The % of red beads

["John McConnell" <wysowl@msn.com.au>]

Raul L. Butland wrote:
Alan's concern is a reflection of the improvement our society has achieved
since Dr. Deming's idea for the red bead experiment. Back then, the defect
ratio was much higher. Now it is much lower. A change in the number of red
beads may be in order.  I am no longer in the management field, but I would
be interested in hearing from some of you "actives" what the proper ratio
should be in order to reflect today's lower rate of defects.
**********************************************

John replies:
I fear we might be missing the point.  What good would a Red Beads
experiment be if the defect (red) rate was (say) 5% ?  At 5% red and 50
beads per paddle, the average would be roughly 2.5 red per paddle.  Many
workers would score zero for a given batch.  Often, a worker's total score
for the experiment would be tiny.  Is this what we want?  Or perhaps we
could use a proportion red of 3:1,000.  In such cases rarely would a sample
of about 50 have any defects and the experiment would fail.  This is true of
attributes data in factories, let alone an experiment.

If we continually reduce the proportion of red beads the distribution would
change shape until the mode reached zero.  For beginners, are we not better
served to avoid these issues?  I have conducted these experiments with low
proportions red and consequential non-normal distributions for advanced
students.  They were fun, and brought their own lessons, but I would never
use them for beginners because data that are actually random can look very
much like the result of special causes.  

The experiment should teach us about stability; should illustrate a system
making errors or rejects; should be demonstrably stable and easy for
students to "see"; should make willing workers feel frustrated and should be
a vehicle to lead into SoPK.  It does that as it stands.  I'm all for
improvement providing the aim is not compromised.  But to significantly
reduce the proportion red seems to have fewer advantages than disadvantages
and stands in danger of compromising the aim.

A final thought.  Do any of us believe that this question completely avoided
Deming?  If we assume not, we are then led to ask why he chose the
proportions he did.  Deming used 20% red.  How long has it been since
manufacturers routinely produced this type of reject rate?  I doubt the
proportion red Deming chose was ever meant to be representative of industry
or commerce.  He had other reasons.

Cheerio!

John McConnell