Hence, I made this post.
     
    It also reminded me of what we learned in our calculus courses.  Remember finding the limit of f(x)/g(x) = 0/0.   An example is the function sin x / x  in the limit as x goes to zero.  Although both numerator and denominator go to zero in this limit, we know that the ratio goes to 1.  I started thinking along these lines when I read point 2 above and the notion of zero opportunities.  Are we missing something? It might just be that we must recognize the fuller meaning of the simple equation, 
     
    Good Parts = Total Parts – Bad Parts
     
    Or, Good/Total = 1 – (Bad/Total)
     
    For each part there are a certain number of opportunities.  This is consistent with the above definition of opportunities. Hence, we can say that total opportunities = k (Total Parts) where k is 1 or some constant greater than 1, specific to the part. (Put many parts together, we get a “product”.) The ratio Bad/Total must scale in some unknown way as Defects/Opportunities and the denominator can never become zero for reason just noted.
     
    Again, thanks for asking. I would be glad to hear your view of this line of reasoning. With my warmest regards.  

    0
    #104399

    Charmed
    Participant

    Dear Tim F:  I don’t know why these multiple font sizes appear after pasting?  Anyway, I have to fix two important points, see bolded, which I missed in editing.
    Motorola’s pager which does (not) go through any inspection (final, as you have pointed out).  Does (zero) inspection mean no opportunity?

    0
    #104407

    Reigle Stewart
    Participant

    To Gary Cone and All Others. The following question was
    posed earlier in this thread: Consider four performance
    opportunities that define a system called “X.” For our
    discussion we will say the opportunities are A, B, C and
    D. Suppose only A and B are “active.” This says they are
    regularly assessed in terms of “performance to standard.”
    However, opportunities C and D are not assessed, so
    they are said to be “passive” in nature. My discussion now
    picks up from here. Lets pretend that we make a single
    system called “X” containing A, B, C and D. To expand
    the discussion, say we observe DPU=1.0 following the
    creation of the first system. We will now say that defect
    opportunity B was found to be defective following the
    production of system X, but opportunity A was not
    defective; however, opportunities C and D were not
    surveyed for conformance to standard. Opportunities C
    and D were never looked at or evaluated for conformance
    to standards! What is the DPO? Seems to me that DPO =
    D / O = 1 / 2 = .50. Why? Because every defect
    opportunity must be capable of yielding a defect. If the
    opportunity is not evaluated for conformance to
    standards, the probability of defect is zero. If the
    probability of defect is zero for a particular opportunity, the
    opportunity is not really an OPPORTUNITY FOR
    DEFECT. Therefore, it should not be counted. This
    implies that the numerator term would be illustrated by the
    binary sequence 0 + 1 + 0 + 0 = 1 and the denominator
    term would be given as 1 + 1 + 0 + 0 = 2.. It would be
    incorrect to say the denominator term should be 1 + 1 + 1
    + 1 = 4 since only the first two “slots” of the denominator
    are “active.” Consequentially, the denominator
    summation would be 1 + 1 + 0 + 0 = 2. Notice that a “1” in
    the denominator term says the opportunity is active and a
    “0” says it is passive. Another way of saying this is that a
    every slot of the numerator must be capable of yielding a
    “1” in the corresponding numerator term. If not, the
    corresponding “slot” in the denomintor must be assigned
    a value of “0.” So each “blank” position (i.e., slot) in the
    denominator represents an opportunity that has been
    created. But if any such opportunity cannot produce a
    defect, then it is not a true defect opportunity – it is merely
    a “production opportunity.” Since it cannot produce a
    defect (for whatever reason), it should be assigned a
    value of “0”. The opportunity may have been created, but
    no defect density can be realized even though a total of 4
    production opportunites were created (recognizing that
    not all 4 were also defect opportunities). So, we should
    only count “active” defect opportunities. Seems to me this
    is a very simple rule for counting defect opportunities and
    ensuring the proper summations (for the numerator and
    denominator terms). Respectfully, Reigle Stewart

    0
    #104408

    Charmed
    Participant

    Dear Reigle Stewart:
    I am truly delighted to see that you have chosen to participate.  Let me say, upfront, that I do find some things quite confusing in your exposition. Please do bear with me since I wish to “count” this as learning “opportunity” for me at least. I beg your patience with me as we review the definition of opportunity, which I found on this Website.  This is given below in its entirety once again.  (Please note the bolded sentences, bolding is added by me.)
    https://www.isixsigma.com/dictionary/Opportunity-102.htm

    OpportunityAny area within a product, process, service, or other system where a defect could be produced or where you fail to achieve the ideal product in the eyes of the customer. In a product, the areas where defects could be produced are the parts or connection of parts within the product. In a process, the areas are the value added process steps. If the process step is not value added, such as an inspection step, then it is not considered an opportunity.Opportunities are the things which must go right to satisfy the customer. It is not the number of things we can imagine that can go wrong .
     
