classroom excersizes for control chart/improvements

Six Sigma – iSixSigma Forums Old Forums General classroom excersizes for control chart/improvements

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
  • #38654


    I teach a intro quality control class. We’ve just finished up the first half of the semester.  I’m trying to set up a lab where the students can generate data to make control charts – and then have them make improvements to their “system” to eliminate assignable cause and then work to reduce common cause error. I’ve been kicking this around for several weeks and can not think of anything that would be fairly easy (within 2 hours) and inexpensive. (No budget!!)
    Anybody have any suggestions or web sites to reference?
    Thanks, Elissa



    I also teach SPC.  What I do usually do is try to draw from the student’s experiences.  Where I teach, students somehow find it hard to arrive on time.  So what I’d do if I were in your shoes is to ask them to list the times the arrived in class for the past 10 meetings.  You can use this data for X bar R charts or for a p chart.  You will have to consolidate this data first, or give it to the class to process for around 15 minutes.  When you examine the data as a class, you would probably find some out of control points.  You can then ask them why there were too many late people, or why people were unusually early during the concerned meeting.  This would be a very good illustration of assignable causes, much more than what Montgomery or Grant could write in their books!
    I hope that it helps.



    Well, we usually conduct a Maryland cookies exercise, which takes about an hour and is much appreciated by our students. If you are not familiar with this exercise I can tell you shortly that all you have to do it to distribute cookies to everybody in the class and ask them to count chocolate chips on the surface. Then plot the numbers, calculate control limits+center line and you have your control chart to work on (make interpretations, etc). We sometimes buy cookies of two different brands (so that the number of chocolate chips would be slightly different) and draw parallels with having two different suppliers of the same material in production.



    Another method you might want to consider is one based in cards and Minitabs random number generator. An advantage of this approach is that you can use the same set of cards for an one of a number of charts; plus you can model assignable and common causes by using some trickery. To demonstrate a shift you would have to keep those cards separate, for example by cutting them down slighly, so that the different size is not immediately obvious but sufficient to sort the cards out for re-randomization.
    It is also important to demonstrate what happens when data is highly correlated and control limits are too tight. This is not the same as a common cause of variation and requires a different kind of corrective action. For me corrective action is key – it is NOT sufficient just to state that a process is out-of-control. The first step is a test of homogeneity, and the second step is to change sampling plans to find out why- even if it means using a Multi-vari chart to identify the cause of variation.
    Finally, will real process behave like a random variate? No! So any form of training is just an approximation of the real world. But I think it is important to teach the correct principles of assignable, special, shifts, etc. Last but not least, an appropriate sample size is at least 30 subgroups. If a process can’t run that long without adjustment or cleaning, then a Shewhart Chart is not the correct tool for the job!
    Good luck,



    I am actually taking an SPC class currently.  My professor seems to like to use M&Ms to do different experiments in class. 
    For example…  Using M&Ms for attribute control charts.  He gave every student a bag of M&Ms and he asked us to open the bag and count the different colors of M&Ms that we had.  He would be around the room and ask everyone what color we pulled out, and whether or not it was a defect (for example if it had the M stamped on correctly, or if there were 2 peanuts in the M&M, etc.)  We would then wait a couple of minutes and then take our next sample.  And so on…  We charted this as we went. 
    The only thing that confused a few people was that we only did one chart and plotted our control limits on the first graph.  As long as you fully explain that this is not the way to do it in industry, then you should be allright.  It worked really well.   took about an hour to go through the whole exercise, plus we got a free bag of M&Ms.  Everyone was pleased and learned first hand how to apply some SPC tools. 


    Tony UK

    Try using a catapult to generate data.  Here’s the weblink:
    To generate special cause data ensure different people fire the catapult and do not let your team hold down the base.  Measure the firing distances with a tape and control chart them.
    Then allow 2 improvements (such as holding the base and fixing the angle of fire).  This will eliminate most of the special cause and the team can concentrate on common cause variation and standardisation.

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

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