Power in Data: Turning Facts into Discoveries
Blog: Smarter CMO Solutions
Mark Cleverley - Mark Cleverley helps governments with technology-enabled transformation to ensure the safety of their citizens. He advises IBM’s public-sector customers and IBM teams on potentials, challenges, and best practices in the evolving use of new technologies related to public safety. He has consulted widely for many government projects and has written and spoken publicly extensively in the United States and abroad. Mark Cleverley has worked with governments in many nations and many areas of information systems.  

Innovator Chat: Smarter Analytics Yield Safer Cities


Keeping our cities safe is essential. Today, police forces nationwide rely on technology to help them anticipate -- rather than just react to -- crime. By connecting previously disparate data and revealing key patterns, analytics can help them yield actionable insights and uncover trends before they become criminal events.

Mark Cleverley, Director of Public Safety at IBM, responds to our questions on how analytics are helping police forces across the country make our cities safer.

1. How is data most commonly used today by law enforcement to yield actionable insights and reduce crime? What's lacking?
Today around the world many law-enforcement organizations are adopting more "data-driven," "fact-based," or "intelligence-led" approaches. Most would agree that we are at the beginning of being able to do this most effectively, and that much remains to be done. Today, data sources are often fragmented, unconnected, and constrained from being used appropriately by technology, governance, and cultural inhibitors. By no means are all potentially relevant sources of information available to those who might benefit from them. Provision of data to the "front lines," in usable form, is often only done in relatively new projects that do not necessarily take advantage of potential technology improvements.

These inhibitors are being removed over time -- For example, the new Public Safety Broadband Network in the United States will remove some key technical barriers, enabling police and other first responders not just to interoperate, but also to access in more effective ways rich new applications and data. Cultural inhibitors are also under siege, with agencies accepting the need to move from "need to know" to "need to share."

But there remains much to do, and we believe that analytics disciplines have a critical role to play.

2. How does your software integrate data that provides a more holistic view and helps uncover criminal trends more quickly? 

We identify five disciplines that many law-enforcement organizations are deliberately trying to improve in this regard.
  • Access to relevant data -- which covers identifying, connecting with, and collating data from a wide range of increasingly diverse sources, increasingly in digital formats, under appropriate governance, so that it can be efficiently processed and used in real time.
  • Ensuring integrated, trusted information -- which is concerned with merging information from many disparate sources and systems, and sorting it to provide a trusted base of information. Building data warehouses and portals so information can be accessed by those who need it, while instilling the data governance processes to ensure quality is enforced and compliance implemented. 
  • Creating a trusted information layer -- from all the information available, making sure that it is cleaned, de-duplicated, etc. -- and that disparate records that may appear to point to different entities are resolved to ensure that all relevant information about an entity is accessible and clear.
  • Delivering responder operational insight -- which is about providing appropriate information in usable formats to those who need it on the front line. This includes real-time video feeds to and from handheld devices, providing access to visualized combined data sets so that responders can make better decisions -- and automating much of the reporting so that field personnel spend more time in the community and less back at the office. Interoperability is a key part of this -- ensuring that information can be shared as needed, across different agencies, and, when appropriate, to and from the community and the private sector. Also important is information management and design -- to ensure presentation in relevant and useable ways. For example: Sending 500 mugshots to a handheld or vehicle-borne device at the beginning of a patrol is less useable than, say, sending mugshots of the five or ten wanted individuals that the patrol officer is most likely to encounter, given other relevant information such as planned patrol routes, notifications of who might have been released from incarceration into the community, etc. Of course this works in both directions, enabling collection of information from the field via these channels -- for example, directly assimilating imagery or reports.
  • Adding proactive planning and decision making -- Making better use of the information available to improve strategic and tactical decision making -- including identifying trends and patterns (e.g. crime hot spots) and using the data to more effectively schedule police resources, share crime maps with the community, track effectiveness, optimize resource allocation, and use intelligence to anticipate and prevent events from occurring.
  • Enabling more unified threat assessment and response -- sharing information across all relevant city departments -- including, for example, those managing transportation, weather conditions, utilities -- so that situational awareness can be optimized and the command center can allocate resources across agencies to optimize the response. 

