Judgment, Evaluation, Feedback, etc. LO9968

Sat, 14 Sep 1996 14:39:34 -0400

Replying to LO9894 --

Ref Michael Ayers question:

> At one point, Deming does seem to say that "individual
> performance cannot be separated from the performance of
> the system" -- with the critical caveat that -the system is
> in statistical control-. That is, if the system's performance
> shows no variation beyond what an enlightened observer
> would expect, then the variation -within- that range of
> performance cannot legitimately be ascribed to an individual.

> OTOH, in his list of management's Five Deadly Diseases,
> his specifically mentions annual performance ratings as
> one of those deadly diseases. Elsewhere in the book,
> however, he mentions talking to the people whose
> performance does fall outside that anticipated range --
> so that under-performers can receive training or transfer,
> and so that over-performers can can help improve the system
> to bring everyone up to their levels.

My interpretation of Demings comments on this was that, just as with
systems under statistical process control, you cannot meaningfully react
to a great deal of variation around an "average" performance because it is
random. Thus, the majority of employees, who's performance is presumably
clustered around the average (BIG presumption I know...) cannot be
correctly differentiated with regard to ranking and rating schemes, salary
rewards, etc.

On the other hand, some people will be so far from the norm, either
positively or negatively, that you can be reasonably sure (although never
positive) that their exceptionally high/low performance is not due to

To me, the implications of this include:

- a very simple rating system with three classes (Exceptional, Standard,
and poor)

- group based rewards for everyone in the standard rating (which should be
80-90%+ of the workforce). This seems natural to me if you want to reward
performance but cannot distinguish between individual performance levels

- immediate action on the "outliers", those with very good or poor
performance. If a machine goes out of process control, you take action
immediately; you do not wait for a periodic meeting with the maintenance
man... If someone is doing exceptionally badly, they probably need some
education (another of Deming's principles) on how to do better

Similarly, if someone is exceptionally good, they should be rewarded
for it AND their techniques/skills incorporated into the education system
to shift the performance of the entire system to a new and higher level.
I would guess that Deming would be against a yearly reviews because
of their periodicity (Process control takes action whenever there is an
outlier, not based on the calendar) and because, as usually implemented,
they result in grading people into numerous levels which are not
objective, not significant (except in extreme cases), and too removed from
the performance for education to take effect.

Personally, I like Deming's thoughts in this area very much, but it is
always worth noting that SPC depends on a stable process that is both
repeatable and under control. Statistical measures need to be watched
very carefully if the underlying process (tasks, methods, materials etc.)
are changing and are not particularly meaningful for tasks that are not
performed repeatedly.

On this general topic: I would also recommend looking at Steven Covey's
material on Performance Agreements. He has a very interesting approach
that is oriented toward having people do self evaluation.

Robert L.



Learning-org -- An Internet Dialog on Learning Organizations For info: <rkarash@karash.com> -or- <http://world.std.com/~lo/>