iSixSigma

Percentages vs. Sigma?

Six Sigma – iSixSigma › Forums › Old Forums › General › Percentages vs. Sigma?

Viewing 10 posts - 1 through 10 (of 10 total)
  • Author
    Posts
  • September 21, 2004 at 3:15 am #36934

    batman
    ★ 10 Years ★
    Participant

    I’m reading a lot of this Six Sigma concept and I don’t understand why couldn’t we just use percentages?
    For example:
    Day 1: 36% defects
    Day 2: 23% defects
    Day 3: 22% defects
    Day 4: 10% defects
    Day 5: etc….
    I can calculate the standard deviation / sigma of the process, but why couldn’t I just use the percentages for analysis?
    Can you tell that I am new to this?

    0
    September 21, 2004 at 10:04 am #107695

    Arun Kumar
    ★ 10 Years ★
    Participant

    Hi,
    Your question is very basic and hence v imp. First of all, you cannot have % defects as we normally mistake. It is % defective. So, if you correct to % defective, then you are not able to compare processes, because thediffer in their complexity and nature (something which is represented in the number of opportunities). This is necessary to create a level playing field- making the sigma scale a universal quality metric.
    However, if you measure on the sigma scale, you can always convert to % defects, if thats what you want.
    Another strong reason is the business case of six sigma. % do not correlate with cost of quality, whereas sigma scale (which is based on number of defects and not defectives) correlates well.
    You can say, while % defective is ok, the degree of damage to the product or serive is in this way measured.
    kind regards,
    arun kumar

    0
    September 21, 2004 at 10:43 am #107696

    vineeth
    ★ 10 Years ★
    Member

    As Arun very well Mentioned..% vs Sigma..
    its the Same difference as Yield  vs DPU ( defect per Unit ) and hence DPMO to See a long term Performance..
    well For the Customer in concerned, it makes no difference , if a Unit have 1 defect or n Defects…Its  a Defective unit for the Customer..
    and for the Supplier is concerned, it makes a Difference, The Probability of a  Defective unit is more when the  DPU is high..
    How come Cost of Quality is Linked with  DPU..Infact Yield is also the same…???
     
     

    0
    September 21, 2004 at 10:56 am #107697

    Arun Kumar
    ★ 10 Years ★
    Participant

    Hi Vineet,
    I don’t know whether your last line is a question, but I will love to err on the right side here.
    I am quoting Mikel Harry (further substantiated by Crosby or vice versa).
    308537 as dpmo – >40% as sales as COPQ
    66807 as dpmo – 25-40% sales as COPQ
    6210 as dpmo – 15-25% sales as COPQ
    233 as dpmo – 5-15% sales as COPQ
    3.4 as dpmo – <1% sales as COPQ
    having said this, you will realize that the cost incurred in correcting defects is the actual cost. With % going down, there is no gurantee on the defects, but the other way round is surely true. Hence COPQ is impacted by defects and the sensitivity is more.
    regards,
    arun kumar

    0
    September 21, 2004 at 2:32 pm #107706

    KBailey
    ★ 10 Years ★
    Participant

    Three other reasons:
    First: Sigma level can be calculated from a relatively small number of data points. Once you get past 3 sigma and get your defect rate < 1%, it becomes much more expensive to get enough data points to measure quality by actual % defective.
    Second: Taguchi loss function says that variation adds cost even if you’re within customer specs. Sigma level and Cp measurements give you an idea how much variation you have within specs – where most of your data points lie – instead of just the ones that fall outside specs.
    Third: Many people see 99%, 99.9%, or 99.99% quality as being pretty good. They just don’t naturally see a big difference between 1%, .1%, 0.1% defect rates, etc. Especially when it comes to processes with many steps and/or products with many components, it’s sometimes harder to get people motivated to improve when they’re looking at a number that they normall think of as “pretty good.”
     

    0
    September 21, 2004 at 10:33 pm #107729

    batman
    ★ 10 Years ★
    Participant

    Wow! Pretty enlightening reasons.
    I especially like your last one.
    Thanks!

    0
    September 27, 2004 at 1:25 pm #108046

    ssp_128
    ★ 10 Years ★
    Member

    Percentages are good but sigma value are better.  We want to know how far we are away from the variation.  36% away from the target tells me that I have a lot of work to but does not tell me if this number is a reflection on the output of the process.  Basically, 36% does not tell me if the process is working or not.  Data also must be normalized as you’ll see in the pareto analysis.
     

    0
    September 27, 2004 at 5:48 pm #108078

    Moore
    ★ 10 Years ★
    Participant

    The third reason is very psychological, but a good one.  In the classes I have taught I frequently explain that the sigma scale is much more sensitive to small gains, therefore making it easier to show there actually was an improvement.
    For example, would it be more clear that you had a real improvement to tell your boss you went from 99.9% quality to 99.99% quality OR to tel him you went from 4.6 sigma to 5.2 sigma?  I think 0.6 improvement on a 6 point scale (yes, I realize you can theoretically exceed the “6 point scale”, but for purposes here…) indicates the improvement better than 0.9% on a 100% scale.
    The other reasons given by people are also correct….just thought I would add to this one.

    0
    September 27, 2004 at 8:39 pm #108085

    mammasaya
    ★ 10 Years ★
    Participant

    I was in a tollgate review of a project in the Define phase, and made the suggestion that the belt state the process Sigma level in their problem statement.  A colleague very strongly pointed out that this “could not be done” in the Define phase, as process capability is addressed in the Measure phase.  I was aghast and amazed … would you have advised in this way?

    0
    September 28, 2004 at 8:35 am #108108

    Arun Kumar
    ★ 10 Years ★
    Participant

    Hi,
    It depends on the level of maturity of the organization. For beginer’s stating sigma level in define could be an issue of methodlogy and quite rightly so. For a mature BB/GB, what you have suggested is absolutely fine.
    Regards,
    arun 

    0
  • Author
    Posts
Viewing 10 posts - 1 through 10 (of 10 total)

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