DLA’s first data scientist to help solve problems by blending traditional research methods, new technology 

By Beth Reece

PRINT  |  E-MAIL

Ten years ago, people were poking fun at Jesse Rowlands, the nerdy kid in the corner hoarding data and running statistical analyses. The laughter has stopped. As the Defense Logistics Agency’s first data scientist, he’s suddenly a big deal. 

“That’s how it was for a lot of current data scientists up until three or four years ago. Now, we’re important,” he said. 

DLA employees have used various research methods to solve problems for years, but maturating science and technology tools present new ways of using increasingly complex data and computing practices to work through tough issues. 

“Traditionally, we’d do simple analysis to solve simple questions, but now the questions and the data have gotten so complicated it requires a lot of extra steps,” Rowlands said. “Today, we create hypotheses and apply machine learning algorithms. It’s an advancement in analytics.” 

The DLA Information Operations employee expects to use his knowledge in mathematics, statistics, computing and database management to help the agency solve problems like demand planning. To determine what customers will buy next week, next month and even six months from now, Rowlands will work with demand planners for DLA’s supply chains and information technology experts to predict patterns in customers’ orders. 

“With the robust data and systems we now have access to, we can ask these hard questions; but it’s going to be a really iterative process and we’ll have to start small,” he said. 

The process will begin with limited experiments, algorithms and data sets. Additional data and prediction intervals will gradually be added using trial and error to increase the problem’s scope. 

“You go back to the drawing board and say, ‘OK, based on what we learned in this experiment, what data, attributes and constraints do we need to include going forward?’ Maybe we need to add new data that costs money and introduces new constraints,” he said. “Or maybe we need to reduce the scope of the problem from six months to three.” 

The goal is to build an iterative cycle where each study becomes the basis for the next one until a robust model is reached. While people often expect quick answers and solutions from data science, Rowlands said it can be a slow process of baby steps that eventually lead to progress.

Despite the cumbersome process, results from data science can be a game changer that transforms entire offices, he added. 

“I’ve deployed tools that have doubled the output of 10 to 20 people or quadrupled their success rate. If you do enough of those around the agency, you eventually get a bigger impact that adds up to grand results.”

Problems at grassroots levels don’t usually rise to senior-leader attention, Rowlands continued, but are exactly the kind of issues data scientists can help solve before they grow into agencywide obstacles. 

“Educating the workforce and managers who are actually doing DLA’s work on what machine learning and data science can do for them will help the agency generate ideas for people like me,” he said.

Though leaders are eager to see the impact of data science in areas like demand planning, the agency is still in the infancy stages of adapting it. 

“The data science capability builds upon IT fundamentals, so we’re making sure we have access to the right data. We’ll get to specific problems like demand planning, but it’s going to take a while as we build the foundation.” 

Rowlands began his career as an internal auditor for the Defense Commissary Agency, where he used predictive models to determine where people were stealing and detect those conducting fraudulent transactions. Before coming to DLA, he was an operations research analyst at the Defense Contact Management Agency. There, he led teams through machine learning projects to create predictive models that solved DCMA business problems.