ICT is widely viewed as enabler of productivity improvement, but new technology adoption is often a horse of a different colour. The process involves a complex interplay of business drivers, IT strategy, organizational priorities and product momentum, weighed against IT budgets, refresh cycles, legacy investments, implementation bandwidth — and just plain old inertia. In public sector organizations, procurement processes and the need for broad scale departmental coordination can introduce extra tangles to this web. What tips the scales? In the case of the Vancouver Police Department (VPD), it took a crisis, strong leadership by management with tech vision, a willingness to push institutional boundaries and a solution that was easy to implement, accessible to a range of users and that could show tangible benefits.
Crisis drives adoption
With the help of IBM’s i2’s Analyst Workstation software suite, the VPD has deployed C.R.I.M.E. (Consolidated Records Intelligence Mining Environment), a solution designed to provide the force with greater ability to monitor, predict and mitigate the impact of crime in the region. VPD special constable Ryan Prox, who managed the implementation, has outlined three key adoption drivers. “The aftermath of the Pickton inquiry and the public backlash against what had happened with Lower Mainland policing” was critical in his view, as one of the issues it raised was the lack of multi-jurisdictional analysis capability. “We were operating in a Vancouver Police department bubble, as were other jurisdictions… but we were dealing with a serial killer that was operating at the metro, Lower Mainland level.” As an alternative to unification of various jurisdictions under one combined management structure, the VPD opted for development of analytics capability that could ultimately scale province wide. The solution was created in the VPD, Prox explained, since the agency was smaller, “more nimble, and so better able to try out new initiatives fairly quickly.”
According to Prox, another challenge which emerged from the Pickton inquiry was the fact that the VPD investigation components and management structure “were not very technologically literate.” Management change at senior levels in 2007 — 2008, however, helped transform the force from one that was viewed as not very progressive to one that would serve as the source of technology enabled policing. In 2007, the Chief Constable’s position was assumed by Jim Chu who became involved with the VPD’s E-Comm project for transition to a new radio system, installation of the PRIME-BC (Police Records Information Management of BC) system and a new mobile computing and data access platform. In 2001, Chu authored a book entitled Law Enforcement Information Technology. Current executive level deputy chiefs are also more inclined towards “risk taking in this area” in Prox’s view, or at least more determined to leverage technology, and the force also benefited from the hire of a couple of “high flyers” in enterprise data base SQL definition, design and management. “This top-down cultural shift really had an impact on the organization’s approach to the use of technology,” Prox added.
A third driver was the need to have up to date policing services in place for the Vancouver Olympics, which Prox observed “really put things into high gear and added a time clock” for implementation. Combined, these three factors enabled the VPD to move forward at an accelerated pace with a new intelligence and analytics infrastructure.
Rolling out i2 analytics
The VPD’s implementation was staged: the first system was built for use by the VPD; this was extended next to the metro Vancouver level, and ultimately province-wide data and records management were incorporated to extend access even further. “The technology implementation was the least of my challenges,” Prox noted: “we went from nothing to something in two weeks.” The initial installation was delivered by IBM i2 system engineers and designers along with VPD IT staff, who supplied technical skill, experience building police security requirements and the knowledge of operational processes needed for successful implementation. This group commandeered 16 on-site computers and were able to stand up a working system in record time.
Gaining access to information at the metro and provincial levels proved a bigger hurdle for Prox: “in British Columbia, we have a multitude of layers of bureaucracies that deal with access to information, such as Prime Corp. which oversees PRIME-BC, which in turn administers the provincial records management system.” To encourage buy in and cooperation from these various agencies, Prox pointed to the need to turn learnings from the failed Pickton investigation into positive change — in the media sensitive, post inquiry climate, no one wanted to accept responsibility for impeding progress on the development of more collaboration, greater efficiency and better results in policing.
The new i2 analytics solution was installed in the force’s main IT server room at headquarters, which acts as a secure repository for files and records that are handled only by police staff due to their sensitivity. To administer its data warehouse and manage infrastructure and other systems, the VPD maintains 6 staff for helpdesk, a database design engineer, a security specialist and
hardware engineer to manage dynamic system demands on its virtual server environment and its NetApp SANs. The VPD data centre is a Microsoft shop, running SQL, the Microsoft server OS and Hyper-V virtualization software.
As Prox described it, deployment of IBM’s i2 analytics platform into this environment was a relatively straightforward ‘plug n’ play’ process; however, the implementation team had to create an ETL process that was “quite convoluted” for importation and cleansing of data from disparate programs and systems, including Excel, access data, PRIME, and the department’s Computer Aided Dispatch and Computerized Booking and Arrest Systems. The fact that all provincial institutions use Versadex for records management simplified data integration to some extent, though tables still required deconstruction and recompilation in SQL, a task that involved months of advance preparation and was in some cases stymied by vendors who proved reluctant to share proprietary detail on table structures. Prox explained, “We are mapping at a very high level — in the 99.3 to 99.7 range for GIS mapping of crime incidents — so we have a lot of cleansing routines, and holding directories that must be checked manually and regularly for errors. We have a lot of processes in place to support the data warehouse.”
