Some of the data science skills necessary to succeed in the profession are obvious. Things like coding, statistics...
and business knowledge are now just table stakes. But what makes a really good data scientist is often much subtler, and businesses interested in hiring data scientists have to look deeper to make sure they get the most bang for their buck.
For example, Alfred Essa, vice president of research and data science at publishing company McGraw-Hill Education, said he looks for the ability to communicate results and work as part of a team in all new data scientists he hires. In a panel discussion at the recent Business Analytics Innovation Summit in Chicago, Essa said the projects his team leads are very collaborative and require a lot of interdependent work. So he asks job candidates to give a presentation on some work they've done in the past to assess their communication skills.
Essa recently hired a new data scientist who had just completed a Ph.D. in physics. Before being hired, the candidate presented his dissertation on the esoteric topic of quantum scattering in the Martian atmosphere. It was a topic that nobody on Essa's team knew anything about going in, but the presentation was effective at conveying the basics. That's how Essa knew the candidate would be a good fit at New York-based McGraw-Hill.
"All our work is done in teams, so in the interview process, it's important to ask that [they] do a presentation," he said. "These are the soft skills that distinguish a good data scientist from a great data scientist."
For Scott Sokoloff, chief data officer at Chicago-based Groupon Inc., a certain level of self-awareness on the part of data scientists is crucial. He said that, despite its reputation as a purely objective practice, data science involves a lot of purely subjective decisions.
Things like deciding which data elements to include as features in a machine learning model, to bigger picture decisions, like what business problems or questions to address, require judgment and data science skills, Sokoloff said. It's also important for data scientists to constantly evaluate why they made the decisions they did and whether different decisions could have led to better outcomes. Challenging your own assumptions and looking for different approaches can actually make analytics more objective, Sokoloff said.
"Far too often, people doing data science don't realize they're making assumptions, and to me, that's the death knell," Sokoloff said.
Ismail Parsa, head of data science at the Turkish online retailer Hepsiburada, said a true scientific mindset is one of the critical data science skills. Too often, data scientists build something and move on, but Parsa said starting with a hypothesis and testing it is important.
Data scientists should continue testing their products even after they've been deployed. This kind of continuous testing ensures optimal performance.
"If you're hiring a data scientist and they don't use the word testing [in their resume or during the interview], you can seriously doubt their work," Parsa said.
Data science jobs are rarer than you might think
Data science isn't just the same old BI
Data science still tops list of best jobs