As companies use big data to better inform decisions, they create more demand for professionals able to grapple with those databases and turn them into assets.
Indeed the competition for recruiting big data science talent is brutal and likely to stay that way for a while. But opportunities exist for CIOs to tap the skills of people in their enterprises.
"This is a massively expanding field and it's going to be that way I think for the next decade," said Fritz Schlereth on a panel at the Global Artificial Intelligence Conference in Boston last week.
Schlereth is head of product at Descartes Labs, a geospatial analytics startup spun out of Los Alamos National Laboratory. He was among a panel of AI experts at the conference who offered advice on how to identify data science talent and how to manage data science teams to yield business results.
Internal data science talent
One way to acquire data science talent: Look homeward. Solution architect Craig Rowley joined the Columbia Sportswear Company in 2017 with the purpose of building a data science team for the $6 billion retailer.
"This is the first enterprise company I worked at where I actually get to see that phase where we're discovering who those people are that are multidisciplinary," Rowley said.
The people at Columbia who have demonstrated the ability to solve big data questions share certain attributes, Rowley told the conference audience. They typically possess a deep domain expertise. That expertise allows them to define a problem worth pursuing, he said, and to understand the kind of data they will need to solve it.
The case for emotional intelligence
When recruiting big data science talent, John Mercer, the head of data science for the video advertising company Pixability, looks for emotional intelligence first -- and for people who would work well in a team environment.
"You want people that are kind, that are thoughtful, that are ultra-collaborative -- people that you want to show up to work with every day, people that you know you can trust, that you know they have your back and that are going to be mutual champions for you as you support each other," Mercer said. "If you don't have that, it doesn't matter how much math you know, or how much statistics you know, or how good a programmer you are because things will fall apart."
The right data science team for a company depends on the organization's size and what it is hoping to accomplish, Mercer said.
"It's really dependent on the context. It's dependent on the industry. It's dependent upon: What is the mission that you're given and the current state of the company and how big it is," Mercer said. At early stage companies, the lines of delineation between different roles are "blurred," he said. As companies grow, the data science talent become more specialized with more clearly defined responsibilities.
Integrating data science talent
At Wolters Kluwer, a global information services provider, director Syed Haider looks for a variety of backgrounds in machine learning, natural-language processing and statistics, along with other more universal attributes.
Syed Haiderdirector, Wolters Kluwer
"We look for basic skills like programming. We look for motivation. You have to be highly motivated because we call it data science, but it's actually data art," Haider said. "Every problem is unique"
Data science teams should learn about how their output will be used in the business and members of the team should interact with the end users, according to Descartes' Schlereth, who said it is a "mistake" to "insulate" data teams from customers, whether they are internal or external.
At Descartes Labs, there is a separate team that works on the data infrastructure for the data scientists, and the scientists are encouraged to be "vocal" about their needs, Schlereth said.
"A lot of data science problems are front-loaded in terms of the data prep and a lot of data handling -- bits and pieces themselves -- and quite honestly that's not the sort of training that goes into being a data scientist," Schlereth said. "The data scientists focus on actually solving the problems and understanding the data itself, and then understanding algorithms that help automatically extract insights from that data. They're not cloud engineers."
The data scientists at Descartes are also encouraged to reach out for help when they become stuck on a problem.
Culture of communication
"They're not going to be comfortable with that unless you've really crafted the culture such that they're invited every day to reach out and say, 'Hey, I want to open this problem up to the floor,'" Schlereth said.
Data science teams are distinct from others areas of an enterprise in that they run experiments and what they create might not look as tidy as other products. That means that the data science talent in your company might need to explain that when there are false positives or false negatives in the output, that is not a bug, it is an example of the "error rate," Haider said.
Communication within the enterprise is essential so that there is common understanding, according to Rowley.
"As we're working with folks across the company, we're finding that we not only have to re-train ourselves and our skill sets to deal with the platform and the new types of data, we have to bring our business partners forward as well on the journey and help them understand the terminology, the tool sets," Rowley said.