RE: STUDENT REQUEST: Deming & performance appraisals

[WMack <WMack@pec.com>]

> -----Original Message-----
> From:	Templeton, Peter [SMTP:ptemplet@alarismed.com]
> If I understand correctly, from Red Beads and the like, we work on the
> assumption that variation 
> cannot be attributed with confidence to the individual (the basis of PRP).
> Instead, attributing such
> variation to the system, we regard it as being, for our purposes random
> and
> unattributable.
> 
> - I am particularly interested in those that sustain a position above or
> below the average.  Do you think it is valid, as
> I think happens in SPC, to acknowledge that this indicates a trend, or a
> signal that the individual is genuinely at such
> a point in the ranking.  
--------------------------------

I believe this may be a misunderstanding of the control chart.  Here are
three points to consider.

1) The control chart differentiates between special causes and common
causes, not between attributable causes and non-attributable causes.
Common causes are most assuredly attributable to something.  Focussing
on common cause problems and resolving them is the primary idea 
behind continuous improvement.

2) The average is almost a meaningless number for anything besides
a normal distribution.  For example, a process with two parallel assembly
lines may show two peaks for some value, one for each assembly line.  
Neither assembly line may ever produce units with the average value.  The
people on one line will be continually above average and those on the
other, below.

3) The control chart should be used to track the system.  By tracking an
individual on the control chart, you have made the apriori assumption
that the individual is the root cause of some problem.  If you have some
data that validates this assumption, you do not need the control chart
to prove it.  If you don't have the data, you have eliminated all other
explanations for common cause variation by trying to attribute it to
an individual.

If you have a system involving 100 people, it is highly unlikely that making
changes to improve 1 person will affect the overall system measurably.
It is much more effective to make changes that affect all, or a large number
of the people in the system.

I hope this helps.  The key is to focus on the upper and lower control
limits and largely ignore the average.

Wayne Mack
=======================================================================