IRS’ approach to big data focuses on business outcomes

The IRS is stopping fraud and improper payments not by gathering new or better data. Rather, the tax agency is using data as a strategy of change.

Dean Silverman, the IRS’ senior adviser to commissioner for compliance analytics initiatives, said the bureau is going beyond predictive analytics to improve its business outcomes.

“For example, on identity theft fraud, it’s not so much about creating a new model or a new tool (it is about that), but it’s also about creating a strategy that holistically eradicates identity theft fraud altogether,” Silverman said in an exclusive interview with Federal News Radio before his presentation at the Predictive Analytics Conference sponsored by Elder Research in Washington. “We are looking at authentication and how to test those kinds of models. We are looking at what are our internal policies, how long do we hold a return before we pay a refund? We are looking at other authorities we might have in terms of accessing data. So it’s a strategic approach that is amplified by using data and analytics of which predictive models are a part.”

Silverman took on this task of developing analytical capabilities for the IRS about two years ago with three specific focus areas:

  • Stop fraud and improper payments;
  • Reduce the tax gap between the number of people who are paying taxes and the number of people who should be paying taxes;
  • Ensure core compliance with tax rules and laws.

“I was brought in by the previous commissioner to basically take analytics to a new level, which is tying it more to strategy and operations than just an analytics answer or piece of analysis,” he said. “Strategy always begins to me with what outcome do we want? You work backwards from what outcome do we want to how am I going to get there and how much can data and analytics solve that problem? At the end of the day, data is a raw material. Analytics makes it into a product. But that product has to have a name or use, and that use has to be an improved or accelerated outcome.”

Silverman said his office’s goal is two-fold: create material improvements in each of those three focus areas and create a data driven decision-making culture in the IRS.

“What I think is really important is that if you run analytics separately and you run operations separately, what you will wind up with is an analytics organization and an operations organization. The two shall only meet by throwing things over the fence. It’s not a productive model,” he said. “A productive model is to embed both operation and analytics together and work together to solve a specific problem. Sometimes that involves multi years of testing or multi years of implementing.”

Saving billions of dollars already

He said his office’s role is three-fold and atypical because it involves analytics, strategy and implementation.

Data and analytics aren’t new to the government. More and more agencies are taking advantage of the new tools and technologies, such as cloud computing, to address challenges, whether it’s improper payments or just better strategic planning.

But the IRS is at the forefront of many of these government efforts, both because of the length of time it has been doing this type of work, and because of its mission that brings in huge amounts of data.

Silverman said the IRS is seeing the fruits from the tree planted several years ago.

He said the IRS has used these approaches to stop billions of dollars in fraud, specifically identity theft, over the last three years.

The data and tools have helped the IRS become more aggressive with combating identity fraud. In 2013, the latest data that’s available, the IRS initiated 1,492 investigations, up from 276 in 2011. Of that 1,492, law enforcement recommended 1,257 for prosecution, up from 218 in 2011.

The IRS hasn’t just focused on fraud, but improper payments more specifically, such as using the comparative analysis approach to find fraudulent tax return prepares around the earned income tax credit.

Silverman said the IRS saved or recovered more than $2 billion over the last three years because of the data and analytics.

Always in need of new tools

IRS business offices have recognized the value of comparative analysis office. Silverman said initially he started small and showed value in exploring all the things data and tools can do. But now, the comparative analysis office has more business than it can handle.

“The key was demonstrating value … did you deliver something new and better to the bottom line and in substantial proportion?” he said. “My organization looks at the really big problems. By doing that and showing progress, you get people to believe in things. You get converts into the value of data and analytics.”

Silverman said his office’s role isn’t always to do the projects, but give the business operations some advice on how to improve their outcomes.

“We want to build that analytical culture so that it becomes second nature and everybody is using and appreciating analytics,” he said.

Even three years into the effort, the IRS is far from perfect. Silverman said the agency always is looking for new tools, especially as vendors continue to improve them.

He said last year his office did some market research for a new network analysis tool by asking five vendors to provide their technologies. Silverman said two dropped out almost immediately and two others showed great promise.

“When we tested the tools, we found their results on non-compliant taxpayers different. So we asked why? Maybe there was room for two tools?” he said. “In the end, we now use the best of both worlds.”

Silverman wouldn’t confirm which vendor or vendors won the contract. In 2011, the IRS hired SAS under a $6.25 million deal for its analytics software to help it reduce tax refund fraud.

“The tools are increasingly important. I liken it to an arms race. We are no longer dealing with one-offs and folks watching cartoons doing identity theft. We are looking at organized rings, and they have sophisticated tools. And they are getting more sophisticated, so we have to get more sophisticated,” he said. “We moved from simple rules to more complex rules to clustering to linking of the clustering to a new approach we are going to try this year, which is return DNA. It is an arms race in a fundamental way, and we have to keep investing in tools.”

Silverman added the IRS also has to invest in training of employees to be problem solvers, not just analysts.

“I think you can find all manner of skills. The question is, what is the right mix?” he said. “If we define it as data, you’ll wind up with data without analysis. If you define it as great modelers, you’ll have analysis without someone taking care to make sure the data is correct. Or if you are doing project management, you’ll have people who can manage processes, but they don’t have the wherewithal to manage the outcomes. So you need to put those skills together in a team, and you need to build that same kind of capability inside operations for the kinds of problems that operations should be solving by themselves.”


IRS struggling to tackle massive surge in identity theft

SEC employees harness big data, cloud to open up savings, new possibilities

IRS to use more data to prepare for long term budget uncertainty


Sign up for breaking news alerts