DOD CDO shares 7 data management best practices

Michael Conlin, the first chief data officer at the Department of Defense, or DOD, was brought in from the commercial side to offset the enormous complexity in the department's data, which accumulated over 35 years. He's here to modernize and implement data management best practices.

Conlin spoke at MIT's Chief Data Officer and Information Quality Symposium. He said the complexity of the DOD's data systems is a result of historic decisions, legislated requirements and a fast-paced operational environment. Conlin manages roughly 10,000 operational systems, each having its own unique data schema.

At the DOD, he is responsible for the entire department's management, analytics and governance of data to answer critical business questions in support of the National Defense Strategy.

A view of departmentwide data enabled the DOD to compare different parts of the organization's performance, in terms of output and cost, with the objective of being as efficient and productive as the commercial sector. With this, they can determine the data management best practices specific to an organization.

Along with this significant challenge comes new expectations for law and policy. For example, the Foundations for Evidence-Based Policymaking Act mandates that every cabinet-level agency appoint a chief data officer. However, that law included neither specific authority nor budget to get it done.

Additionally, the law requires departments specify learning agendas to support their budget request. All the decisions and policymaking now must happen based on evidence. If you don't have data to support a decision, the law says you must have a formal agenda to get it. This creates a demand for the departmentwide oversight and management of data, which a chief data officer role enables.

To address some of these challenges, Conlin and his team created a repository of curated data to show how his team works with data owners through metadata management. This created a common data strategy and built a common enterprise data store. The team built a database of commercial costs for items commonly purchased by the DOD to compare what the department spends. The database can be queried by a data visualization tool that allows users to see information in real time, overcoming the common idea that says all you have to do is generate a static report to reference over a given period of time. Live data is always best.

"All the business mission exists to create readiness," Conlin said. "So, storytelling becomes very important. Helping people learn how to consume information in a different way is very important, and we had great success with this."

Three major things created significant change. First, the DOD began focusing on delivering a consumer experience -- something more personally catered to the user than, for example, the understanding or opportunities presented on social media or in consumer markets. Second, the team adjusted some of its processes to operate closer to the same efficiency and as the private sector. That explicitly requires the adoption of best practices from industry. Finally, the department implemented a formal mechanism that allows Conlin to bring in people from the commercial sector on a temporary basis to have them work on some of the department's trickier challenges.

Conlin shared some tips for how other organizations can implement data management best practices in a similar way:

Direction is more important than speed. Knowing where you're going is more important than how fast you get there. You will stumble and make mistakes. Conlin said as long as you learn in the process, speed doesn't matter. Direction makes a bigger difference.

Learn by doing. Don't wait for the answers; find them. Become comfortable with that. You have to make learning a measure of success, Conlin said. Perhaps you didn't achieve what you wanted, but if you learned something along the way, the exercise was worthwhile.

Walk toward the uncertainty. There is no cookbook for what data managers do. You have to go invent the art of the possible, Conlin said. If uncertainty scares you, you wouldn't work in an area of data management.

There is no truth in data. There are just a lot of random facts. Conlin said it takes a human being to interpret and understand them. There are a range of quintessential human capabilities that must be brought to bear on data before it has any value.

Data is more important than algorithms. There is no shortage of open source algorithms, and almost all of them are available through libraries in popular machine learning platforms. The data is your competitive advantage.

Buoys not boundaries. "I mark the safe shipping channel, the buoys, from a tools and techniques approach," Conlin said. His team has implemented a variety of tools in production, but he said he still lets individual employees make their own choice with how they want to execute their task. It isn't about limiting people with boundaries.

Automate everything. Conlin advised to write scripts, digital recipes and to automate everything the first time you do it, making it repeatable. "It's never a throwaway," he said.