    See Defect.———————-[From the discussion forum:]”Some folks think we measure opportunities by counting how many ways something can go wrong. That is a bad approach, because it inflates the denominator. Motorola had a fairly simple approach: count number of parts plus number of connections. Period. In another discussion thread I read that Allied-Signal used a formula of multiplying bill of material part counts by three. Both approaches are straigthforward and repeatable.”If your process is adminsitartive, it probably will be very difficult to be as simple or as repeatable. Don’t sweat it. You can use defects per unit as easily as defects per opportunity. The idea is to measure the right things, and to understand how the process varies over time, so that it can be improved.”———————-Opportunities from a customer’s standpoint really do not make any sense. When you hand something off to a customer the opportunity is once – just like you recieving supplied material. They either sent you the correct thing or not, and if it isn’t perfect then it is defective. At the customer level you need to treat it as one.You use opportunity counts to account for complexity differences internally (or benchmarking something externally). If you make jet engines, light bulbs and loans the opportunity count will level the playing feild. If they are all 1 defect per unit they are not equivalent operations. If I calculate rolled throughput yield from defects per unit it still says they are equivalent. If I put in the opportunity count I can differentiate betweent them.

    Posted By: TomModified By: JonathonLast Modified: Sep. 1, 2002
     ———————————————
    Hereon are my thoughts.
    ———————————————
    I would like to offer the following example, very respectfully, which may be similar to the system “X” that you have defined for us, with opportunities called A, B, C, and D. 
    Let the system “X” be the system that produces Cadillac’s.  These must be painted. We want to produce Black Cadillac’s.  But, our painting system, for some reasons unknown is producing four different colored Cadillac’s: White, Red, Blue, and Green.  Any color other than the Black color means “nonconformance” and the Cadillac will not be accepted by the customer.   I know this is an extreme example and no factory will operate and continue to allow a painting system to function in this manner.  But, I can readily comprehend what “conformance to standard” and “nonconformance” means.  I hope you will permit me to go along with this notion of what you have called system “X” and the states of conformance or nonconformance.  I will give other examples as well that are easier for me to understand.  Sorry, for breaking this message up into two parts.  I do this, humbly, to make this a bit more clear and beg your indulgence.  With my warmest regards.

    0
    #104409

    Charmed
    Participant

    Dear Reigle Stewart:
    I continue here the discussion from the last post. The first part is based on an exchange between Tim F and myself since I had been thinking about Gary Cone’s two articles. So, here it goes.
    Total Parts = Good Parts + Bad Parts 
     
    Bad Parts = Total Parts – Good Parts
     
    Bad /Total = Defects/Units = 1 – (Good/Total)
     
    So, if DPU = 1, Good = 0; all parts are Bad.
     
    Number of Opportunities = k (Total Parts) where k = 1 or some number greater than 1.
     
    Defects/Unit = DPU = Bad Parts /Total Parts
     
    Defects/Opportunity = Bad Parts /k (Total Parts)
     
    So, perhaps, let’s say k = 4 in Reigle Stewart’s example.
     
    Defects/Opportunities = DPO = 1/k = 0.25 to 1 depending on value of k.
     
    In my humble opinion, the value of k is uniquely fixed for a “system”.  It does not depend on how we choose to “evaluate” it for nonconformance, or how we choose to observe its performance. If you roll a pair of dies, the probability of getting two 6s is 1/36 = (1/6) for first die multiplied by (1/6) for the second die.  This assumes that the two events are independent and the occurrence of one does not affect the other. We might not observe the event “two 6” for a long time.  But, this does not change the probability of occurrence of the events (within the context of the classical assumption that all events are equiprobablity of occurrence of all events). We must wait until two dice are rolled at 36 times before we can determine the probability of the event called “two 6”. The “evaluator” and whether or not something is being “evaluated” does not fix the probability of occurrence of an event or the opportunity for the defect. 
     
