The U.S. Department of Veterans Affairs has been collecting data for nearly a century. But Dat Tran, the VA's deputy assistant secretary for data governance and analysis, says it's not enough to just collect data. He says the VA needs to transform its data into information that gives decision-makers insights into veterans' needs. Tran is leading the effort to do that, to establish strong data governance and apply predictive analytics so the department can better serve the country's veterans. It's a monumental task, but Tran says it's essential so the government can help its veterans in the most effective, efficient manner possible. Here he discusses the challenges of big data at the VA and elsewhere.
Favorite tech toy: Samsung tablet
Do you have any outside interests? "Both of my kids are trained in tennis, so I spend all my time [away from work] playing tennis with them and driving them to tournaments."
What's on your reading list? The Bible, The Signal and the Noise: Why So Many Predictions Fail -- but Some Don't, by Nate Silver, and Naked Statistics: Stripping the Dread from the Data, by Charles Wheelan
Hometown: "I was born and raised in Vietnam, but I consider my U.S. hometown Columbus, Ohio. That's where my family first settled after the Vietnam War. We were on one of the last helicopters out of the U.S. Embassy [in 1975]. I'm a big Buckeyes fan."
What's the most common problem with big data? We get overwhelmed with it, because we don't really know what we have from an organizational perspective. We have a lot of silos of data, but we don't look at it or understand it. A lot of organizations I've dealt with, the data we capture and use is for operational purposes, but we don't look at it from a business-process perspective and understand how to use it for better decision-making. And many organizations aren't organized to have true data governance to manage and govern the data and information they have. When I talk about data governance, I'm talking about how to deal with people, process and technology.
What does good data governance look like? The framework for good data governance must address four critical domains: infrastructure, knowledge process, human capital and the fourth, the most important, is the culture. We need to have the infrastructure in place, the software, the hardware, the network. And databases need to be streamlined to eliminate redundancy and enable us to have a single version of the truth. The knowledge process: We basically need to shift from an operational focus to a more analytic focus, not only to report what has happened and what's currently happening but to use what we know to apply predictive analytics to understand what will be happening in the future.
With human capital, oftentimes we see people just looking at data from their perspective, they collect and update data, but we need to look at it at an enterprise level. And many don't have a culture that places a high value on data and information. We need to instill a culture where data is viewed as an asset.
What was the challenge you faced when you stepped into your current role? When I came into the job, one thing I learned is that we as a department have a lot of data. Big data is not new to us. It's been around a long time in the VA, but we didn't have a customer-centric view of our data. Most of our customers are veterans and family members, but we don't have an information environment that allowed us to see a 360-degree view of veterans we are serving and veterans we aren't serving, i.e., those who aren't applying for VA services. And if we don't have a 360-degree view, it makes it tough to do proper planning and proper budgeting. And it was very tough for us to do good analytics on the veteran population.
Was this a people, culture or technology issue? It's a combination of all of the above. The Veterans Administration was first established 83 years ago, in 1930. The department started with one or two business lines, but eventually Congress and this country added on more services for veterans. So as a result, what we ended up having is an organization with many business lines that are basically stovepiped, so the culture as well as the technology were built and grown from a stovepipe fashion.
It was no one's fault, but what we recognize today is we need to be able to take care of our veterans from a veteran's perspective. When veterans come to the VA, they don't look at us as veterans healthcare or veterans home loans. They look at us as the veterans department. But we don't have the data environment to support that. We don't have that one integrated data environment that allows us to see a veteran as one individual. That's what we're working on with the Customer Data Integration initiative.
How far are you into this initiative? We're about to wrap up the work of our skunk works with recommendations to the senior leaders on what the needs are for the upcoming fiscal year, what pieces we need to start putting the big picture together.
You've talked about the importance of a chief data officer. Do organizations really need a C-level executive to harness the power of their data? Personally, I believe that an organization needs to have a chief data officer as a dedicated role that reports to the highest possible level, either the CEO or the COO. The chief data officer should be someone who knows about the business and the technology.
I work with folks who claim to be data people, but they're data people who have no business experience, they're more experienced on the IT side, or they have no experience with the technology, they only understand the data from the business side. The chief data officer would be someone who could speak to both the technology folks and the business folks and serve as the link that pulls business and IT together.
Are you seeing more of them? I believe you're starting to see some. There are a couple progressive companies [that have] chief data officers. There are some where the CIO and chief data officer are the same person.
This story, "VA data guru Dat Tran on turning data into information" was originally published by Computerworld.