As one of the founding members of the State Chief Data Officer Network, Tyler Kleykamp talked to GT about the increasing importance of his role as state and local governments ramp up their use of big data.
The role of the chief data officer has grown increasingly important as more state and local governments ramp up their collection and use of big data for a range of problems, from criminal justice and health care to transportation and community engagement. CDOs in local government have been in the forefront with their initiatives, but states also have CDOs and their numbers are increasing. At last count, 18 states and the District of Columbia had a CDO, according to the Pew Charitable Trusts.
One of those is Tyler Kleykamp, who became Connecticut’s first CDO in 2014. He also is one of the founding members of the State Chief Data Officer Network, a voluntary, self-organized group of state CDOs, similar to the Civic Analytics Network, which represents city CDOs.
Most of us are the first person to do this job in our states. There’s no blueprint to follow; nobody has preceded us to explain how it’s to be done. There are a lot of unknowns. It’s also very different
in the public sector compared to the private sector. Having a peer network, where you can ask someone questions, get advice, bounce ideas off somebody, is really important. It not only helps us do our jobs better, but also it appears that states are going to continue to hire CDOs. We keep seeing more states add the position — Florida, Vermont, Virginia are recent examples — so as these people come on board, it’s helpful to have a group to work with.
A lot of this starts with a use case or problem statement. For instance, if we are looking at an opioid issue, we might start at a high level — address the number of deaths, for example — and look at what we have: what different systems might agencies have; what’s the data we might have; what’s the relationship between all of this?
For example, we have data on overdose deaths, which in Connecticut is fairly public information. There’s also a criminal justice policy division within our office, so we are able to take that and look at people who have been involved in the prison system. One of the things we found through matching records is that in 2016, about half the people who died from an overdose were at some point involved in the prison system [in Connecticut]. That tells us there’s another place to dig in and start to plan how to help these individuals, such as in re-entry programs.
We are decentralized here. Our IT agency provides enterprise services, and they have a role in procurement, but there are some things that agencies manage on their own. That presents some challenges, where it might be easier if it was just the CIO and me who figure something out. But the reality is I have to work with the agencies.
My role is a little less involved with some of the back-end data management components. It’s a little more high level for us, so we’ll issue a policy or administrative rule that says, “This is how you should do things to ensure some best practices,” but we’re not getting into the weeds with everybody on how they follow that practice.
There are always legal restrictions on what we can and cannot do with some of the state’s data. ... If we can talk to other states through the network about how they are able to use some data in a legally permissible way, or work within constraints of a universal law to do critical work, that can be really helpful.
Another challenge is data quality for analytics. In the private sector, there is a concept emerging called data ops, which takes the principles of agile and applies them to analytics. I’ve been working to modify that to be more applicable to government. It starts with a series of principles you operate under. Next, you want to involve everybody that touches data. … There’s a lot of work in government that goes into making data suitable for analysis, so there are issues around older systems and data quality. This is a way to address that. Data ops starts with some simple questions and then uses the data and finds the best way to address those data quality issues. By taking a team-based, agile approach to it, you can get things off the ground sooner in terms of data and grow that over time.