Day one of the 2010 CFO Corporate Performance Management Conference in New York is in the books, and while the day’s presentations and discussions should rightly merit being the prime subjects of this post, those events have been overshadowed by one of even greater magnitude: dinner with Thornton May. Where does one start? Where does one end? Before the salad course was finished the topic for the evening had been laid on the table for critique: “SIM City for business decisions.”
Thornton and I were joined in this dialogue by Douglas Hubbard, a specialist in decision analytics and the author of several books on the subject. The impetus for this topic was the acknowledgment, supported by Hubbard’s data, that very few businesses are measuring and monitoring metrics and KPI’s that actually matter. Most businesses are completely surprised to find that the primary driver of their principle objective is one that they have not considered nor recognized as being even slightly important. While many organizations struggle to trim their initial 400 KPI’s down to a more manageable 30 or so, a statistical analysis of their data would generally show that fewer than a dozen are truly meaningfully significant from a cause and effect perspective, and would suffice to manage their performance against.
The unreliability of human judgment is borne out with studies that show judgment-only forecasts are less accurate than the most basic of simple forecasting techniques, and further, that a model that simply averages these human judgments is more accurate than the judgments themselves. In short, human judgment tends to be too optimistic, too confident in its own abilities, and completely miscalculates associated risks. As they say, “Experience is inevitable, learning is not.”
The conversation narrowed its focus a bit during an informative discussion regarding the high efficacy of Monte Carlo simulations in aiding forecast accuracy and in improving decision making. Nuclear power plant simulators are a great example of using simulation to train operators to make better decisions under pressure and over a wide range of simple and multiple system failures. Captain Sullenberger’s heroic landing of his plane on the Hudson was most certainly not a case a good fortune; it was a case of hours and hours in a simulator practicing just such maneuvers.
As the final coffees were being cleared from the table, the consensus was building that management science, and specifically decision making, would only come to fulfill its promise when the power of analytics and simulation was brought to bear. “SIM City for business decisions,” was Thornton’s summation of what is needed to augment human intelligence and judgment in order to significantly improve business decision making. The learning may not be inevitable, but with analytics it certainly becomes much more probable.