Bureaucracy, Efficiency, Learning LO149

From: John Conover <john@email.johncon.com>
Subject: Bureaucracy, Efficiency, Learning LO149
Date: Sat, 18 Feb 95 02:06 PST

Recently, there has been some comments here about Bureaucracies, and
their inefficiencies. In a past life, I made an attempt to use
algorithmic analysis, (the science that computer programmers use to
optimize execution time of a process running in a machine,) to
evaluate organizational efficiency. The rationale was that if
administrations were mechanizations of work flow in an organization,
then the organization's work flow could probably be analyzed by the
principles of algorithmic analysis. What prompted this was that I was
having to turn a dysfunctional engineering organization, (where the
previous management's concept to hurry things along was to "throw
resources at problems.") I was aware of Fred Brooks "Mythical Man
Month," and some comments by Stanislaw Ulam that the management issues
that have to be addressed in an organization grow exponentially with
the number of individuals in the organization. Some of the comments
made here about bureaucracies prompted me to dig out a paper that I
wrote (but never published,) about a decade or so ago on applying
algorithmic principles to organizational analysis. I abandoned the
concept because it did not offer a precise methodology of
organizational metrics since the value of the metric variables had to
be inferred through indirect means, and, in addition, I had to made an
assumption, that worst case, organizational performance would be
exponential (NP) on the complexity of what the organization was
attempting to do-which I felt was a far to stringent constraint.

However, there was some interesting things that came out of the paper.
Starting with a very simple two person "mom and pop shop"
organizational model, the model is extended into a flat organizational
structure, and finally, a hierarchical structure. Some of the
interesting formalities that fall out of the math:

        1) In a hierarchical organizational structure, the optimum
        number of people in each sub-tree of the hierarchy is 6, which
        agrees closely with military doctrine.

        2) If you calculate the magnitude of resources required to
        complete a project by summing the the sub-components in the
        pert chart, you will make an optimistic error of a factor of 2
        in the resources or time required to complete the project,
        which agrees with Brooks and Ulam's observations. (In NP
        problems, in some sense, the sum of the parts is larger than
        the whole.)

        3) Flatter organization structures are superior, except when
        the complexity of what the organization is attempting to do
        passes a certain point, then hierarchical structures are more
        efficient, which seems to be supported by military doctrine.

        4) In flat structures, the economic optimum, and the minimum
        time to solve a problem are coincident solutions-in
        hierarchical structures, they are not-but are close (eg., in
        hierarchical organizational structures, you may do a project
        with minimum cost, or minimum time, but not both-but in flat
        structures you can.)

        4) There exists an economic optimum, for an organization
        solving a problem of a given complexity, and this economic
        function has a law of diminishing returns on the number of
        people that are assigned to solve the problem. The point at
        which adding more resources actually delays the solution of
        the problem might be a mathematical definition of bureaucracy.

        5) When I dug out the paper, (to look at the bureaucratic
        content,) I was somewhat astonished that the best methodology
        of increasing organizational performance was through learning
        of skill sets-which, oddly enough, completely dominated
        technical, structural, and organizational methodology, at
        least in large organizations. (It is the only variable which
        has a linear effect on the complexity of the problem-all of
        the others have decreasing returns.)

I was fascinated by that-even though the organizational models are
probably too simple to be of any practical or quantitative value, it
would seem that their may some qualitative significance.  The LaTeX
sources from the beginning of the paper are attached. There is a lot
of optimization calculus (LaTeX supports math symbols very well,)
which I can not ever remember proofing.  But if you want (eg., you
have insomnia,) you can print it, if not file it in /dev/null.

For what its worth ...



John Conover, john@email.johncon.com, http://www.johncon.com/

From: john@email.johncon.com (John Conover)

Host's Note: John included the full text of his paper in the message, but
I've taken the liberty of deleting it. If you'd like the full text in, I
suggest you mail John directly.

  -- Rick Karash, rkarash@world.std.com, host for learning-org

Copyright © 1995 John Conover, john@email.johncon.com. All Rights Reserved.
Last modified: Fri Mar 26 18:57:41 PST 1999 $Id: 950220110730.9891.html,v 1.0 2001/11/17 23:05:50 conover Exp $
Valid HTML 4.0!