    Now let’s read Reigle Stewart. I quote verbatim, “…..however, opportunities C and D were not surveyed for conformance to standard. Opportunities C and D were never looked at or evaluated for conformance to standards! What is the DPO?” I find this line of reasoning strange and the reasons have been stated.  Let’s also go back and re-read Reigle Stewart more completely. 
     
     Because every defect opportunity must be capable of yielding a defect (Let’s think about this within the context of rolling a pair of dies.) If the opportunity is not evaluated for conformance to standards, the probability of defect is zero. (This is a very new philosophy to me and I am not familiar with this approach.  The process of “evaluation” and therefore the evaluator, in a sense, we are being told, determines the probability. Does this mean, if a giant tree falls in a forest and no one was around when it happened, it did not make a sound! Or when it rained during the huge deluge and water submerged all the land and nobody was around, there was no water on the land!) If the probability of defect is zero for a particular opportunity, the opportunity is not really an OPPORTUNITY FOR DEFECT. Therefore, it should not be counted. (If you do not observe two 6s, we should change the rules for the probability of getting a 5 or a 3 or a 7. Sorry, again, I must say, humbly, that I find this very hard to accept.) 
     
    But these are just my thoughts.  May be this is the reason why nobody really understands statistics-based arguments; please also see my post where I quoted a recent discussion on the Atkin’s diet on the O’Reilly Factor.  I was looking at the discussion purely from the point of view of the credibility that statistics, as a science, seems to have.  “Studies can be made to show almost anything you want”, said Mr. O’Reilly, twice on that show and the good doctor, who is the Director of the Atkins Center had no defense.  Now, I seem to have no clue either.  Perhaps, it is my inability to comprehend the language. I am really trying to understand this point of view of how to count “opportunities”.  May be, one more, will present itself, soon.  I look forward to a congenial and professional exchange on this topic, without some of the recent acrimonies. With my warmest regards.

    0
    #104410

    Charmed
    Participant

    Dear All:
     
    In the previous post, Defects/Opportunities = DPO = 1/k = 0.25 to 1 depending on value of k.
     
    From this simple equation, we do get the value DPO = 1/2 = 0.50, given by Reigle Stewart, if we take k = 2.  But, we also get DPO = 1/1 = 1 (which is the same as DPU = 1, which Reigle Stewart begins with before converting to DPO), or DPO = 1/3 = 0.33 and DPO = 1/4 = 0.25 for the values k = 1, 3 and 4. 
     
    Now we must all agree on the correct value of k.  Reigle Stewart’s arguments mean that when C and D are simply declared as being “passive” the system X has changed permanently and from now on we are only permitted the value k = 2. This is the only point of disagreement. I think, this is something that can be resolved easily by the experts in this forum. Let ALL of us agree, at least once. With my warmest regards.

    0
    #104411

    Charmed
    Participant

    Dear Gary:
    I am coming to your post again and I think we are in full agreement.   Let’s review your remarks together.  Your message is pasted below with parenthetic comments.
    Charmed,
    Good post. Thank you for trying to bring up the level of discussion here. I get the message.
    I have two thoughts on this.
    First on the question of opportunity counting. If you do it, it should only be for one purpose; comparisons of capability to be able to know where to learn or where to apply additional resources. It has no value to the GB, BB, MBB in terms of working projects. It also should not be used for choosing projects. Projects flow from strategy and should only be on those processes that have an effect on the gap between who we are and who we want to be. That’s all strategy is! (I AGREE)
    Second on the question of the Taguchi loss function. Everyone should understand it and it’s complementary capability measure Cpm. There are very few cases where Cpm should not be the dominant metric instead of Cp and Ppk. If you don’t know what the target should be assume it is the center of a bilateral specification or the “desired” end of a unilateral specification.
    Clearly the passive and active opportunity discussion is wrong (I HAVE TO AGREE, based on the posts in response to Reigle Stewart, or is it RS debating himself as Charmed, I am confused!) Clearly Taguchi, in all of the things he taught us, is right. Look beyond loss function at using DOE in design, inner/outer arrays, etc. If we comprehend these things, questions like active/passive opportunities, the shift, using 4.5 to represent 6 on attribute data, etc. will become passe. (Charmed, no Reigle Stewart, no Charmed, agrees. 
    Gary
    P. S.  Good Lord, or Merciful Lord, Help Us All Thy Children here on Earth. This guy does not even know his own identity anymore, he has morphed into a higher being with no body and just a soul and whom no one can recognize anymore!

