She knew from the literature that a key to increasing graduation rates is reducing student absenteeism. So by digging into school records, she found the main reason for absences was for health reasons, predominantly asthma attacks. So she took existing data from the schools, emergency rooms and social-services agencies and plotted the absent students on a map, checked emergency medical service calls and other records, and identified which neighborhoods had the most incidences. In response, the city launched a campaign in those neighborhoods, sending out public-health and social-services staff to work with parents to come up with solutions.
So what does it take for your jurisdiction or agency to create a corps of data detectives? A recent report, "From Data to Decisions: the Promise of Analytics," jointly sponsored by the IBM Center for the Business of Government and the Partnership for Public Service, calls on government leaders to employ data analytics and invest in trained data analysts--the people we call "data detectives." The managerial framework for analytics is the use of transparency (which includes accessibility by non-technical staff); accountability (which includes creating a clear "line of sight" so employees can see how what they do fits into the broader picture) and a focus on results.
We have found that there are four actions government leaders can take to use analytics to create data-driven decision-making in their agencies:
Collect better data. The most common approach to increasing availability of data is to compile existing data into a single compendium, or single web portal, and publish it on a regular schedule. But just publishing random data isn't helpful. Agency leaders will need to prioritize their data collection and sharing by linking data to clearly defined outcome goals and identify what performance information is needed to track progress against them.
There will also be a need for leaders to allow easy feedback, via social-media tools, from employees and--where appropriate--the public. This creates an early-warning system to alert program managers about possible performance and data reliability problems.
The best model of a cross-agency, data-intensive web portal that includes interpretive tools is the federal government's Recovery.gov, which tracks the spending of federal stimulus grant and contract money. Many states, and some cities, have created similar websites.
Conduct better analysis. Executives will need to be able to make sense of the flood of new data using analytic tools that can help both decision-makers and the public. This means executives will have to understand who their users are and what kind of data and data displays are useful for providing meaning to the data. Information seen as useful to program managers might be very different than information seen as useful by city council members, legislators or the public.
Many organizations have adopted "dashboards," often using maps, to display information for decision-makers and the public. A good example is New York City's My Neighborhood Statistics, which map all 311 service calls and other city-collected data by neighborhood.
Make better decisions. "The challenge today is no longer in collecting information," says performance expert Harry Hatry. "The challenge now lies in using the information that is regularly collected." He found that the effectiveness of programs can be improved by taking timely corrective action based on information collected.
In the United States, this has fueled a movement to use "performance-stat" systems based on frequent goal-focused, data-driven meetings to support decision makers in reaching priority goals. The "Stat" approach began in cities like New York City and Baltimore and has since spread to states as well as federal agencies. Hatry and his colleague, Elizabeth Davies, have prepared a guide to these data-driven performance reviews that summarizes best practices in conducting such meetings.
Take smarter action. Real-time data should not be collected and used solely to react to past events. Using smart sensors and interconnected data sets will allow more sophisticated analyses of data that are predictive in nature.
For example, Santa Cruz, Calif., launched an experiment in 2010 using large sets of data and a sophisticated algorithm to forecast when and where crimes were most likely to take place. The city then began to deploy police officers preemptively to stop them before they occurred. Does it work? According to ABC News, property crime dropped 27 percent.
Because of the success in the use of analytics, states, localities and the federal government are all engaging in greater use of these approaches. Look for a data detective in a government agency near you.