CompStat: From Humble Beginnings
The snow was falling, heavily.
Gene Whyte, John Brancato, Richard Mehia and John Yohe wanted to go home, before the weather turned impenetrable. But Brancato, now a narcotics detective, had to save the file the New York City policemen were working on. So the four-man band had to come up with a name, first.
They had eight characters to work with. This was January 1994, and their computer used 5.25-inch floppy disks to store data and used the now-ancient Disk Operating System that Microsoft had originally re-branded to instruct standard IBM-type personal computers how to work.
So someone called out "CompStat." It could be interpreted to mean Computer Statistics. Or Comparative Statistics. Or even Comparative Computer Stats. It didn't matter.
What did matter was getting home.
"We had no idea it would end up meaning something around the world," said Lt. Whyte, who now works in the office of the deputy commissioner of operations for the New York Police Department. "It just meant, 'compare stats, let's go home,' " at the time.
Thus was born a piece of technology that has come to be identified as the core means by which the NYPD turned endless rising crime into consistent reductions of murder, rape, larceny and other felonies. The use of an off-the-shelf program for small businesses led to one of the few revolutions in a major industrycrimefightingthat has continued unabated in the past eight years.
For the week ending June 2, 2002, for instance, the number of murders, rapes, robberies, assaults, burglaries and grand larcenies totaled 2,905, down from 2,977 in the same seven days a year earlier. That was a decline of 2.4% year over year. Over the span of two years, the decline is 17.6%; and the percentage drop over the nine years of tracking is now 65.7%.
But there is precious little technical originality to CompStat. It started life as nothing more than an elementary database, created in a set of desktop office software called SmartWare, from Informix.
When CompStat got geared up, that approach was junked in favor of entering all the statistics into files on an early version of Microsoft's FoxPro database for small businesses, according to Detective James Callinan.
"The technology is fairly rudimentary," says Vincent Henry, an associate professor at Pace University and retired member of the NYPD. "But it is a lot more advanced than it was" when Whyte and cohorts began working in the police department's typewriter-based environs. "It really was a revolution."'s Quest for Stats">
NYPD's Quest for Stats
The idea behind CompStat also was quite simple. The Rudolph Giuliani administration had just come into power in January 1994, and its first police commissioner, William Bratton from Boston, wanted "stick counts," as Whyte would put it, of major crimes as they occurred.
At the time, New York would report its numbers on murders, robberies and other major crimes to the Federal Bureau of Investigation; six months later, it would get Uniform Crime Report statistics back, summarizing trends. Bratton's "stick counts" were initially to be delivered every weekand compared to the counts for the previous year in the same time periods.
Pulling together the counts was no easy sell. Deputy commissioner Jack Maple, Yohe and the startup team had to push, pull and demand that commanders deliver disks of hand-counted data to the police commissioner's office every week, for uploading into CompStat. Even if it took screaming their demands into the middle of the night, when precincts didn't take the need for statistics seriously.
"You would have thought they had been asked to walk barefoot over hot coals," Yohe says of some precinct officers.
The precinct commanders then would have to stand behind their numbers, explain why there were clusters of certain kinds of crimes and specify what they were doing about it. If they did not have a grasp of what was going on, they were reassigned or gone.
"Precinct commanders grew up in an administration where they could all go hide under their rocks," Bratton says now. "They weren't used to being brought in literally under the bright lights. Was it uncomfortable? It certainly was."
In effect, Bratton was bringing into the practice of policing what business schools might call "management by objectives" or an information technology department might call "managing to the baseline."
"For us, crime was the bottom line," says Yohe, now a private consultant. To establish a baseline, the early crew of CompStat compilers had to count by hand the incidents of major crime for every week in 1993.
Within a year, mapping would be added. Yohe would spend 18 minutes a day pushing pins into a physical map for one precinct, to prove that, even without mapping software, clusters of crime could be identified rapidly and addressed.
That made managing crime visible. "We take the biggest dot and make it go away," says Yohe. "Then we go to the second biggest dot and make it go away."
