iSixSigma

Another control chart question

Six Sigma – iSixSigma Forums Old Forums General Another control chart question

Viewing 7 posts - 1 through 7 (of 7 total)
  • Author
    Posts
  • #52425

    TJ
    Member

    I am measuring total population (system data for 10,000 units per week), and the data is discreet.
    I am struggling to work out how to represent this in a control chart – which one should I use?
    Help!

    0
    #184455

    Gandalf
    Participant

    TJ,
    The most common answer to any Six Sigma question is “It Depends”…
    You said your data is discrete, but is it categorical, pass/fail, ordinal, … ?  What is the widget and what characteristic are you measuring?  Are you measuring or counting defects, or defectives?
    Let us know and we can give you a good answer…
     

    0
    #184460

    TJ
    Member

    Thanks,
    I work in Financial Services, not manufacturing, and we are measuring First Call Resolution. Either the call is resolved, or it isn’t – this is measured on a system with a marker to say if the customer query has been resolved on that call or not. There are 3 markers, ‘A’ and ‘R’ represent a resolution and ‘I’ represents a call no resolved and therefore a defect.
    So we report from the system, and can track a sigma score easily enough, and now we want to get some SPC in place.
    As we get the data from the system, we are reporting on total population, not a sample. So my question is really around the best way to display this for our process owners and us as Black Belts to identify special cause etc.

    0
    #184461

    Gandalf
    Participant

    Since you are counting “defective” calls and the number of calls will vary from day to day, you should use a “p-chart”.  Attribute SPC charts like this only have half of the tests that variable charts do, but you’ll have to just recognize that the warning capabilities are less…
    I would encourage you to track employee id, time-of-day, day-of-week, length-of-call, and issue group in parallel to the ‘A’, ‘R’, and ‘I’ result.  If the SPC chart ever shows that something has “changed” (which is all it’s supposed to do anyway), you’ll then have additional data to help identify root causes.  You can also perform additional analysis to see which of those characteristics correlate and perhaps make proactive process improvements.
    Good Luck

    0
    #184469

    Sloan
    Participant

    TJ,
    I’m a little concerned that you don’t know this since you used the phrase, “…us as Black Belts…” implying that you are yourself a Black Belt. This question about which control chart to use is Green Belt 101. As a BB you should be able to zero right in on either a P-chart or an NP-chart (or under certain circumstances an I/MR-chart). At the very least I would hope that if you need to ask a question, it would be to get some clarification about which of these charts would best to use under the circumstances that you have rather than a totally open ended question.
    I work in the Financial Services industry too so that should not be an excuse.
    I hope you don’t think my criticism is too harsh, but this isn’t the type of question I would expect from a BB.
    Sorry for the rant.
    Outlier

    0
    #184471

    TJ
    Member

    Outlier,
    I didn’t ask for your opinion on me, I would hope that as part of this forum I could get some advice, not be told that you feel I should know this.
    I am a BB, and yes, I do know which control chart to use for discreet defective data. I’m not using Financial Services as an excuse, it was just a passing comment.
    I do appreciate however that I should be more specific in my question, so as not to appear like I don’t know which control chart to use for discreet data! I’ve worked it out for myself now anyway – my question was meant to ask how practically I could display this for process owners, and yes I should have been more specific in my questioning.
    Thanks Gandalf for good advice, thanks Outlier for putting me in my place. You’re the man.

    0
    #184496

    Darth
    Participant

    TJ,
    I know I am jumping in a little late but I would be curious what you figured out on your own. You indicated that you are using population information to do SPC. That cannot be a true statement. A control chart measures subgroups over time. Please clarify. If you are using the entire day’s worth of call resolution information, then that is still a sample not a population. Furthermore, if you are using a p chart as was suggested for a large n the chart is going to look really strange if you do it correctly. How about sharing with us.

    0
Viewing 7 posts - 1 through 7 (of 7 total)

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