A question to Reigel Stewart

Six Sigma – iSixSigma Forums Old Forums General A question to Reigel Stewart

Viewing 17 posts - 1 through 17 (of 17 total)
  • Author
  • #35396


    Reigel Stewart:
    I would like to ask you a question since based on your notes it appears that you know Mikel Harry very well.  In Mikel Harry’s Six Sigma Book and also in Six Sigma Academy’s training materials, there is a picture of an apple tree and I believe the title says something like  Fruit Of Six Sigma (I don’t have the book in front of me). Anyway, the picture says that
    3 sigma: Beat Up Suppliers
    4 sigma: Improve Process
    5 sigma: Improve Design
    6 sigma; DFSS
    Does this mean that process improvement will get you to 4 sigma at best, and that to get to process capability of 5 or 6 sigma, design needs to be improved or changed? Do I understand this correctly? If this is true, isn’t there a conflict here since DMAIC is a process improvement approach, and we teach DMAIC under Six Sigma process but using DMAIC we can never achieve Six Sigma?
    I appreciate your answer.


    Ken Feldman

    Good question John.  I have seen that OH for years.  The explanation always given to me is that you can maybe hope to achieve 4s with DMAIC but after that you reach process entitlement and you need something more dramatic like DMADV to get the remaining improvement.  Since DMADV/DFSS is also a tool under the SS umbrella it has always been presented as a step rather than an incline.  Look forward to Reigle’s explanation.



    How come it’s some sort of revelation that 80% or more of quality, reliability, efficiency, safety etc etc is designed in ?DMADV/DFSS/DMAIC and so on are only common sense written down and their benefit is to allow common sense to be taught. The benefit of that is apparently that common sense is not so common.
    If you have a badly designed anything there is only so far you can improve it. Maybe it’s so badly designed you can’t even get it to 3 sigma. Equally you can have a perfectly functional design capable of 6 sigma that doesn’t achieve that because of how it is used. Take a jet engine the military flies it at a guess to 80% of it’s design limit the same engine on a commercial jet maybe operates at 30% of its operationl limit. Commercial jet engines can demonstrate a 6S reliability in part because of how they are used but the fact is they are a very good design in the first place.
    Strikes me that the apple tree is a good selling gimmick not meant for detailed criticism.


    Ken Feldman

    Ah, Grasshopper, a picture may be worth a thousand words….or a thousand bucks if you are a consultant.  You are correct, the apple tree makes for a simple graphic and not for indepth intellectual discussion.  But, let’s withhold judgement until we hear from the expert himself Johnny AppleStewart.



    Such a tree is ludicrous.
    I’ve witnessed teams achieving 6-sigma levels by doing any combination of the actions mentioned. And the same is true for teams that never achieved 3-sigma.
    Actually, the whole sigma level discussion is filled with assumptions and inconsistencies that detract from the commom sense approach of measuring improvement before and after. Is that just too simple?
    To Mia’s point, the Chinese already had a proverb for this: “Common sense is the least common of the senses”


    Reigle Stewart

    The tree to which you refer is a generalization, not a
    “hard” principle of application. Dr. Harry fully recognizes
    and confirms there are different tools for different
    purposes, but in general, there is a “flow” to the tools
    when one looks at the big picture. The general flow is
    from basic quality tools (low hanging fruit) to DFSS (sweet
    fruit). Again, the tree analogy helps the novice grasp the
    general heirarchy of Six Sigma, but from 10,000 feet
    (enterprise level). Such a high level view generally says
    that the application processing-for-six-sigma (PFSS)
    tools, you will get you to about 5 sigma (overall). At this
    point, you often hit the limit of collective entitlement
    capability (the average capability of the processes that
    define an enterprise). To surmount the “5 sigma wall,” we
    must redesign the processes (and perhaps the products)
    to progress on to 6 sigma. Remember, we are talking the
    “pooled” capability of an enterprise, not individual CTQs
    … people often miss this idea … many individual CTQs
    can be brought to a 6 sigma level of capability using only
    process improvement methods, but the tree you refer to is
    “wholistic” in nature. Respectfully, Reigle Stewart



    I appreciate your reply. I understand that that this is a high level view. But I am still unclear about this. You wrote, “we are talking the “pooled” capability of an enterprise, not individual CTQs … people often miss this idea … many individual CTQs can be brought to a 6 sigma level of capability using only process improvement methods, but the tree you refer to is “wholistic in nature”
    Are you saying that process improvement by itself will NOT get you to six sigma level at an enterprise level but at an individual CTQ level process improvement may get you there? The picture itself does not say anything regarding enterprise level versus CTQ level. It actually gives you the impression that for more improvement, you need to go to DFSS. Thank you.


