leverage points

[Jay Forrest]

At SDSG, LLC (The Strategic Decision Simulation Group), we have been
refining methods for identifying leverage points in systemic models for
several years. James Ritchie-Dunham of SDSG will be presenting a paper on
our methodologies at the International SD conference in Istanbul in August.

In Donella Meadows original query regarding leverage points, she proposes a
hierarchy of increasingly effective places to intervene in a system. We
would note that Donella proposes a hierarchy of increasing effectiveness
which begins with SD inputs and proceeds "upward" through such parameters as
delays, "strength", gain, and structure, culminating in the psychological
models of the people as the most effective points for intervention.

Traditionally, SD practitioners have focused predominantly upon the
characteristics of the system. While Donellas list is appealing as a
general interpretation of system behavior, we have developed specific tools
that help practitioners identify useful leverage points. From a purely SD
perspective we would propose that the "effectiveness of a leverage point"
should be essentially the derivative of system behavior with respect to the
change in the element under consideration. More effective leverage points
would have more impact. This requires an operating model for evaluation.
For our work, we prefer to think of leverage points as those points in the
model with high "influence" based upon the structure of the model. (By
focusing on the structure we can seek leverage points in preliminary causal
and stock-flow models and can begin giving the client useful, thought
provoking output much sooner than is required for a calibrated/validated
model.) In our work leverage points have relatively high statistical impact
on other variables in the system, implying (but certainly not guaranteeing)
a large impact on the behavior of the system. Once high-influence
(leverage) points are identified, we further classify them as either having
high exposure to other variables (relatively difficult to control) or
limited exposure (relatively easy to control). It is our experience that
highly exposed "levers" are very difficult to change in complex systems --
the system itself often resists the change -- such that even if the variable
would be an effective lever IF one could change the it, substantial and
coordinated effort are required to acheive the change. As a result, a
portion our work focuses on differentiating "leverage points" which are
within local control (of departments or divisions) and those requiring
broader control (across departments or divisions). Our tool kit automates
the (mathematical) determination of exposure and influence from causal and
stock/flow diagrams and effectively determines a variables leverage in the
system.

A second important distinction we would suggest is that the structural and
systemically-determined leverage points are secondary to the clients
mental model. As consultants working with senior management on real world
problems, there "is no leverage" unless the clients mental models
accommodate the lever. This focuses our initial efforts on capturing the
clients mental models and paradigms, and then on helping them understand
the leverage points within their understanding of the world. Please note:
this does not mean we do not challenge our clients to arrive at an
enlightened model -- we do. However, it is imperative that the client "own"
and understand the model if they are to use it to adjust their world. Once
we arrive at an agreed model, then we can begin to help them consider and
understand the structural implications of their paradigms and thought
processes. Client participation in the process is mandatory. Only by
changing or developing the clients mental models can the systems consultant
impact on the client behavior. We would suggest that the logical sequence
when working with clients is to begin at the top (number 1) of Donellas
list and to work down to the model details. As far as the specific items on
Donellas list are concerned, we find the specific sequence relatively
unimportant as the value of any specific intervention site will, in our
experience, shift with model structure and client paradigms.

I look forward to your comments!
Jay Forrest



The Strategic Decision Simulation Group (SDSG, L.L.C.)
jayf@sdsg.com
http://www.sdsg.com
(281)493-5022 (voice) (281)558-3228 (fax)

[Robert Glitz]

Jay Forrest wrote:
> Once high-influence
> (leverage) points are identified, we further classify them as either having
> high exposure to other variables (relatively difficult to control) or
> limited exposure (relatively easy to control). It is our experience that
> highly exposed "levers" are very difficult to change in complex systems --
> the system itself often resists the change -- such that even if the variable
> would be an effective lever IF one could change the it, substantial and
> coordinated effort are required to acheive the change.

Jay, could you give me a real-world example each of a high exposure
leverage point and a Low exposure one, please?

Also, the toolkit to which you refer. Is it a software add-on to an SD
program or is it methodology? Is it for sale?

Thanks for the enlightened post.

