10/12/98 WHEN BAD MODELS LEAD GOOD COMPUTERS ASTRAY

WHEN BAD MODELS LEAD GOOD COMPUTERS ASTRAY

Am I the only one who noticed the connection between ''Misfire'' and ''Virtual management'' (Management, Sept. 21) in the same issue? I wrote my first computer model in 1953. It slowly produced the wrong answer! Since it was a physical system we were modeling, it didn't take long to see that it was the wrong answer because we had a general idea of what the right answer should look like. It took much longer to figure out why our computer model didn't work. We decided that we didn't know enough about the details of the processes underlying the physical system we were trying to model, so we gave up.
It seems obvious that nobody will ever learn enough about the details of how human beings will react to new (to them) conditions to build a valid predictive computer model based on such reactions, no matter how much statistical ''history'' was built into the model. I would never bet any of my own money on the results of running such a model, whether the model is pointed toward Wall Street investments or toward customer behavior patterns in a department store. I guess some people will never learn that just because ''the computer says so'' doesn't make it true.

Murray Lesser
Yorktown Heights, N.Y.

I wonder if others caught the irony between ''Misfire'' and ''Virtual management.'' The cover story was all about the failure of software to make good investment decisions, while the management story was about developing software to make complicated management decisions. Forecasting anything means taking information we know today and using it to predict what will happen tomorrow. The concept is the same whether we are talking about the stock market or the retail industry.
There are two basic keys to good forecasting: being able to quantify stable cause-and-effect relationships and/or having a stable background environment for the relationships you cannot quantify. It would seem that the wealth of quantitative stock market data would make it possible to develop forecasting software that could outperform the market. But ''Misfire'' points out that just maybe the investment software gurus relied too much on a stable background for relationships that they could not quantify.
If we can't develop forecasting software for the stock market that can eliminate embarrassments, then we shouldn't expect to develop forecasting software that will eliminate embarrassments in the area of customer behavior. Can we develop software that will help us gain a better understanding of important factors? Absolutely, but don't expect a final answer from your computer.
Good companies already have a tool ''to safely test hunches and scenarios without major investments--and embarrassments.'' This tool is called ''talking with your customer.'' Unfortunately, it is a tool that is not widely used.

Robert J. Chambers
Norcross, Ga.

Thanks for giving prominent coverage to what has been viewed as a ''far-off and far-out'' development: management tools based on complexity science. But your readers should know that these tools are more within the reach of the average company than your article implies. For example, the ability Macy's is seeking--to use the computer as a laboratory by running simulations of the behavior of complex systems--is already in place at Japan Central Railway. They use faster-than-real-time simulation to find the best solution to snafus on the line and work around them before any commuter is affected. Our own simulation work with Hewlett-Packard Co. focuses on the behavior of a crucial asset: the labor pool of knowledge workers HP needs to retain and recruit.
Moreover, simulation isn't the only business application of complexity science. We have seen it at Deere & Co. in a new approach to production scheduling, at Marks & Spencer in their credit scoring of charge-card customers, and in many other settings. In the longer term, the lessons of the adaptive-systems approach will change the way we view corporate strategy and organizational behavior.

Christopher Meyer
Director
Ernst & Young
Center for Business Innovation
Cambridge, Mass.