If, for instance, the dots indicated a slight uptick in old ladies getting mugged on the No. 4 subway train, Manhattan South could put a decoy on the train in "some kind of drag," says Henry.
The data was soon visible on screens of all sizes, through a no-frills software selection process. The CompStat team got the green light to map dataas long as it used the MapInfo software sitting on shelves unopened from a previous initiative.
The stick counts formed the basis for constant attention to crime-fighting detail across all 76 precincts. The counts provided what Bratton would call "timely, accurate intelligence," and then it was up to all chiefs, commanders and street cops to respond rapidly, come up with effective tactics and relentlessly follow up.
"The process only works if you're willing to ask hard questions. Why are you having this crime? And what are you planning to do about it?" says Whyte. Not too much to ask of a commander who is running the equivalent of a $20 million-a-year company.
The process was played out in management conferences. CompStat meetings constantly addressed old issues and tackled new ones. As many as 150 commanders, detectives, information specialists, officers and agency heads might pile into a room at One Police Plaza to hash out what to do about the latest stats.
"It allowed me twice a week to literally walk into the engine room of that very large ship that I was steering and meet with all levels of the organization at one time," says Bratton. "Everybody is included in that meeting that has anything to do with the goals we are seeking to achieve."
Over time, the department began feeding in information on recovered vehicles, general complaints and even reports of shots fired, whether from crime or not. The idea: Disparate data, tied together, provides clues to the right tactics to take. Eventually, reports from bus drivers and taxi drivers could be added, in an attempt to enlist more eyes on the streets.'s Successes">
The process now has become institutionalized in the NYPD. After all, CompStat was created out of "collective police wisdom" and then ingrained through years of CompStat meetings, tactics and follow-up, notes Phyllis McDonald, director of research for the division of public safety at Johns Hopkins University. The result: New York reported another 4% drop in homicides in 2001, after 12 straight years of previous declines. Meanwhile, in Boston, where Bratton previously served as chief of police, homicides last year went up 67%.
So celebrated is the CompStat success that Maple, Bratton, Yohe, John Linder and others have been able to create private careers based on implementing it in other citiesand even in private businesses.
Indeed, in January, shortly after leaving office as mayor, Giuliani formed a consultancy called Giuliani Partners, with the express purpose of showing how to use CompStat as a means of monitoring corporate performance.
Since that time, Giuliani's communications director, Sunny Mindel, has declined to return at least nine separate invitations for the former mayor to describe how he planned to apply CompStat to businesses; and, one on how he thought CompStat could be used to counter terrorism.
At the annual convention of the Direct Marketing Association at the Jacob Javits Convention Center in New York in June, Giuliani himself brought up that potential. The "war in Afghanistan is being fought" the CompStat way, he said, without elaboration.
He called the CompStat makeover of law enforcement a blend of people, process and technology "in equal measure." He said he called for CompStat statistics every day, so crime got constant attention. "If you measure something, you can manage it," he said in responding to a Baseline question. "If you can't measure it, it floats away, you can't manage it."
But Giuliani may not be the best person to actually turn around and implement new CompStat programs, suggests Bratton, who fell out with the former mayor after 27 months on the job, partly due to battles over credit for the success of CompStat.
"He literally attended, I think during my time, one meeting" of CompStat, Bratton says. "He had nothing to do with its initiation, design, implementation or resulting manifestations. But he understood how significant it was to ensure that the strategies that we were putting into place" were effective.
Bratton still criticizes Giuliani for not investing in technology to achieve permanent reductions in crime, saying the mayor preferred to pay overtime to officers to get short-term results.
He expects the current commissioner, Raymond Kelly, to weave the fight against terrorism into the CompStat process, and possibly establish an entire antiterrorism unit, using the measurement and accountability approach to identify and prevent future terrorist attacks.
"At the NYPD, we changed the mission from measuring our success on our response to crime," says Bratton, "to basically the record of success is, 'How many crimes did we prevent?' That's the guts of the mentality change that CompStat was essential to bring about."