Robert J. Glitz, LtCol, USAF
From: Robert Glitz <rglitz@erols.com>

[Fred Cagle]

>I use a simple 4 dimensions of action list for intervention, evolved
>
>1. Act
eact to whatever you do; that is the simplest intervention.
>2. Script
escript how you react to whatever you do: that requires you
>to be your script writer, a more responsible role.
>3. Cast
ecast the actors within the system: can you make other actors
>in the system more concerned about the events important to you?
>4. Organise
eorganise the system you are part of! Most drastic changing
>the potential itself because new actors interact together to form newer
>systems.

These four dimensions help me to understand a leverage point in one of my
current projects. I would like to see them reststed in SD terms. Thank you
Fred

Fred Cagle
Tijuana Watershed International MAB Project
3751 Ray st. ,San Diego,Ca 92104 619-297-0931
internet Fred.Cagle@sdsu.edu

[Jay Forrest]

Robert Glitz wrote:
>Jay, could you give me a real-world example each of a high exposure
>leverage point and a Low exposure one, please?

A low exposure variable would have limited exposure to influence by other
variables. In certain mental models, selling price may be a low exposure
variable -- for example, for the selling price of a new software product
which has no competition and is sold over the internet so that manufacturing
and distribution costs are negligible. In that case the selling price may be
relatively arbitrary. Whether it is high influence or not will depend upon
the structure of the model in which it sits but one would tend to expect
selling price to have significant influence on the reception of a such a new
product. We find that our analytical techniques are very helpful in
validating the structure of mental models because if we (us and the client)
expect price to be highly influential and it is not, then something is askew
between the mental model and the structure on paper. This indicates a need
to explore the mental model with regard to that variable in greater detail.

General business indicators often show up as high exposure variables.
Profit is always highly exposed except in very simple models because almost
anything that changes influences profit. However, it is hard to generalize
without having a specific structure in mind. We are in the process of
assembling some generic models and analyses to use as public samples. I
will gladly send an email when we have those online for downloading off of
our web site.

>Also, the toolkit to which you refer. Is it a software add-on to an SD
>program or is it methodology? Is it for sale?
The full toolkit we are using consists of a combination of software and
methodologies. The influence/exposure issue is specifically addressed with
software. We are currently exploring our options regarding the marketing of
the software. I should note that we do not intend to release a full blown
SD program - rather it will be a stand-alone which will work with SD models
from existing vendors.

We are in the process of finishing brochures and newsletters for distibution
in Istanbul and will gladly provide more information soon.

Thanks for the questions.
Jay Forrest
The Strategic Decision Simulation Group (SDSG, L.L.C.)
jayf@sdsg.com
http://www.sdsg.com
(281)493-5022 (voice) (281)558-3228 (fax)

[Mohammad Mojtahedzadeh]

The discussion on the leverage points is expanding from guidelines for
identifying them, to methods for conveying them to the client, and of
course, to ways of classifying them according to their effectiveness on
changing the systems behavior, clientss mental models etc.

The multidimensional nature of the concept of leverage points, as the
foremost outcome of effective analysis, suggests that understanding complex
systems is not a single process; it is a continuum from a prosaic cognition
to esoteric insights.

To a large degree, one can classify understanding dynamic systems into
three classes of insight. The first class of insight comes from organizing
and structuring available knowledge about the system through operational
thinking, skillful conversation, and mapping basic plumbing (stock-and-flow
diagram) along with essential wirings (important connections). The second
class of insight originates from model building processes. And finally,
the third class of insight results from a careful analysis of the model and
understanding of the linkages between systems behavior and its feedback
loop structure. It is in the third class of insight that beneficial
discussions on the feedback loops including their gain, strength and role
in generating the behavior, are developed. One cannot expect that the
insights gained in the first stage would be as profound as what can be
gained in the third stage.

The hope is that the first class of insight could shed some light in
discovering potential leverage points as it provides a clearer picture of
the system. I am interested in a collection of heuristics, that if one
follows them, can potentially lead to the first class of insights. This is
particularly useful for those who dont have access to modeling expertise
and is a preparatory exercise for model building. My understating is that
Donella Meadows is trying to create such heuristics. Is that right
Donella?

Mohammad Mojtahedzadeh
High Performance Systems, Inc.
mohammad@hps-inc.com