iSixSigma

Special Causes?

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
    Posts
  • #246729

    Ale
    Participant

    Hello everyone,

    I have a question about using run charts to find special causes (trend,  cluster, …). I have a dataset that collects the orders of goods released by a company to its suppliers. Every day, several orders are made to different suppliers, who deliver the goods: on time, in advance or late (even for many days). The specification limits are +/- 5 days from the scheduled delivery date.

    The CTQ used is the number of orders delivered within LSL and USL divided by to the total of orders issued on the same day, in this way I have a percentage variable that changes daily (% On-Time).

    Is it correct use a run chart to find out special cause related to the time (image attached)? Or Should I just use the number of orders within the specification limits that vary from day to day?

    Thank in advance

    Regards

    Ale

     

    Attachments:
    1. You must be signed in to download files.
    0
    #246780

    Strayer
    Participant

    It sounds like you may be conflating specification limits with control limits, which is a common misunderstanding.  A control chart tells you whether or not the process is statistically under control, but says nothing by itself about whether or not the variation is within specification limits.  The UCL and LCL are calculated at +/- three sigma from the mean.  You can find free templates for control charts.  Plug in the data set from your run chart and use the rules for identifying variations of special cause.

    p.s. I like to add the USL and LSL lines since this tells me whether or not variation of common cause is within spec limits and whether or not the mean is centered.  I’ve had arguments with other practitioners about doing this.  They say it’s a distracting complication.  I say it gives me a lot more information in one chart.

    0
    #246784

    Jam Haron
    Participant

    You can use run chart and address the outliers first.  Focus on making delivery within control limits. That will make process more capable to meet the specification.

    Once you have eliminated outliers (with reference to control limits), then implement changes to improve delivery consistently within specification limits.

    0
    #246829

    Strayer
    Participant

    Once again.  Control limits are calculated at +/- three sigma from the mean.  A run chart will identify outliers but says nothing about control limits.  We do statistical process control because with an ill-controlled process trying to eliminate outliers is likely to lead us down the garden path.

    0
    #246985

    Ale
    Participant

    Thanks a lot for your advices @Straydog and @JamHaron!

     

    0
    #246986

    Fausto Galetto
    Participant

    @Alessione

    @Straydog

    @JamHaron

    Control limits are calculated at +/- three sigma from the mean.

    makes NO sense, for Exponential or Weibull distributed data!!!!

    SEE the ERRORS in T Charts (from MINITAB)

    0
Viewing 6 posts - 1 through 6 (of 6 total)

You must be logged in to reply to this topic.