    0
    #104415

    Anonymous
    Guest

    I hope you don’t mind me cutting but I am not aware of any defect counting in Taguchi Methods. I have only ever encountered %defective. From my own interactions with Gen-ichi, he seems to place more emphasis on the cause of defects.
    Tim’s list raised one of the most important issues with DPMO, which is the assumption of independance. In my experience defects are not always independant, which means that it is important to distinguish between those that are random and those that are systematic.
    The idea that a quality engineer can measure the ‘quality’ of a supplier by counting defects rather than %defectives is a fallacy that can be easiliy checked using a simple model. This method was widely used in Motorola in the mid-80s to identify low yielding products within product families and was one of the reason why they invested so heavily in failure analysis.
    (In 1987/8, Motorola Austin had one of the best semicondcutor failure analysis laboratories in the world .. we could remove passivation while keeping a product fullly functionally, ion mill down to first level poly, deposit a tungsten film and probe the gate!)
    As an example of the two types of defects, consider a diffusion process there are obviously random point defects due to particulate cantamination. On the other hand there are also dislocations formed at the edge of silicon wafers as they cool down, depending on the wafer separation, the cooling down rate, and the pull rate. (These defects are not visible to the naked eye.)
    By way fo contrast, the step-by-step confirmation method of TPS would perform a final check …. even if z > 12 sigma. This is called process self-assurance at Fujifilm.
    Isn’t it about time we all got back to quality engineering!
    Regards,
    Andy U

    0
    #104419

    Charmed
    Participant

    Dear Andy U:
    Do I necessarily have to follow the current Six Sigma Dogma? I don’t think so. That’s what I hear from you in other posts when it comes to 1.5 sigma shift and what you lament SS has now become. Do I have to follow what they did at Motorola? Then, it appears you have not not read my posts carefully. I am trying to do something else which is explained in the posts. I started with Gary Cone’s articles and tried to make sense of them.  That’s all. No more no less.  Eventually, I am not there yet, I would like to see how to connect this with what Taguchi taught us.  Right now, I am trying to find out about the correct way to count defects, DPU, DPO, and DPMO, all that good Motorola stuff that is being applied in ALL fields of human endeavor – without any justification. You have a different perspective and I have mine. If we cannot meet, that’s fine too.  
    And, I repeat, this is not Reigle Stewart. Such speculations that are going here are, in my humble opinion, unprofessional and unbecoming of eminent people like yourself. I hope, at the very least, we can agree on that.  Let’s focus on the message not the messenger, or even the origins of the messenger. With my warmest regards.    
    P. S. This is my last post, at least for the day.

    0
    #104422

    Mikel
    Member

    Reigle,
    Are you having fun?

    0
    #104425

    Anonymous
    Guest

    Dear Charmed,
    I’m glad that you’re not willing to swallow the Six Sigma dogma – this is what I’m trying to tell everyone:
    WHAT WE NOW KNOW AS SIX SIGMA WAS NOT WHAT WE USED IN MOTOROLA WAFERFABS IN MID-1980.
    Given the claims of certain indivuals who are always citing Motorola, I believe this is an important message! In fact, I’m surprised that there have been no product liablity suits.
    Moreover, Motorola mobile phone products have a pretty dire reputation these days, notwithstanding the ‘improved’ Six Sigma program.
    Yes, this is lamentable – so is the economic performance of many corporations in the USA. What is going to happen when the USA becomes a second-rate economy?
    In answer to the question you brought – use %defective or DPU. DPO is not sufficently general to be applicable to every industry. My final advice is to GO LEAN ……and use step by step confirmation process self-assurance.
    Cheers,
    Andy

