Specification Limits

Six Sigma – iSixSigma Forums Old Forums General Specification Limits

Viewing 4 posts - 1 through 4 (of 4 total)
  • Author
  • #44416


    While this topic may seem better placed in a beginner’s guide, there appears to much misconception in this area. It appears to be the root cause of incorrect statements such as “Customers evaluate quality with reference to defect rates.” and “sigma levels”.
    Specification limits as most know, are “the voice of the customer”. This term is used to contrast them with “the voice of the process” or control limits. Control limits are fixed by the process. However, specification limits are not fixed in this way. Specification limits are set arbitrarily or by choice.
    Most product characteristics vary continuously. However, specification limits are a step function. At one point a characteristic is deemed “good” and at the next increment it is deemed “bad” or defective. The decision for a characteristic to be “OK” at one point then “NOT OK” at a minute increment away is made by choice, not because of any inherent product characteristic. Specification limits are artificial boundaries at which point an arbitrary decision is made to call a product “OK” or “defect”.
    Consider the following situations (Note that “service may be substituted for “product”):

    Company “A” manufacturers a high quality product for company “B” to company “B”’s specifications. Company “A” considers itself a 6 sigma company. Company “B” uses the product in it’s manufacture of other products. Company “B” implements a quality program and decides to reduce variation in it’s raw materials as a means of improving overall quality. It tightens it’s input specifications. Company “A” is now faced with narrower specifications and suddenly finds it is no longer a 6 sigma company despite the fact that it’s process has not changed !
    Company “A” manufacturers a product for company “B” and company “C”. Company “B” and company “C” have different specifications for company “A”’s product. Company “B”’s specifications are tighter than those for company “C” because of different applications for the product. Defect levels will be different for each specification. If defects are used as a measure of quality, which defect level should be used?
    The marketing department of company “A” publishes specifications for it’s product. The company receives complaints about defect levels, so it decides to publish broader specifications to reduce defects. It tells production that it should find ways to improve quality so that it can publish tighter specifications. Marketing, together with management, can set specification levels at whatever level they choose. If defects are used to measure quality, does this make the marketing department responsible for the level of quality?
    It can be seen that in each of these situations, specifications and their corresponding defect levels may change without change in the process. Defect levels are only a reflection of where specification levels happen to have been set. They give no indication of quality.
    In situation “2” above, the problem in using defects as a measure of quality may be compounded if company “A” sorts product for it’s two customers, to meet each customer’s specifications. If company “B” created histograms for it’s input, it would find it to be truncated, indicating that product had been removed. If company “B” recognises that good quality has nothing to do with defects but rather, good quality is “on target with minimum variation”, it might choose a different supplier, whose product histogram was not truncated in this manner.
    Because specification levels are arbitrary and because defects directly relate to specifications, measures of quality based on defects have no value. Suggestions of “4s” or “6s” companies are meaningless.



    Dear Anon,
    I have started to reduce all wastes in LEAN concept since Jan,06.
    Now, I have already succeed in 7 of 8 wastes but the most difficult
    is Defect reduction. How I can start to combine Six Sigma and LEAN
    together to get the best quality during operate machines all times.
    The specification is not the best solution, now. How we prove them
    the less effectiveness  items?
    Thank you very much.



    I’m not Anon, but I can tell you from first hand experience how it’s done in a typical Japanese company. Throw way your books by Womach and read Liker’s ‘The Toyota Way’. Everyone else is doing it. Why not you!
    In the TPS scheme of things Six Sigma equates to process excellence. What this means is the process has a natural ability not to manufacture defects. It achieves this by having Cpk > 2.
    But is this enough? According to one Japanese manager I worked with …  it is not. You also have to have process self-assurance – a kind of insurance – even if you have process excellence. (There are other more complex arguments, but I won’t mention them now.)
    This is why most Japanese companies use ‘single flow’ or ikko nagashi – because this is the only way to ensure ‘one-by-one’ confirmation, or 100% autonomous inspection. (This is not inspection by human beings.)
    Once you set up one-by-one confirmation you’ll generate a pile of waste and you’ll stop the line as soon as you find a defect. In the short term, you may have to ‘disposition’ defective parts on the basis of what the customer feels they can use, or else they’ll go out of busness if you’re a sole supplier.
    You then have to work hard to eliminate all the waste as well as trying to achieve process excellence.
    Good luck,



    Hi Andy,
    Thank you very much.
    Now, we have the problem in our process, the Cpk is 1.3 only.
    We want the process excellent and TPS work well in our product.
    Moreover, M/C are so old and always make minor stoppages at least
    twice per hour and we are not discission to identify these events to
    be downed by operators. It’s just reset and re-touch and let’s them go!
    We do the things per specification and the defects are not reduced yet!, What shall we go next? Step by Step for improve our process
    and machine performance. We’d like to see ZERO defect!
    Note: For LEAN concept, The MCT was reduced from 3.4 days
    to 0.7 day within 3-month. I have realized in theory!
    Thanks, again.

Viewing 4 posts - 1 through 4 (of 4 total)

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