Access to VPD data depends on a number of factors. Some information carries security classifications associated with file sensitivity or a person’s unique security clearance, but access by role, departmental unit, and investigative team also have to be accommodated. In some cases individual teams may be authorized to see only a subset of the data, and access permissions can change on a daily basis. Prox noted: “we are pushing the limits of what software can do on role-based access. It is very difficult to set up security provisions using traditional software” since this depends on the security level of the information itself, staff role, personal security clearances and association with a particular unit or case. The VDP has found elements of required security in the Versadex software, in the ibase database application which offers some security provisions as it imprints its own database structure on top of SQL, and in the role-based security contained in Microsoft’s enterprise server management software.
Remote access to the VPD’s main crime and intelligence analytics platform, backend infrastructure and billions of records that are housed in the data warehouse is delivered to partners, analysts and internal staff through Entrust dual factor authentification via a 256 bit encrypted secure tunnel through the Internet. The VPD has also created a law enforcement graphic dashboard interface (the “GeoDASH”) which draws on parts of the analytics database, transmits this to a DMZ (‘demilitarized zone’ or separate, small network) and from there to mobile device terminals running on 2G cellular networks. This system offers field staff some GIS-based crime analysis and query capabilities through the install and regular update of tile cashe files on local MDTs; however, capabilities are limited by cellular communications, which are not robust enough for data and graphic intensive ArcGIS software.
In its current transition to predictive analytics, the VPD has partnered with academics at the University of Victoria and some smaller ISVs that develop custom apps and interfaces. The force will also leverage a special crime enforcement module of ESRI’s ArcGIS technology, which integrates into i2 data mining and query tools and with Analyst Notebook, to create predictive policing based on geo-spacial analysis. “We’re leveraging our data mining capabilities, our data warehouse and pulling that information into an ArcGIS Server environment and creating an interface that automates this predictive capability and pushes it out to our mobile terminals,” Prox explained. This project is now in beta testing, running algorithms based on property crimes, and Prox expects to soon equip patrol officers with a new version of the GeoDASH interface that will enable them to self deploy to crime locations predicted for the future. Unlike similar implementations in Los Angeles and the U.K., the VPD will not deploy an analyst team which delegates tasks to patrol officers, but rather empower individuals with the tools they need to monitor predicted crime hotspots within their own areas of responsibility.
The VPD was attracted to the i2 analytics solution for its ability to address a key issue in policing: need to manage the data deluge which is now part and parcel of disclosure requirements in modern investigations (in Prox’s last organized crime case, the photocopy budget alone was $15,000, bringing the total budget with overtime for court disclosure prep to $100,000). One adoption goal was to break down the data, distill it into meaningful pieces and build intuitive access to it through data visualization. In Prox’s view, i2 is the leader in this field, “renowned for their ability to do link charting, link analysis and representations of organized crime, telephone calls and interception in a very graphic, visual manner. They were game changers in the intelligence and law enforcement industries in terms of their ability to pull out present information in a nice, clean easy to read chart.” The i2 solution also provides a data mining tool that Prox described as similar to “writing SQL script using wizards. It is a very visual, drag and drop query building tool” that the non-IT specialist can use to write multi-level, multi-table relational queries across databases, and which generates pectoral queries/results that are easy for a broad audience to comprehend.
When Prox was assigned responsibility for the VPD’s analytics program, the force employed two analysts. Today, that number is 25 — all civilian specialists who have met stringent hiring criteria which Prox claimed are the most demanding in the industry. Interestingly, though these “best and brightest” are constantly working on ways to leverage the technology,” they do not necessarily have a tech background, but may have specialization in a wide range of disciplines, including geography, math and criminology. Critical thinking skills, the ability to frame the right questions and to operationalize the results are favoured in hiring, a luxury that has been made possible through i2 solution accessibility. Prox explained, “We have built the technology and infrastructure in such a way that you don’t need to be a computer programmer to use it. I can teach a constable with cursory knowledge of his iPhone how to do our data mining in an afternoon.”
With the analytics solution in place since 2009, the VPD now has several years of experience using the system under its belt. Measuring ROI is a different game in a policing environment, but Prox did point to productivity metrics that are appropriate to the organization — high profile breakthrough cases that have been attributed to analytics. The capture of Ibata Noric Hexamer, a rapist and child predator who was operating multi-jurisdictionally, “pretty much sold the system,” he added, because the force had investigated the case for a year without results, but after applying analytics was able to identify the offender in six weeks. The department had a similar result with Shalendra Kumar Sharma, a sexual predator who had been operating for 10 years and was escalating the violence of his attacks, before being identified through application of analytics capabilities. The VPD now also reports on patrol deployment and resource efficiency — an exercise that was not possible before implementation of the analytics platform, which has provided a systematic view of organizational benchmarks and deliverables. Analytics have provided the basis for monitoring key performance indicators, such as solve rates, 911 response times, the achievement of service delivery goals, resources for patrol units, measurements that can serve as the basis for performance improvements.
So what key takeaways can be applied by other organizations looking to adopt analytics capabilities? While the VPD situation is unique, its implementation highlights the importance of critical business drivers and strong executive sponsorship. Advance preparation was also key — in terms of ensuring institutional collaboration, building data integration and defining security protocols. Deployment of the right tools was also important: with i2, the VPD was able to quickly deploy a law enforcement module that was accessible to non-IT staff, allowing these to focus on policing rather than tech questions. With analytics, Prox believes the VPD’s ultimate transformation to “an intelligence-led police organization” is within reach.