3. How is information compiled and used in a way that is respectful of privacy?
Law-enforcement organizations (LEOs) take privacy very seriously. There are risks that the increasing use of technology in these areas presents a challenge to privacy. Like any technology or process across history, if it is misused in the pursuit of privacy violation, then there is ipso facto a threat to privacy. But just because the technology can do something does not mean that it should be done. We have institutions whose role it is to represent society in making that kind of decision, and LEOs are bound by the laws that come from that set of decisions.

In addition, technology can provide levels of protection against privacy violations. For example, with digital data it is easy to establish mechanisms that can unequivocally record what was done to a piece of information, by who, when, etc. -- and thus create an audit trail that is more detailed and accurate than for many paper records or manual processes. And appropriate processes can be implemented around technology deployments that provide other levels of safeguard -- for example, ensuring that street or traffic cameras are not installed where they may be able to see into private property -- or making sure that individual faces that might be captured on video are by default "pixilated" out, and that access to the original imagery is strictly controlled by legal process.

4. How do law-enforcement agencies see information being used to increase their operational efficiencies?
Historically LEOs have attacked the problem of reducing crime by developing their response capabilities, investigating faster, and ensuring high quality hand-offs to corrections and courts. Today they are adding the ability to find ways to keep crime from happening in the first place. By becoming better at capturing and managing data and turning it into actionable insight, agencies can improve outcomes without necessarily having to invest commensurate new resources.

In order to do this, agencies must work to access and exploit all available data, and then get that information to those who need it. Analytics can help to identify trends and improve situational awareness. With this insight, prevention and enforcement can be better targeted. And, with good interoperability and a common understanding of the situation, speed of response and coordination across agencies can both improve. Real-time insight provided to the front line makes first responders more productive, giving them the ability to focus on the mission and perform it more effectively. New technology in itself is not enough to enable this -- a more holistic approach, which brings all relevant information and resources together appropriately, is important.

5. Which user communities other than front-line officers benefit from your software?

Many user communities can benefit directly or indirectly, but perhaps most benefit in the early days accrues to the analysts. The creation of broader, better, faster access to more diverse sources of data, together with the analytical toolsets at the desktop and in powerful servers, gives analysts a much richer landscape of solutions with which to do their jobs.

Other communities that can really benefit from the increasing use of information and analytics include LEO management -- from those responsible for deploying patrol resources, to those whose job it is to communicate strategies and benefits to the population that the agency serves -- and their associated collaborating organizations (like Offices of Emergency Management) and stakeholders (the Mayor or the City Manager role).

6. How does your software provide improved situational awareness? How can information be used to promote trust between police and the community?
If relevant information is better understood and better shared, in more timely ways, then situational awareness is improved. This can be at tactical and operational levels, where, say, officers or citizens are sharing live video from the scene of an incident, or where personnel from different agencies are able to exchange tactical information securely and quickly. It can also be at the strategic level, where planners and policymakers take advantage of more accurate, more powerful sources of information and predictive models in creating their own shared understanding.

In promoting trust between police and community, nothing replaces clear, consistent, and frequent communication between those involved. Arguably those interactions can be enhanced and made more effective, when the leaders involved have access to knowledge and understanding that is demonstrably fact-based, etc.

7. How do you ensure that police officers understand the power of these tools? Do you use training sessions to brainstorm data sources that cops have always intuitively relied on but never codified?
One way to think about the predictive analytics discipline is exactly that -- it codifies experience and knowledge that one finds in experienced officers with knowledge of particular geographies or communities, but does it more quickly and at greater volume. Hence it acts as a "force multiplier" and a collator of institutional knowledge.

