Byte cop

Brett Goldstein maps out Chicago’s crime-ridden spots with his predictive-analysis system.

By Elizabeth Flock
Photography by Dan Dry

Body of work
At the Chicago Police Department, Goldstein wanted to design a computer model to replicate an officer’s intuition.

When Brett Goldstein was six years old, he started writing programs on his parents’ single-circuit-board computer. In fifth grade he was teaching himself programming languages. At Connecticut College a professor told him he’d “maxed out” the curriculum.

In his early 30s Goldstein, SM’05, became IT director at OpenTable, an online restaurant-reservation company. He had been an early employee at the start-up, founded in 1998 in San Francisco. By 2006 he had moved to Chicago, completed his computer-science master’s at the University, and was making six figures at OpenTable, when he surprised colleagues and friends by leaving the company to join the Chicago Police Academy.

It wasn’t his first foray into criminal justice. In 1999 he received his master’s degree in the field from Suffolk University, with the goal of working in IT security; part of his role at OpenTable involved network security. But he was inspired to do something different after the attacks of September 11, 2001. “I realized I wanted to be doing some kind of service that went beyond just volunteering,” Goldstein says. Unsure of which direction to go, on a whim Goldstein took the police-academy exam in 2004. “I like to take tests,” he says as explanation.

But his choice to actually enter the academy to become a beat cop wasn’t a quick or an easy one: “It was a big decision that required a lot of thought,” he says. “I became more committed with each step of the police screening process.” He knew that, as soon as he graduated from the academy, he’d be working a beat; that’s just how it was. But he “had high hopes” that he’d eventually be able to use his data-analysis and computer-science training for the Chicago Police Department.

In early 2007 Goldstein was assigned to the Harrison District, one of the toughest neighborhoods on Chicago’s West Side. (Goldstein, raised in Boston, now lives in Chicago’s Pilsen neighborhood.) After 13 months, he moved to headquarters. While working on the West Side, according to an August Chicago Sun-Times article, Goldstein had “started thinking about how he could design a computer model that could replicate” an officer’s intuition. He had this in mind when he transferred off the street.

In 2009, with the help of a $200,000 National Institute of Justice grant, Goldstein launched his predictive-analytics project. The group would analyze crime data to focus manpower where trouble was most likely to occur.

The germ of the idea was born earlier, back when he was getting his master’s at Chicago. Computer-science professor Leo Irakliotis had Goldstein in his 2005 intensive data-mining course, where students learned to extract patterns from data.

The two worked together on a project to analyze call records from the Oak Park (Illinois) Police, to see who frequently called 911 and hung up. They then predicted the corners most likely to have hang-up offenders.

It’s finding these patterns in seemingly random events that drives Goldstein’s work at the CPD, where he makes forecasts using the entire data system. Data analytics uses time and location patterns to calculate where crime might happen. Goldstein’s group, for example, might show “a wave of burglaries in one neighborhood,” says Irakliotis. “They would use the data to see how it might expand to another neighborhood.”

Although a number of U.S. police departments now use such figures to draw conclusions about crime, Chicago has one of the largest data sets in the country, with more complex breakdowns of crime details. The city, says CPD spokesman Michael Fitzpatrick, began collecting electronic data earlier than most departments, in the late 1990s, partly thanks to former assistant deputy superintendent Ron Huberman, AM’00, MBA’00. While working in information services, Huberman helped create a central database to house all police intelligence: in 2001 the department partnered with the Oracle Corporation to create CLEAR (Citizen Law Enforcement Analysis and Reporting). The system—now recognized by the U.S. Justice Department as a law enforcement best practice—is used to streamline processes such as filing arrest reports and provides officers with easy-to-access information.

Goldstein, teaming up with the Illinois Institute of Technology and the Rand Corporation, has helped take this data to the next phase. After about five months in beta, the Predictive Analytics Group officially launched in August 2010, with Goldstein as director. Not all Chicago cops, however, are enthusiastic about the program—or about Goldstein. In August a popular police-department blog, Second City Cop, reacted to the Sun-Times story about Goldstein, titling its post “Golden Boy Interviewed.” It generated 173 comments, most negative. Concerns ranged from the unlikelihood of success to personal attacks on Goldstein and anger over his quick promotion—he’s been with the department for four-and-a-half years, and by age 36 he holds a director position.

Goldstein, named to the 2010 Crain’s Chicago Business 40 under 40 list, stresses that the program is a tool for the officers, not a replacement. “You can’t underestimate the guys’ relationship with the community,” he says. He has invited skeptical beat cops to visit his office to discuss how it works.

Part of the officers’ concern comes from the difficulty of evaluating the program’s accomplishments. “It’s hard,” says Fitzpatrick, the department spokesman, because “if there’s a likelihood for something to happen and it doesn’t, that’s a success. So how do you gauge that?”

Goldstein doesn’t claim his analytics would be the only reason for department successes or lower crime. Every day his group sends out intelligence showing the areas with the biggest potential for violence. Then the department that deploys officers produces its own information, based on gang conflicts and human intelligence. The two elements are combined with feedback from the patrol division to come up with that day’s patrol plan. The “key idea,” says Goldstein, “is to give the field commanders the latest and best information to be able to deploy their resources intelligently.


Return to top