On the Commutecast this week, I talked about how Gerg Gigerenzer and colleagues have shown that a simple algorithm can out-perform a complex one in predicting future events (PDF). It’s not just that getting the information is expensive — whether you are hunting it down or finding it, collecting data costs you effort and often resources. It’s that even where data collection is free, having a more complex model produces worse predictions.
J. Scott Armstrong and Kesten Green have discussed this in the forecasting field. Simple (but not simplistic) models often outperform their more statistically complex cousins. You simply get worse predictions using more complexity.
The simple is better proponents point out that businesses often get complex forecasting models done, with obtuse and opaque statistics that are barely understood by the forecasters, because they want need to absolve themselves of accountability for their decisions. “Hey, it’s not my fault that it didn’t turn out!” they say. “I used this massively complex forecasting model and it failed.”
All too often, managers want to get rid of the uncertainty in their decisions. That’s not decision-making: it’s wishing for childhood. If we didn’t have uncertainty we wouldn’t need to make decisions. Everything has some level of uncertainty. Even things like walking down the street. There is always a real chance that you will trip on your own feet or that your legs will give out suddenly. Ask any elderly person. We think that things can be “sure things” but they can’t.
If you want to have the management job, you have to live with uncertainty in the results. The greater the uncertainty, the greater the reward, of course, if things pan out.
Bosses want things to always work out. They expect zero uncertainty, no need to replan or reconfigure for success.
No, let’s make that “bad bosses” and “poor performing managers” expect zero uncertainty. Real high-level managers understand and deal with uncertainty.
Image credit: “Football men exercising, Harvard”, Bain News Service, ca. 1910. Library of Congress (USA) collection