8. How do you populate databases and verify data to promote trust and common standards between police and budget staff?
There are different nuances to be addressed within this question.
  • Databases are populated in accordance with, and only under the aegis of, the legal and regulatory framework within which a LEO operates. Common standards, for example in the United States the National Information Exchange Mechanism, underpinning such applications as Uniform Crime Reporting, etc., are of great importance in creating common understanding and building shareable sources of information. We support and recommend the use of relevant standards in the pursuit of these objectives.
  • In 2012 IBM released the results of some research into the return on investment in technology within law-enforcement organizations (in which the five disciplines mentioned earlier were elucidated). This linked with and supported our position that a more holistic view of public safety was emerging, underpinned inter alia by technology enabling connections to be made between more and different organizations and communities that could play a part in creating a safer environment. The research explored the benefits of technology investments that were traditionally the purview of budget officers -- those "hard" benefits directly accruing to the agency making the investment. But it also explored the "softer" benefits to society as whole -- including, for example, the impacts of crime reduction on victim costs, on insurance rates, on the costs incurred in other parts of the criminal justice systems such as the courts and corrections, etc. It has been our experience that this perspective helps constructive communication between budget-oriented staff and operational policing leaders, and we surmise that it will also help communication between the policing organization and its funding stakeholders.

9. How do (or don't) personal data like prior convictions factor into predictive analytics?
In general, predictive analytics in policing is not about individuals. Almost by definition prediction is more powerful when applied to large populations of events or attributes. So we do not expect personal data about specific individuals to play a role in the way these capabilities are being used. In the aggregate, it might be possible to draw conclusions with degrees of confidence about the likelihood of certain types of offense to be associated with others, or with geography, or with certain conditions, etc. But it is beyond the scope of the techniques discussed here to make any predictions about specific individuals -- that may remain the domain of science-fiction movies.

10. What are some ways smarter analytics helps police departments reduce costs and optimize resources? Does it eliminate unnecessary drives, prevent the need for backup by preventing crimes, or simply save admin cost?
At the risk of a cliché -- all of the above! Perhaps you can reduce costs by predicting service schedules for vehicle fleets. Perhaps you can optimize resources by pre-deploying them to areas where need is predicted to be greater/more relevant. In doing so, you may be able to retain some resources in appropriate reserve. Patrols can be planned and directed, and rosters developed, more effectively.

We believe that the broader use of analytics, in all the disciplines described below, offers a very rich landscape of potential benefits to many areas of policing. It is the start of a journey which can help LEOs and their stakeholders achieve significant operational and strategic gains.

11. How are other analytic capabilities such as content analytics, video analytics, and social-media analytics benefiting law-enforcement agencies?
Analytics disciplines cover increasingly varied capabilities:
  • Content analytics can help to generate insights into entities and their relationships across very large amounts of unstructured textual data, such as arrest or incident reports, after action reviews, etc. It is very time-consuming and potentially error-prone to attempt this at scale manually.
  • Identity analytics is specifically designed to locate all the information in disparate places that refers to a single individual, and make sure that all the right connections are available when and where needed -- the classic "joining up the dots." The importance of this capability is being more and more recognized at the local levels of the criminal justice environment. (It is important, for example, for an officer encountering an individual under certain circumstances to know as much as possible about the person, who may have records in local databases, but also records in other places under slightly different names).
  • Video analytics is a powerful discipline to assist and enhance the use of video information, typically from street, traffic, or vehicle cameras. It can allow real time alerting of various configurable events -- a vehicle traveling the wrong way, a person standing too close to a platform at the railway station, a package being abandoned in a public place, recognizing the location of a license plate, etc. And it can be used in a forensic sense, for example, answering very quickly questions such as "Was a white truck parked at this location at this time/date?" Once video information is in digital form, it can be treated with whatever the latest algorithms might be -- license plate recognition has been around for years.
  • Social-media analytics is a relatively new part of the picture, with the rise of social media channels and the proliferation in society of devices from which to access them. At one level, being aware of the equivalent of "consumer sentiment" can help LEOs to understand what issues the community thinks are important. At another, publicly available social media can support other intelligence-based policing capabilities. In a more challenging scenario, the use of social media by criminals is presenting new opportunities and difficulties for policing (e.g. the rise of "flash mobs" sometimes "flash robs," where group crimes can be conceived, orchestrated, executed, and dispersed in very short spaces of time).

