In complex systems, problem symptoms are often far removed from the problem causes.
In the 1950’s Stafford Beer managed the largest Operations Research firm in England. A large corporation would come and say “Our inventory is out of control. Can you develop a system for us to get the inventory costs down?”
Stafford always said something like “We certainly can. You have come to the right place. However, to make sure the new inventory control system will really and truly fit your company, we had better poke around a bit. We need to interview some folks and spend some time observing how your plant actually operates.”
After a bit of poking around, Stafford’s team would discover that theinventory control problems were merely symptoms of some other problem. The key problem always turned out to be an inappropriate compensation scheme for the sales force or an inadequate production control system or some thing else equally far from the inventory control system.
For example, in one manufacturing company the salesmen earned the same commission for selling a manufacture-to-order item as for selling their most popular product. This compensation scheme insured that the inventory would never be adequate, because the salesmen kept selling unique items. The inventory problem was fixed, along with the firm’s overall profitability, by making the sales commission depend upon the profitability of the item sold.
Stafford’s firm had dozens of these inventory control engagements. They almost always carried them to a successful conclusion — and they never built a single inventory control system. The real problem always turned out to be something else and the inventory problems were merely symptoms.
Now imagine what would happen if the consultant engaged to fix the inventory control problem was the world’s best expert on inventory control systems but didn’t know much else. In the example cited above where the salesmen kept selling unique items, the firm might sink massive amounts into improving the inventory control system. They might become extremely good at keeping precisely the optimum quantities of all of their standard products in the warehouse. And all of this effort would mask the real problems and make it even less likely that anyone would stop and examine the structure of the sales commissions.
Companies are complex systems, with multiple circular chains of cause and effect. For example, reliance on sending orders down the chain of command discourages initiative on the part of the lower ranks – and this loss of initiative means that you can’t get anything done without issuing lots of orders.
In such a company, the executives and the workers are engaged in an unrecognized conspiracy to ensure that the executives stay busy issuing orders and the workers don’t do anything except what they are explicitly told to do. The workers become cynical and irresponsible, and save their thinking for when they are off the job. The executives lament the sorry state of morale, the low quality of their work force, and the excessive time it takes to actually get anything done.
The executives never realize that they have “trained” their work force to be irresponsible and unmotivated.
Peter Senge provides many other examples of circular causal chains in companies. He says that the symptoms that we struggle with are generally far removed from the real causes. Consider our two examples above:
(1) The symptoms are that inventory is out of control. The real problems turn out to be the structure of the sales commission or the production scheduling system.
(2) The symptoms are unmotivated and irresponsible workers. The real problem is a management team that has unwittingly encouraged these behaviors on the part of the workers by an over-reliance on issuing orders.
To paraphrase Russell Ackoff, it is far more useful to fail at solving the right problem, than to succeed in solving the wrong problem.
Determining the right problem requires skills in observing and modeling. All companies do this, but usually only in an ad hoc way.
Observing includes gathering data, statistical analysis, examination of processes, and often just plain old fashioned conversation and direct observation.
Modeling is making sense out of our observations. It puts the data in a context for making predictions. It is often facilitated by tools, such as systems dynamics (a method for modeling in terms of circular causes and effects) or statistical analysis. A good model interprets our observations, provides a road map for action, predicts results, provides measures, and can even be useful for supporting or helping to define a desired vision.
For a company to consistently produce good models requires a set of skills. These skills can be facilitated and learned. They include the use of powerful tools such as those mentioned above. More importantly they include the abilities to determine the focus (appropriateness), to exercise rigor (reproducible observations and valid generalizations), to develop an open and true sharing of ideas (dialog), and to define expectations (measures for success).
When such skills are explicitly facilitated and learned, observation and modeling become more consistent, predictable, and less dependent on personalities. The company willing to make such an investment will find itself more consistently attacking the right problem. Most importantly, it will have developed a distinct competitive advantage: it will have learned to learn.
Originally published as
Volume 1.8, November 4, 1998 of The Corporate Forum
Department of Engineering Management, Old Dominion University