    0
    #104426

    Charmed
    Participant

    Dear Andy U:
    Thank you, sincerely, from the bottom of my heart. Out of my deep respect for you (I know about your Motorola connections and that you worked with Taguchi himself, based on my reading of the material posted right here on this website), and mindful of the very professional and civil nautre of your post, I have chosen to reply. However, I still continue to be totally disgusted by all the snipping that went on yesterday – and literally meant it that I will sign off for good.  (Not that anyone is going to give a hoot – I don’t put myself in any elevated category but there is life beyond this forum – that’s all I mean.)
    Yes, I am concerned about many of the things that you have mentioned (Motorola products like cell phones are a joke and I don’t use them, many U.S. and for that matter many European corporations are not profitable, etc.) and am sincerely trying a new track. The sorry state of U. S. and European corporations has nothing to do with Six Sigma – just bad policies with no long term vision.
    Now, please do us all a favor and tell your friend(s) Stan(s) to learn some civility.  If he/she/they want to exchange personal messages with you, that should be offline. After what we went through yesterday, I am quite sure many in the forum would agree. With my warmest regards. 

    0
    #104427

    Charmed
    Participant

    Dear Andy U:
    If you read my posts carefully, you will see that we are in full agreement on the point that you have mentioned below.
    In answer to the question you brought – use %defective or DPU. DPO is not sufficently general to be applicable to every industry. My final advice is to GO LEAN ……
    When I say “Parts” or “Units”, it is the same as what you call “defectives”. A single part, or unit, or a defective, can have more than one defect. So, DPU should be the right measure, not DPO or DPMO. Or so, it seems. 
    But, you have been at Motorola and you well know that “opportunities” and therefore DPO and DPMO are also good metrics to use if we want to control the process and improve its capability.  When do we use DPU and when do we use DPO?  That is what everyone has to understand.  And, following this trail, as one who has worked with Taguchi, you will soon see how to relate DPU and Taguchi’s Loss Function.  Again, I take a slightly different view of Taguchi’s teachings.  I also tend to take a slightly different view of many other great works. I call it exploration and learning – all legitimately part of Six Sigma.

    0
    #104429

    Anonymous
    Guest

    Charmed,
    Don’t take anything in this forum personally. As I’m sure you appreciate – forums are not real, and many posters are not real either – such is the brave new world in which we live .. or is it?
    Best wishes,
    Andy

    0
    #104430

    Anonymous
    Guest

    Charmed,
    I do not regard DPO as a good metric. It was not known when I was at Moto. We used % defective.
    The relationship between % defective and DPU is well known – it was developed by SGS engineers in Utah in about 1982/3.
    However, I am open to another view point …
    Cheers,
    Andy

    0
    #104432

    Mikel
    Member

    Reigle,
    Thank you for you on going concerns about my civility.
    We live in a society where there are many who misrepresent themselves and the truth. It is difficult for me to smile and pretend, especially in light of the bullying from the otherside. Of course we are not supposed to be talking about the state of US politics, so I’ll drop that diversion. I would however suggest Al Franken’s Lies and the Lying Liars who tell them. It is very good in my humble opinion.
    I am trying hard to program myself, but I think I’ll need a lobotomy (or maybe four of them – you think they give group rates?).
    Also thank you for your concern over my “private” communication with Andy U., I felt obligated to let him and others know he was right when he saw through your posts as “Charmed” last week and I came and said he was wrong. Now I know his BS detector is more sensitive than mine.
    As always, love you (in a manly cowboy way).

    0
    #104433

    Charmed
    Participant

    Dear Andy U:
    Again, my sincere thanks.  If I had taken anything personally, I would not have responded to your posts. 
    Having said that, there comes a point where disruptive, rude, and potentially hurtful conduct cannot and must not be tolerated. Schools and colleges and many other institutions will fall apart if we tolerate, or tacitly tolerate, such behavior. I stand by Nick and admire him/her for what he tried to do yesterday.  I some ways we all, and that includes you, must learn to stand by a man when he is right and depart from him when he is wrong.
    You are judged by the company you keep and who your friends are. I am sure you know that. Some conduct that we saw yesterday is unacceptable and I remain disgusted by it.  Today, I have chosen to speak up – but only because of your kind replies. You have done a lot of good with your responses.  But, there is still work to do to weed out the pollutants here.
    If someone offered you a glass of milk, and as you were about to drink it, told you that it has just one drop of poison, would you continue to drink it?  The same goes here.  One drop of poison spoils a whole glass of milk.  Wisdom with statistics must be complimented with other types of wisdom as well.  The best jokes are those where you yourself, not somebody else, is the butt of the joke, etc.
    I will adhere by your well meaning words. But, you too have an obligation to help this forum by dissociating yourself from such folks who stand by you now. With my warmest regards.