12. How do your solutions cope with the huge volumes of real-time data that need to be analyzed and combined with information in data warehouses in order to provide a complete set of information ?
The world is becoming more instrumented, with therefore more diverse sources of information about the physical and digital worlds providing more information at a pace which we have never seen.  The "information explosion" provides many challenges, and specifically, in some arenas, the pace and volume at which information arrives is staggering and difficult to assimilate. And the universe of data in which LEOs are interested is expanding. Thus, we continue, in research and development, to enable public and private sector entities in many domains to deal with this phenomenon.

One avenue is a set of capabilities that radically extend the state of the art in big data processing. They can deliver a high-performance computing platform that allows applications rapidly to ingest, analyze, and correlate information as it arrives from thousands of real-time sources, at rates up to petabytes per day, across heterogeneous data types including text, images, audio, voice, VoIP, video, web traffic, email, GPS data, financial-transaction data, satellite data, sensors, etc. -- and to do this analysis at speed, leveraging sub-millisecond latencies to react to events and trends as they are unfolding, while it is still possible to improve business outcomes. This we believe is a key direction which technology in law enforcement will take.

13. Can other public-safety agencies, like fire or EMS, use predictive analytics to similar effect? Do they? How is the ability to share information impacting other city agencies and departments (such as water, transportation, social programs, etc)?
Predictive analytics can offer benefits to other public-safety agencies and beyond. For example, it seems that in some cases fire departments can use these techniques to suggest where fire events are more or less likely to occur, and can thus adjust their inspection schedules, moving from a time basis to a risk basis -- which can target resources more effectively and thus save money.

There is significant scope for other agencies to use predictive techniques as well -- and they are doing so, in diverse disciplines -- from systems doing traffic prediction, to those trying to estimate the likelihood of recidivism in certain kinds of offender.

14. What's the next phase in analytics for public safety -- more use by smaller cities, more use across more variables, more use for reentry and sentencing, something else? Or improved coordination of law enforcement across departments -- even at state and national level?
Several avenues of development present themselves.

Most importantly, we must find ways to "bring the best to the rest" -- by which I mean, if a large, well-resourced city police department can access certain valuable capabilities in the area of predictive analytics, how can we make sure those capabilities are available to the smaller, less well-resourced cities? Aggregation of demand, driven perhaps through private cloud implementations and shared services governance, offers a way forward.

Sharing knowledge in the domain is next. For example, this might include the sharing of predictive models, from city to city, perhaps on a restricted "open source" basis, which may give a department a head start on creating its own models.

Sharing skills and experience is also important. Not every department can afford to devote resources to the analyst function or to the development and maintenance of models. One aspect might be enabling collaboration between law enforcement and academia, perhaps a local college or university with criminology or sociology skills. (The renowned Blue CRUSH program in Memphis, for example, would not have achieved what it has without this kind of close cooperation between the PD and University of Memphis).

Building broader models of a city, say, may well drive interesting insights into how different aspects of city operations interact and affect one another. The dynamics of the "system of systems" which is an area very much worth exploring with these and other techniques in the future.

Infograph_RightRail_Promo.jpgDiscover more. Click to view our infographic.

Making Sense of Big Data to Fight Crime

Police work is still about catching bad guys--but it's also about understanding crime patterns and acting proactively to prevent incidents before they happen.

A New Tool in Law Enforcement's Arsenal

Technology helps law-enforcement officers better predict where crimes will occur based on historical information and trends.

Join the Discussion

The Atlantic does not moderate these reader comments, except to the same extent comments are moderated pursuant to the Terms & Conditions generally applicable to all content on The Atlantic’s sites.
After you comment, click Post. If you’re not already logged in you will be asked to log in or register. blog comments powered by Disqus