From the Army to the FCC, agencies are learning how to use analytics as a means to measurable ends.
Where: FCW Face to Face on Big Data Management and Analytics
Who: Col. Bobby Saxon, chief, Force Management Enterprise Division, Force Management Directorate, Office of the Deputy Chief of Staff, U.S. Army
Naba Barkakati, chief technologist, Government Accountability Office
James Miller, associate CIO, Office of Managing Director, Federal Communications Commission
Adam Stalker, National Digital Director, Enroll America
Why: The Army has some 3,500 different IT systems. So for Saxon, who is tasked with pulling all relevant data into that service's Enterprise Management Decision Support system, the challenge is obvious. "If you want to use that data," he said, "it's simple -- all you have to do is contact 3,500 different people."
Saxon's job, of course, is to provide senior Army leaders with actionable analytics while sparing them all those data calls. At a Dec. 3 FCW event on big data and analytics, he and the other speakers discussed the challenges of wrestling disparate data sets into usable form -- and of making sure an agency's analytic efforts are being driven by the mission, not just big data for big data's sake.
Stalker said that his group -- a nonprofit working closely with the administration to identify the uninsured and steer them to HealthCare.gov -- had been guilty of letting big data "drive our engagement, rather than just inform it."
In 2013, Stalker said, Enroll America was able to use predictive analytics and big data to identify those likely to lack health insurance -- with such precision that they could target individuals for personal outreach. After a few months, however, it became clear that this precision was expensive and unnecessary. The vast majority of the uninsured are congregated in specific counties, and even specific ZIP codes. It was much more effective to focus on those communities as a whole and "just be there," Stalker said. But initially, "we were just so excited about going and finding that individual and knocking on that door."
For Miller, the big challenges have come in the form of government-specific constraints. When the FCC began working with nationwide broadband data, he said, there was "a lot of policy and law involved. ... We had to think about [Paperwork Reduction Act], and about the Privacy Act."
And beyond the letter of legal obligations, Miller said, the government faces the practical hurdles of individuals' "Big Brother" concerns. Simply saying you will anonymize the data is not sufficient, he said. "You're going to run into privacy issues if you're doing anything interesting at all with data."
Agencies are learning, however. Barkakati cited inter-agency discussions among inspectors general about analytics-driven audits, and the Centers for Medicare and Medicaid Services' use of predictive analytics to better process claims. Such knowledge-sharing is key, he said: "Tell people how you're succeeding. Articulate the value."
Saxon concurred, adding that with predictive analytics, it's critical to manage expectations about how well-informed and reliable those predictions will be. "We share that information with our senior leaders to help them make more informed decisions," he said. "When you're using predictive analytics in my world, it's extremely important to make sure that the stakeholders know exactly what they're getting."
Ultimately, Saxon said, "it is not about the data. It is about the solution to the problem. Stay focused on that."
NEXT STORY: Make open data actually work