Analytics professionals spend more time gathering data than they dedicate to analyzing it, according to a recent survey.
Gathering and analyzing data to drive decision-making in government has received ever-increasing focus, and this trend will continue with the enactment of the Foundations for Evidence-Based Policymaking Act. But—funny thing—there isn’t much evidence about the use of analytics or its value.
With that gap in mind, we set out to gather information about the use of analytics in government, using a survey as a starting point, with support from ACT-IAC. We wanted to find out: Does the government get value from analytics? If so, how much value? And how does it get the value?
Findings from our research were presented Oct. 2 at Johns Hopkins and were the basis for an illuminating discussion by a panel of agency experts.
4 Key Findings:
1. When making significant decisions, most agencies use analytics. For the three most significant decisions made by their agency in the last year, 82% of those respondents say analytics were the “dominant” factor, “of significant importance,” or at least “equal to other factors.” Analytics, therefore, seem often to guide strategic decisions.
2. Analytics professionals spend more time gathering data than they dedicate to analyzing it. Survey respondents reported spending 23% of their time gathering data, which is more than they spend analyzing data (8%), communicating results (14%) or acting on the analysis (10%). This finding suggests that streamlining and automating data collection should be a priority. Panelist Michael Conlin, the Defense Department’s chief data officer, suggested that better data sharing may be a time-saving solution—while data-sharing may run counter to the “loose lips sink ships” ethos, broader collection and collaboration efforts could yield markedly better outcomes.
3. Artificial Intelligence gets a lot of talk but not much action. Survey respondents agree that machine learning (26%) and artificial intelligence (25%) hold the most promise for improving government, but very few respondents use AI right now (less than 4% reported doing so). Panelist Patricia Hu, director of the Bureau of Transportation Statistics, said that this deficit could be addressed using "reverse mentoring," meaning senior officials are paired with new analysts, who can educate senior staff on emerging techniques such as AI and machine learning.
4. Staffing is the biggest hurdle to the government’s ability to gain more value from analytics. Attracting/retaining staff was named the biggest challenge in government analytics; and, more than 50% of respondents indicated that their agency plans to develop or recruit staff with a specialized certification in analytics. For new staff and continuing, panelists advised that agencies should give employees the tools and space to explore and experiment with new approaches if they want to retain talented analytics staff.
This is mostly good news: Agencies are using analytics and getting value from doing so. Nonetheless, the results suggest several areas where government can improve:
Create an analytics staffing plan and strategy. Give more thought to how to compete for talent, and how to develop the capabilities of agency employees. Panelist Collin Paschall, senior lecturer in government analytics at Johns Hopkins, explained that millennials are eager to develop their analytic skill sets, but expect to hold many different jobs over their careers. Agencies should foster a culture that encourages employees to expand their toolkits and even participate in rotations that allow them to experience a variety of analytics roles.
Streamline/automate data collection. Invest in standardizing data collection, and sharing data—both within and between agencies, as well as across various levels of government.
Get ready for machine learning and AI (really). Don’t just talk about it. Find and plan test cases to unlock the value of these new technologies. Look and learn from new employees, and from peers and in other sectors. One example was shared by panelist William Wiatrowski, Deputy Commissioner at the Bureau of Labor Statistics. He reported that BLS is experimenting with natural language processing (a form of AI) to read routine datasets and generate standardized press releases. This frees up time of the agency’s data scientists to tackle more complex analytics challenges.
Dr. Jennifer Bachner leads Johns Hopkins University’s Program in Government Analytics. She holds a PhD in Government from Harvard. Jeff Myers is a senior director at REI Systems.