Home ] Up ] Next ]


Origin of Schneiderman's Half-life Method


Arthur M. Schneiderman

In 1984, on a flight returning from a study mission to Japan, I made a remarkable discovery.  Like many such discoveries, this one was also serendipitous.  I was reviewing materials from a presentation given by Kenzo Sasaoka, president of Yokagowa Hewlett-Packard.  One display showed the multi-year efforts of a quality circle working on the reduction of soldering defects in the manufacture of printed circuit boards:  

Starting at a defect level of 0.4 percent, which corresponded on average to two defects on each board produced, the team systematically reduced the defect level to near zero, or so it appeared.  Each of the "pointing fingers" represented the implementation of corrective action resulting from one turn of the PDCA cycle.  Recognizing that their near perfection was a graphical illusion, the team generated the second graph which now measured the defect level in the smaller units of parts per million, or ppm (1 ppm = .0001% defects).  The incremental improvement trend continued.  Over a period of five years, the team reduced the defect level by a factor of more than 10,000.  Now instead of having to rework every board, only one in 600 required this costly and non-value adding step.

Interesting data, but how does it help reduce the boredom of a 15 hour flight?  Well, I'm an engineer, so faced with those two graphs I recognized that they could be combined on a single graph if a logarithmic scale was used for the defect level.  But I'm also a manager, and no longer travel with semi-log graph paper in my briefcase.  Therein lay my first challenge.  The next few hours were spent creating semi-log paper using my calculator and a ruler and re-graphing the YHP data.  This was the result:

The most remarkable aspect of this new graph is that the first three years of data lie on a very straight line.  To an engineer, this instantly signals that the rate of improvement is constant.  Radioactive decay has the same behavior.  We characterize it by the constant amount of time it takes for the radiation to drop by 50% and call this the half-life.  Using that analogy, I calculated the "half-life" for the printed circuit board defect rate, it was 3.6 months!

In other words, starting at a defect level of .4%, after 3.6 months, it was down to .2%, in another 3.6 months it was down to .1%, in another, .05%, and on and on.  In fact, the team went through more than 7 halvings of the defect level before something happened and the rate of improvement slowed dramatically to a new and constant half-life of around 8 months.  What happened?  At a subsequent visit YHP, I shared this graph with Mr. Sasaoka and his vice president of manufacturing.  After much discussion, we discovered that around the time that the improvement rate slowed, the team discovered that most of the root causes of the remaining defects lay not with manufacturing but with the design of the boards themselves.  The team needed to be expanded to include the design function, thus increasing the organizational complexity of the process.  By the way, on my last visit to YHP, they were measuring defect levels in ppb or "parts per billion."

With this discovery, the remainder of the flight went quickly.  I included this finding in my first Quality Progress article, and  started researching other exemplary continuous improvement results and published my findings in the second Quality Progress article.

back to top

1999-2006, Arthur M. Schneiderman  All Rights Reserved

Last modified: August 13, 2006