Since 9/11, any big technology initiative in law enforcement has likely linked to counterterrorism. It was a hot-button issue that everyone understood. There were, and are, federal dollars available to buy high-fallutin’ and mega-expensive technologies that promise to help us fight crimes that would harm “soft” targets.
Big Data is a term you’ll hear more and more. Like last year’s “cloud,” Big Data is a buzz-phrase that describes the process of aggregating, correlating, analyzing and otherwise mining exceptionally large datasets.
What does Big Data do? In his book, The Power of Habit: Why We Do What We Do In Life and Business, Charles Duhigg explains how, knowing nothing but what a woman buys, Target can state whether she’s pregnant and accurately predict within two weeks her due date. That’s Big Data working.
Another example: By taking video and capturing wireless network information from every navigable street in America, Google created some seriously big datasets, which it mines very well—your mobile phone’s location probably derives its precise location by asking Google where it is, and Google knows.
These are big data success stories. But like any other new technology, Big Data has tankers-full of hype. There are many latchers-on, marketers and charlatans who would have you believe that Big Data will solve all your problems. Rubbish.
Or, more precisely, “Rubbish in, rubbish out.”
How It Really Works
“Ask any analyst what they do for a living and they’ll tell you they find actionable patterns in data,” says Eric Olson, vice president at the intelligence firm Cyveillance. “Ask them what they do all day, however, and they’ll tell you, ‘I throw data away.’”
In law enforcement, we’re now being sold Big Data solutions to problems we don’t have, and we’re not applying Big Data mining techniques to solve the problems we do have. From agencies like Nassau County and Rochester, N.Y., we’ve learned that you can do a whole lot of mining with off-the-shelf software and hardware. Ask me about sophisticated data mining and we’ll tell you just how much you can do with humble Excel and some pivot-table-fu.
Outside Help
One very important development in law enforcement in the past 24 months has been the rise of our understanding of the value brought by firms doing data aggregation. These companies do what’s called “bridging,” or “ingestion.” They’ll bring in data from across silos—records management, computer-aided dispatch, court records, NCIC, etc.—and make it instantly accessible from a single pane of glass.
This is monumentally valuable. We’ve all seen the deadly consequences of having life-saving information not get to an officer because it’s sitting in a database somewhere. Totally avoidable tragedy.
It’s important, too, to understand the limitations of data aggregation. While it’s fantastic in terms of retrieval (if I run a driver license, these technologies will check multiple and previously unconnected sources to get all the available information on that person), it’s not doing any kind of automated analysis, predictive intelligence or data reduction. At its heart, this is the same kind of reactive, pull-only kind of data that law enforcement has had for years—albeit with massively better, and better-synthesized, information stores to pull from. So even though it’s connecting large stores of data, it’s not Big Data.
This isn’t a semantic distinction. It’s an important concept when you’re considering what you’re buying and why you’re buying it.
Aggregation will get you massive efficiencies in some areas, and it’s the first step to getting connected to data from your own and other agencies. But the story doesn’t end there. In fact, it just begins.
What to Think About
The important thing is to use this to think about all your data—from all your repositories—and how you can best look at what it can offer you. If you were to try and find meaningful ways to evaluate how your patrol officers are doing beyond conventional numbers, you might want to start by looking at the humble citation.
By taking all the data from every citation each of your officers has written, and analyzing patterns, you can start to see which of your officers are writing tickets that contain certain kinds of information, and you’ll immediately spot the ones who seem to be asking more questions and recording the answers.
When an officer captures a civilian’s answers to questions like, “What do you do?” and the answer is, “I’m a student,” your citation is going to show you whether the officer is following up and asking, “Oh, a student? What school are you in?”
Or the one who, while writing a citation, asks: “Oh, you’re self-employed? What do you do? You do that from home?” These things sound trivial, but they’re an indicator of engagement and even of patrol emphasis that can be counted and averaged and result in understanding your officers’ strengths and weaknesses.
As you start to look at the corpus of citation data, you’ll start to see your officers who are “good” at this particular function. More important, you’ll start to understand something much larger about what these officers do. Are they doing traffic? Patrol? Interdiction? This will be startlingly clear from the citation data.
What’s more, you can take these statistics and start to predict whose citations will be dependable for other things—like which officers are capturing the kind of information that make it easier to do follow-up investigation or locate the person when a warrant is issued.
This isn’t to say that Officer A is any “better” than Officer B. Maybe Officer A is a motor jock, and Officer B is doing interdiction. These officers have vastly different reasons for making contact and for writing citations in the first place. These can’t be seen as “job performance” exercises: Rather, they give you insight into how you can best leverage each officer’s skills.
The point is, once you get into the data, patterns that are much more meaningful than “arrests” or “citations written” begin to emerge. You can start to group officers by functional role, so that numbers are being compared to similar numbers, rather than an arbitrary measurement.
You can do this in Excel. When you get lots of data, you can do it in an access database. Now, is this Big Data? Not quite, but you’re getting there.
Conclusion
Once you start getting good at it, then you can start talking to Big Data vendors about what they have to offer. Because you’ll understand the basic value proposition they bring to the table, you’ll also know that no technology is “the solution” to your problems.