igor - Fotolia

Public-sector analytics teams struggle to implement innovation

Bureaucratic rules and risk-averse culture make it difficult for public-sector agencies to take advantage of data-driven decision-making, despite clear benefits.

Shortly after Robin Thottungal took over as chief data scientist and director of analytics at the federal Environmental Protection Agency (EPA) in 2015, a staff member wrote a phrase on a white board just outside his office: "Culture eats strategy for breakfast."

This one quote neatly describes many of the challenges that public-sector analytics teams face as they try to move into the new world of data-driven decision-making. They may have the will and the tools in place, but without changing the culture -- which can be a challenging and time-consuming task -- it can be hard to move the needle.

"Disrupting public-sector behavior isn't easy, but we're making progress," Thottungal said in a presentation at the recent Big Data Innovation Summit in Boston.

For Thottungal, part of the effort to create change started with focusing more on building data tools. He said that in the past, it could take up to five years for federal IT teams to build an application due to heavy bureaucratic rules. And while that's good for some of the EPA's data use restrictions, such as privacy and accountability, Thottungal said these rules were too often used as an excuse for not doing anything.

Rapid development breaks barriers

Thottungal said he has implemented a rapid development approach that values building and releasing reports and data products, even if they aren't perfect. Getting them in the hands of staff members shows that the department is capable of innovating and taking risks within certain boundaries.

One example of a new product his team has created is a dashboard that utilizes the agency's geospatial data of the Chesapeake Bay to monitor changes in pollution levels. The dashboard is designed to let program managers identify troubled areas and direct their resources accordingly.

And this is just the start. Thottungal said he's trying to move the department to embrace more emerging analytical methods like machine learning and artificial intelligence, and to rely less on basic descriptive statistics like mean and median.

"What I've seen happening at EPA is the way we're able to change is through technology," he said. "What I've seen so far is technology enabling innovation."

Data is being collected all over the place. Using it is a different story.
Richard Culattachief innovation officer, state of Rhode Island

Richard Culatta, chief innovation officer for the state of Rhode Island, said in a presentation that he's also trying to encourage a rapid prototyping environment. Like Thottungal, he's seen many projects take so long to get developed that by the time they're released, they're irrelevant. Even worse for public-sector analytics teams is when a product takes a year or more of planning and development, only to find out once it's released that a developer made an incorrect assumption about a data source that invalidates the tool.

Culatta said this is why he favors a minimally viable product approach. It makes it easier to spot bad assumptions early in development. And if a product fails, it hasn't wasted very much time.

He compared his approach to using data within the state government to that of the Lockheed Martin Skunk Works team that developed some of the nation's most advanced military aircraft. That team was set up to move quickly and innovate rapidly. With new sources of data coming online and new methods of analyzing that data evolving rapidly, Culatta said it's imperative for public-sector agencies to take this approach.

"How do we create the Skunk Works of big data?" he asked. "Data is being collected all over the place. Using it is a different story."

Leverage skills on your analytics team

For Justin Antonipillai, a counselor at the U.S. Commerce Department, the key to moving public-sector culture is putting people in a position to take full advantage of their skills. He said since he's been at the department, he's seen highly trained data scientists being asked to do simple data reporting tasks. This stifles innovation by making it hard for the most skilled workers to engage with the trickiest and most impactful problems.

"We had world-class experts being asked about the type of [Microsoft] Access database they should set," he said.

But now, leadership is looking for more difficult tasks to give to data scientists. For example, his team recently led an initiative with the Commercial Foreign Service department, which is set up to help U.S. businesses do more business overseas. The goal of the project was to identify companies that could be exporting more. Before, program managers would simply reach out to businesses offering assistance based on their intuition. Now, the department has algorithms that identify companies that aren't exporting at capacity.

Thottungal said the public sector is generally averse to taking the kinds of risks required to change and embrace new data-driven methods. But to stay relevant in a changing world, agencies are going to have to. "The way the public sector can change is to have people come in and say, 'Let's take some risks within the boundaries,'" he said.

Next Steps

Efficient infrastructure is key to public-sector analytics success

BI in the public sector is different than in the private sector

Don't let management keep your company from making data-driven decisions

Dig Deeper on Big data analytics