iSixSigma

How do I calculate the DPMO on documents?

Six Sigma – iSixSigma Forums Old Forums General How do I calculate the DPMO on documents?

Viewing 13 posts - 1 through 13 (of 13 total)
  • Author
    Posts
  • #49187

    Mark Leach
    Participant

    I need to determine the potential erros for documents. We have documents that are followed by individuals, and in some cases data is entered onto them. In order to improve the quality of our processes we are looking to improve our documents and therefroe reduce the error potential, but I need some way of measuring this. Any helpful advice would be appreciated?
    Mark.

    0
    #167680

    Cordell Fife
    Participant

    You could do a Gage study.  This would give you the repeatability etc from Gage analysis and also zero in on the specifics areas of the form that are driving errors.

    0
    #167682

    Haugen
    Participant

    Mark,Determine what is important to the customer, and then measure that. For example, in our document process, numbers are very important (address, phone, SS#, etc), and the order as well as the proper digit are important, but we only count the whole number as a defect – not each digit, because the root cause of the issue is all the same. Proper lighting, correct posture, legible copy, etc (manual data entry). does not do the process or the operator any good to count 3 defects in one address when he/she just fat-fingered the address because they were trying to type too fast.

    0
    #167718

    SiggySig
    Member

    Mark,In the strictest sense of DPMO calculations, each field on a document being updated would constitute an opportunity for a defect. I agree with JimH that you probably wouldn’t want to capture multiple defects per field. That seems like it could be a data collection nightmare.To calculate DPMO, you would need a count of total defective *fields*, total fields, and simply take the % * 1,000,000 to get DPMO. That would also be a great metric to track the improvement of the process over time.Where I’ve seen this get tricky is when not all fields are being touched each time a document is processed.

    0
    #167725

    MrMHead
    Participant

    There is also the discussion of Defects vs Defective.
    You may have 15 opportunities for defects on a document, but does just 1 defect render the document “Defective”?
     

    0
    #167728

    SiggySig
    Member

    Excellent point – in most cases, if anything is incorrect on a form, it would be considered defective, and would have to be reworked.I’d still go with DPMO so that I could understand whether the process is improving or not. Because 10 wrong fields would be worse than 1 wrong field, and simply counting defectives would not give you insight into this.

    0
    #167729

    JimmyJam
    Participant

    Great points..and can I add another question, and forgive me if it is a stupid one. 
    For discrete capability calculations….total number of units, defect opportunity per unit and total defects……if I do have 15 opportunities per unit….should I be using 15 or 1 (if just one defect renders the doc defective?)…I am thinking 1….because if I use the 15, I get a higher sigma, but that is not truly reflective of my process and the “defective” issue..
    Can someone clarify….

    0
    #167730

    JimmyJam
    Participant

    Siggy…
    Understand your point…but for my clarification…..wouldnt it be better to count it as a defective unit, as opposed to defects per unit…because, I may fix one defect out of 10…but the entire document is still defective, and has to be fixed. 

    0
    #167731

    MrMHead
    Participant

    For DPMO, you would count all 15 opportunities per unit.
    Once you start counting Defective Units, then you are calculating Yeild (which should take re-work into account.)
    Ignorance can be fixed, Stupidity is Forever!
    (aka, there are no stupid questions  ;)

    0
    #167735

    JimmyJam
    Participant

    Thank you for the help…so in my case I should be counting each as a defective.   And in that case…is it still safe to use the discrete capability calc…for sigma…it will be a sigma on my yield? 
    Another dumb question…..I heard you would use discrete calc for nonnormal data (if you do not transform).  Is this true / okay (use data points above the upper spec as a defect, 1 defect opportunity per unit). 

    0
    #167737

    SiggySig
    Member

    Jam,In this case, I would suggest using a straightforward dpmo calculation to get process capability. Here’s one: https://www.isixsigma.com/sixsigma/six_sigma_calculator.asp?m=advancedYou aren’t counting defective documents here, but rather defects per opportunity. This is a critical distinction, because if you move from 15 defects per document to 1 defect per document, your process will have improved, but if you only count defective documents, you will not see this improvement.To answer your other question, I have heard different opinions on that. Personally, I have done process capability in Minitab with non-normal data and applied the appropriate distribution to get my DPMO. You can also transform, but I try not to transform data if I can avoid it. If I select the right distribution using the non-normal option, I get the same answer as when I transform the data.

    0
    #167759

    JimmyJam
    Participant

    Thanks Siggy for your input..!! 
    What are your thoughts on……if I am trying to improve the yield….does it make more sense to use defective rather than defect per unit?  To your point, I may decrease the defects…but if even one defect renders the unit defective and it has to stop in the process……I may have improved my DPMO but not my yield. 
    Thanks again …and am interested in your input…!!

    0
    #167779

    annon
    Participant

    The point is to capture both, using both internal (failures occuring before shipment/sending) and external failures (failues occuring after shipment) to assess the quality of the process output. In this case, using DPMO (or better still – DPU) to baseline and assess internal quality efforts as well as a yield measure for external assessments.
    As to process capability measures, using a DPMO or DPU assessment avoids the requirement of normality and would be appropriate here, noting DPU is often easier to calculate and explain, as well as being more sensitive to change….it is also a snap to calculate RTY (ie inverse natural log of DPU) from it if desired.

    0
Viewing 13 posts - 1 through 13 (of 13 total)

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