Predictive analytics: the future of successful law enforcement?
Connecting state and local government leaders
Picking criminal cases to investigate and prosecute can be a subtle, subjective process, former special agent John A. Cassara writes. Analytics software can help prosecutors make the right choice.
Criminal investigators and their managers are engaged in a constant struggle to identify and pursue successful investigations. For example, today's complex financial fraud cases sometimes take years to complete. From a management point of view, it is a tremendous investment. They can't afford to waste scarce resources that lead to investigative dead-ends.
What most outsiders do not realize is that a very large percentage of investigations are unsuccessful.
Although commentators argue the point, the bottom-line metric that quantifies success for law enforcement is the number of investigations that result in successful prosecutions and convictions. Another key metric for certain crimes is criminal assets forfeited. It is hoped that those numbers will correlate to deterrence.
Within law enforcement, there is a subtle and sometimes selective weeding of cases that are chosen to be pursued. For example, cases are constantly by-passed or bureaucratically pigeon-holed because of issues of jurisdiction or venue. Resource constraints, departmental policy and sometimes even political considerations are also considered.
In my law enforcement career, I was frequently frustrated when a promising case was rejected for prosecution by an assistant U.S. attorney because it did not meet the vagaries of "prosecutorial merit," including a dollar threshold (how much money was actually lost in the alleged criminal acts). For example, in a September 2012 report, the Institute of Medicine estimated that criminal health care fraud costs taxpayers approximately $75 billion a year.
Yet depending on the jurisdiction and extenuating circumstances, prosecutors could impose a limit that unless more than $1 million dollars of Medicaid fraud has been committed, a case would simply not be accepted. Or sometimes long-term, complex, financial-crimes investigations are not accepted for prosecution because the case is simply too hard to understand or difficult to put together. And, of course, prosecutors always want cases that have jury appeal.
At the same time, the situation is becoming increasingly complex because more and more data is being made available to criminal investigators. To take just one example, about 18 million pieces of financial intelligence or Bank Secrecy Act data are filed with the Treasury Department every year. This paper trail often helps law enforcement analysts and investigators follow the money. Other big data sets include law enforcement records, commercial database entries, travel and immigration records, trade data, and motor vehicle registration information. However, turning increasingly large amounts of data into useful insights and discovering future successful cases remain a challenge.
Yet despite the challenges and constraints, every law enforcement agency or department has many examples of successful cases. They might involve narcotics smuggling, white-collar fraud, stolen vehicles, weapons trafficking, welfare fraud, counterfeit identification or intellectual property rights violations.
Successful cases tend to have common denominators. Some of the criteria include the way or method in which a case was initiated, data sources, investigative techniques used and successful prosecutorial strategy. These winning cases could act as models or templates for future cases.
A way to improve this process is with an emerging breed of software that can explore and analyze data to help uncover unknown patterns, links, opportunities and insights that can drive pro-active, evidence-based decisions. Often referred to as “predictive analytics,” it is available to help law enforcement sort through big data sets to find nearly identical data elements that match successful cases. I believe this technology could revolutionize law enforcement decision-making.
For example, data elements in a human trafficking case in Los Angeles could be matched to a similar, successfully completed case in South Florida. Investigators and their managers in California would feel confident initiating a costly investigation knowing that a similar case proved successful and that they already have an investigative road map.
The next step in the process is prosecution. According to a ranking Justice Department official, prosecutors generally do not use empirical data in case selection. Rather, the case has to "feel right." The official continued that case selection is currently "more art than science."
Using predictive analytics as a state-of-the-art decision-making tool can help prosecutors feel more confident about prosecuting a case. They will know beforehand that the case has been vetted, initiated and investigated using elements that proved successful in the past.
The bottom line in this era of diminishing resources is that predictive analytics could boost productivity for criminal investigators and prosecutors. It could fundamentally transform the way cases are selected, investigated and prosecuted.