November 9, 2011: The increased use of UAVs, armed with Hellfire missiles, to attack targets in Pakistan's tribal territories, has led to the deaths of hundreds of Islamic terrorists. The question has often been raised about how the targets were found. Vidcams and electronic sensors on the UAVs, plus monitoring Internet traffic and recruiting local informants have played a role. But another key tool has often been ignored in the media. This tool is predictive analysis, and the CIA, which runs this UAV campaign, has been using this for decades. Actually, this use of patterns, detected from the air, sometimes supplemented by information from people on the ground, has been in use for nearly a century. During World War I (1914-18), when aerial reconnaissance first became a major factor in military operations, it was quickly noted that you could not get as much detail from aerial photos, or eyeball observation, as you could from the ground. But it you collected a lot of photos, over time or just from different angles, and analyzed them, you would uncover patterns that indicated what was a military target and what was not. This was often later confirmed when your troops advanced on the target area, or captured enemy troops who had been there. This form of analysis was further developed, and much more heavily used, during World War II. Since then it has evolved even further, into predictive analysis.

There's another aspect of all this that most civilians, and many journalists, don't understand. In combat, and wartime in general, there is a lot of uncertainly. Sometimes the wrong people get killed. This is why you still have friendly fire deaths. Civilians are aghast at this, but anyone who has been in combat understands that it happens and why. Combat is chaos, and it's easy to open fire on the wrong target. This is even more the case when you are attacking from the air, at an enemy that wears civilian clothes and often uses women and children as human shields.

The way predictive analysis works is quite simple. With more data (from the vidcams, electronic eavesdropping and informants on the ground) it's possible to create a model (or simulation) of what terrorist activity on the ground looks and sounds like. Thus, if the CIA analysts see certain patterns of actions on the ground, they can accurately predict where the Islamic terrorists are, what they doing and, often, exactly who (like a key Taliban or al Qaeda operative) is down there. At that point, the Hellfire missiles are applied. The track record of the accuracy of these predictions has been striking. Few civilians have been attacked, nearly all the targets have been, as the predictive analysis indicated, terrorists.

A key factor in making all this work was the U.S. government changing its policy, in the last two years, of only attacking terrorists who fit a known pattern of terrorist behavior. There are lots of these patterns, many of them specific to a particular individual or group. Predictive analysis cannot always guarantee that a target will be a specific individual, but it can, with near certainty, indicate that the target is an Islamic terrorist.

It all began back in the 1970s, when some CIA analysts discovered a new way to analyze the mountains of information they were receiving. The new tool was predictive analysis. What does this do for intelligence analysts? Predictive analysis was the result of a fortuitous combination of OR (Operations Research), large amounts of data and more powerful computers. OR is one the major (and generally unheralded) scientific developments of the early 20th century. OR is basically applying mathematical analysis to problems. OR turned out to be a major "weapon" for the Allies during World War II. OR, like radar, was developed in the 1930s, just in time for a major war, when whatever was available was put to work to win the conflict. OR is also, half jokingly, called a merger of math and common sense. It is widely used today in science, industry and, especially, in business (it's the primary tool of MBAs, where it's called "management science".) With predictive analysis, the most important OR tool was the ability to "backtest" (see if the simulation of a situation could accurately predict the outcome of something that had already happened, if the same historical decisions are made). For predictive analysis of contemporary situations, the backtest is, instead, a predictive tool that reveals likely outcomes.

Predictive analysis, like OR in general, creates a framework that points you towards the right questions, and often provides the best answers as well. Like many OR problems, especially in the business world, the simulation framework is often quite rough. But in war, as in commerce, anything that will give you an edge can lead to success over your opponents. A predictive analysis is similar to what engineers call "a 60 percent solution" that can be calculated on the back of an envelope.

War is all uncertainty and risk. Whoever can manage the uncertainty and risk the most efficiently will prevail.