    0
    #104437

    Charmed
    Participant

    Dear Andy U:
    Again, again, Andy U, who said we must only follow what was invented at Motorola, whenever or wherever that may be? Please get off that trap. That’s point one.
    Second point.  I do think DPO is a legitimate metric.  There is a world of a difference between DPU and DPO. Let me try this. Let’s say my doctors say that I am a very sick person and have only six months to live. (Lance Armstrong, who won the Tour de France for a record six times was also diagnosed similarly, but he lives!)  So, I am “a” sick man.  That’s like counting one defective.  But, why I am sick? Why did my doctor give me just six months to live? 
    The reasons are: I am obese. I smoke and will not quit. I am diabetic. I have high blood pressure. I have high cholesterol levels. I have a family history, a sibling younger than me is already dead. So, what is going to kill me in six months? We do not know.  There are many opportunities (six here) but I remain just ONE Sick Man, not Six Sick Men.  Doctors call them risk factors. Who knows the future.  Like Lance Armstrong, I might live and be able to cure my illness by doing some other things that nobody does – like pray and meditate regularly.  That’s an opportunity as well.  Armstrong has survived his illness with his passion for biking.
    So DPU and DPO both have their role.  We must be able to tell the difference and use it wisely as we apply and extend the glories of what was once developed at Motorola and spread it far and wide. With my warmest regards.

    0
    #104439

    Sean
    Member

    Charmed Riegle
     
    Charmed = Laxmanan
    ( Also Nobrainjack and Stellar and how many others??)

    0
    #104440

    Mannu Thareja
    Participant

    Hi Charmed,
    Nice explanation…
    Another example can be:
    Case 1 – A General Physician works on a DPU (say) 2 per 100 patients
    Case 2 – A Heart Surgeon works on a DPU (say) 4 per 100 patients.
    From the above two cases, it is very clear from DPU’s that “Physician” is a better doctor…
    But if u consider the Complexities or Risk Factors or Opportunities in both of the doctor’s processes, certainly a Heart sugeon would be having  x100 more risk than a Physician.
    Hence, DPO of Heart Surgeon would be much lesser to the DPO of a General Physician.
    Therefore, DPO helps in nullifying the process complexity.
    Thanks & Regards,
    Mannu
     

    0
    #104441

    Anonymous
    Guest

    Sean,
    Yes .. . Stan was originally right and I was wrong originally. But now I’m pretty sure …
    Cheers,
    Andy
     
     

    0
    #104445

    Peppe
    Participant

    I agree. In electronic assembly have been developed a “complexity index” based on similar things (technology of components, numer of comp, layer, soldering methology, etc…).  Rgs, Peppe

    0
    #104447

    Peppe
    Participant

    Just to return to origin of this post, try to think about Taguchi loss function applied to internal flow (step 2 is a customer of step 1) and “loss for society”  as  “loss for dpt or company”  and let me know what you think about it.     Rgs, Peppe

    0
    #104453

    Anonymous
    Guest

    Mannu,
    Thanks for providing your example. I now have a better understand of the proposition.
    My own view, for what it is worth, is that this is yet another Six Sigma fallacy. The notion that a process with less steps is less susceptible to defects than one with a greater number of steps is not necessarily true.
    Neither is it true for the number of components. While I respect Peppe’s own experience; my experience is to the contrary.
    I could cite many examples, but let’s take one from the company I’ve mentioned already too many times. We were priveleged to compare the circuit design methodologies of both Hitachi, MOS 7 (Japan) and Toshiba. While the American company believed in designing the smallest circuits with the minimum number of components, all these Japanese companies took the view that robustness was worth a few extra components. This gave them a distinct yield advantage over our circuits. (See Phadke’s book Robust Design for other examples.)
    In other words, the most number of steps or the most number of opportunities does not imply a potential for a great number of defects. In fact, the American company I alluded to used two ‘multiple process layers’ to reduce defect density and improve reliability.
    In summary, having more opportunites can actually reduce defect density.
    I’m afraid I cannot comment on your medical model, but it helped explain the ‘concept.’
    Cheers,
    Andy U