    Reigle Stewart

    John: You are correct. At the enterprise level, it is very
    doubtful that process improvements alone will get an
    organization to a Six Sigma level of performance (3.4
    defects per opportunity, on average). Usually, when the
    process improvement efforts have yielded a 4.5 to 5.0
    sigma level of quality, the business enterprise hits a “wall”
    and then stagnates. This happened at Motorola. In fact,
    George Fisher (then CEO) made a tape about this
    phenomenon (1991). At this point, the emphasis must be
    placed on redesigning the system, not tweeking the
    existing processes. Remember, the “tree” you refer to is a
    slide, not a thesis on the subject. Take it for what it is … a
    high level generalization. Reigle Stewart



    I have a question too.
    Why would I bother (and waste a lot of effort and resources and time and money and…) improving a thing (system) that I know forehand I will throw out and replace by a new redesigned thing?
    Why not to go stright to the redesign stage?



    One word…Cost. Redesign may involve significant costs that may offset the savings associated with the improvements therefore causing the redesign to not be cost effective.  Process improvement actions must also be economically justified as well.



    Of course. What I was questioning was whether it was logical to invest in improving the system to reach 4.5 sigmas IF you will then invest again in a new design of the system to reach 6 sigmas. It will be investing two times to get nothing better that if you went directly to redesign, then saving the investment in improving the existing system. If investing in redesign was not economically justified, then investing in improvement and then in redesign could only be less economically justified.



    I think this is a valid point. What I should start with is the definition of state that I want to reach (and its associated business case of course). If I can reach it by tweaking my existing process then I should do that. if I have to redesign my process to achieve what I want to achieve then there is little point wasting time tweaking my existing one.
    The problem is that if the definition and benefit of the desired state are not known then process tweaking adds ever increasing and uncontrolled cost. Someone has to be able to call the tune.


    Reigle Stewart

    Rog: As you may have guessed, the ability to “call the
    tune” comes with experience. Very generally, when
    is high, but is low, you have a control problem, not a
    technology problem (given that the aims of rational
    subgrouping have been fully satisfied). On the other
    hand, if is low and Z.shift is also low, you are in
    control of poor technology. Such reasoning may assist
    you in your efforts. I strongly recommend you set up a
    three variable full factorial experiment to study these
    conditions. Factor A =, set at a low and high level.
    Factor B = Z.shift, set at a high and low level. Factor C =
    Number of such “opportunities,” also set at a high and low
    level. The response variable is Y = DPU = -ln|( –
    Z.shift)^C|, where C is the number of total opportunities
    subject to said capability. Taking the natural log provides
    the expected DPU. Thus we DISCOVER first-hand
    several things. For example, we learn that when
    capability is low, the influence of shift and complexity is
    high, but when capability is high, the system is robust to
    shift and complexity. The interactions are also interesting.
    Dr. Harry first showed me this experiment several years
    ago and how to use such an approach to further develop
    a deeper understanding of the fundamental equations
    governing process capability, DPU, and so on. Once you
    gain some experience with this approach, you can
    include other factors such as cost, design tolerances,
    reliability factors, and on and on and on. Respectfully,
    Reigle Stewart


    Reigle Stewart

    Rog: I forgot to mention that = – Z.shift and
    must then be converted to “yield” by way of the Excel
    equation: = normsdist( Then the resultant yield is
    cross multiplied over all of the C number of opportunities.
    Reigle Stewart


    Jim Cotten

    Reigel – you may find that you need to make design improvements in order to make processes more efficient (engineers and mechanics working together).So you have reached level 5 already while at the same time acheiving level 4 process improvements through some “reverse engineering”. Just a thought….
    James Cotten


    Kim Niles

    Dear John, et al.:
    This is a very interesting conversation to me as not much is written on the 5 sigma wall and or obtaining very high sigma levels. 
    Since any process sigma can also be considered to depend upon how close the outer edges of the process distribution (what is) comes to the specs (what should be), achieving high levels of sigma must depend upon the following factors I envision as follows:
    1-     Product or service complexity (process improvement flexibility),
    2-     Process improvement resources (realistic ability to improve processes),
    3-     Customer and market demands (design flexibility),
    4-     The company’s design resources (realistic ability to design in robustness).
    I welcome other thoughts along these lines but suggest that if a company wants to achieve high levels of sigma, they must first fully address these four factors. 



    Dear All,
    I agree with KIM Niles’ wholistic business view. We need a structured approach and possibly a specific model / methodology on similar lines. 
    Innovation and its need in today’s fast paced changes in customer preferences & business environment & IT etc., has added muliple complexities in an organizations decisionmaking machinery as regards choice of a long-term quality policy/methodology/model.   
    Dr Mikle Harry speaks of the MFSS (Management For SS), which, I believe, wud comprehensively address the whole question of propriety of spend on 6 sigma and which model and what level is desirable and at which point do we start thinking of DFFS  etc., from the business strategy and customer rqrmt point of view.
    Any thoughts ?? – any structured literature on MFSS and how it encompasses DFSS & PFSS, etc., I wud be interested to learn more about how to work on the choice of 6 Sigma methodologies looking to business strategy and customer rqrmt (which is environmental and non-controllable factors). What is MFSS all about ?

Viewing 17 posts - 1 through 17 (of 17 total)

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