    0
    #104454

    Tim F
    Member

    One point that sems to be under-emphasized in all this is that any summary – DPO, DPU, %defective, etc – necessarily leaves out a lot of information.  These numbers are good for executive summaries or daily progess reports, but they do little to know how good a product truly is. 
    A single number doesn’t tell you how serious the defect is.  It doesn’t tell you how hard it would be to fix.  It doesn’t tell you how the defects are distributed.  It doesn’t tell you how much sales you will lose. 
    If you really want to know what is going on, you are going to need a lot more than a single ratio!  No sports fan would try to convince you how good his team is simply by looking at the win/loss ratio, and no QE should try to convince you of the quality of his product based on a single ratio.
    DPO is a useful number; DPU is a useful number; %defective is a useful number – there are potentially dozens of useful numbers to describe a product.  But they are all just a piece of the total understanding of product quality.
    Tim F

    0
    #104455

    Mannu Thareja
    Participant

    Hi Andy,
    Please allow me to give some hunches:

    May be the process that American’s are following has too many opportunities for defect… each step of the process has more than one opportunity of getting an error.Whereas though the number of process steps are more for Japanese, there each process step is “Mistake Proofed” or “Poka-Yoka”.
    May be American’s process is manual  and chances of Human error is more… though the steps are less. Whereas Japenese process is mostly automated.
    Lets take the Medical model that i took…There is a possibility that in a heart surgeon’s career, not even a single patient died as the doctor uses some fandoo equipments at the operation. Whereas in Physician’s case some patient would have suffered due to ill prescription of medicines.
    or.. A heart specialist using latest modern techniques has got a better success rate than his competitor, who uses out dated equipments.Yes, It can be reverse also… :) Depends on the capability! :)
    Thanks & Regards,
    Mannu
     

    0
    #104456

    Savage
    Participant

    Amen.

    0
    #104459

    Anonymous
    Guest

    Mannu,
    The ‘MPL’ layer I mentioned is a reticle that is used twice on the same type of substrate. Since the ‘opportuntiy’ of one layer is already known, then one would expect that if the same reticle is used twice then that would result in twice the number of opportunities and for the number of defects to increase, but this is not the case, which is why the company devised this extra ordinary process in the first place.
    Similarly, the number of steps in a plasma process are often increased in order to reduce defectivity – even though opportunity has been increased by the extra steps.
    Unfortunately, I can’t provide a diagram for an electroncis example, but if you are familar with the idea of variance scaling, then you will appreciate that the addition of one or two transistors (extra opportunities) to scale a voltage level can make a circuit less sensitive to variations in transistor gain, gm. (This is a well known case in robust design.)
    I don’t mean to be argumentative but I’ve already given my reasons for my concerns … one of the great limitations of Six Sigma is that it does not use parameter design. (I’m not claiming that all Six Sigma is bad, but I think it is appropriate to discuss some of these issues .)
    Your pont about poka yoke is well taken …
    Cheers,
    Andy
     

    0
    #104460

    KBailey
    Participant

    Andy U has got the idea. Perhaps I can add something to help clear through the mud on this topic.
    One criteria for metrics validation is, “Does it drive the desired behavior?” DPMO can drive quality improvement in process steps or in particular opportunities. However, DPMO does NOT drive lean in the same way defects per unit/million or defectives per unit/million do. Overemphasizing DPMO rewards the behavior of including unnecessary complexity and opportunities, and it discourages driving process quality for each component to the level necessary for the desired customer experience.
    If I buy a new car, I don’t really give a darn how many opportunities there are. If I have to bring it back to the dealer to be fixed, I’m going to be unhappy – regardless of whether it needs a new windshield wiper blade or a new battery or a new engine and transmission.
    It’s good to keep score and be able to compare the effectiveness of different processes, but what I’ve seen on this thread makes me wonder if some people are focused too much on that, and not enough on delivering a defect-free product or service.
    Other than that, the question relating to the Taguchi Loss Function is interesting. About the only thing I can think to add to that topic is that DPMO can again be misleading, because the cost of a single defect (or even variation within specs) will tend to increase with the number of opportunities or level of complexity of a product due to cascading through the system. It also tends to be more difficult to isolate and correct problems that have more complexity/opportunities.
    kb

    0
    #104480

    Mikel
    Member

    I’m sure.
    The real question is where is he trying to go?

    0
Viewing 41 posts - 1 through 41 (of 41 total)

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