Benjamin Haas - Fotolia
Interest in enterprise data science initiatives continues to grow, but as more and more businesses start thinking about bringing new employees aboard to develop and run advanced analytics applications, the question of how to hire data scientists is coming to the forefront -- and it's creating some confusion and differing opinions on what skills to prioritize.
"There's a lot of buzz around data science, and I think there's a pretty big gap between what people think it is and what companies actually need," said Frank Lo, director of data science at Boston-based online home furnishings retailer Wayfair LLC. "There's a lot of disagreement about what makes a good data scientist."
Lo and other speakers at the 2015 Big Data Innovation Summit in Boston said they focus more on soft skills than specific technical capabilities in assessing candidates for data scientist jobs. For example, Lo, who currently manages a team of 18 data scientists, said he had to do some serious thinking about what he and the business at Wayfair need from those workers. Eventually, he settled on intellectual curiosity as the top trait to look for in new recruits.
It can be challenging to test people for that in a traditional interview process, but Lo said he has found ways to get at the issue. For example, he asks applicants to describe an analytics project they worked on outside of school or previous jobs. Lo is looking for data scientists who are passionate about analyzing data to solve problems, and he thinks that passion should transcend academic and professional responsibilities.
Frank LoDirector of data science, Wayfair LLC
"This is something you can't find on a resume or in a degree," he said. "If you think about data science, it's about what inferences you can make. I look for people who are not interested in answering questions, but in asking their own questions."
Lo said he has even turned down job candidates who had impressive lists of technical skills because they didn't appear to have a curious nature. This isn't to say that tech skills aren't important to him in hiring data scientists, though. Lo said people aren't a good fit for jobs at Wayfair if they don't have at least basic programming skills in Python or Java. He's less insistent that they have specific skills in using Hadoop because SQL interfaces to the distributed processing framework are so common. Today, very few people actually need to know how to program in MapReduce to analyze Hadoop data, according to Lo.
New hires need to know the business
Riley Newman, head of data science at lodging rentals website operator Airbnb Inc., said soft analytics skills are also one of his top priorities when he goes to hire data scientists. Specifically, he looks for strong communication skills and business acumen. "We wanted to find people who are very effective communicators as well as being good at Python and R," Newman said in a presentation at the big data conference.
Currently, Newman manages an analytics team consisting of 70 data scientists. They're split up into smaller teams that are embedded within separate business units. Periodically, the teams rotate to other lines of business in order to give them new challenges. Newman said that approach helps keep the data scientists interested in the work they're doing, but it also demands a high level of flexibility when engaging new business problems.
That's why San Francisco-based Airbnb has a fairly extensive interview process for prospective data scientist hires. It starts with a one-on-one interview; from there, candidates are given take-home tests to assess their programming and data analysis skills. Next, recruits are brought back into the office and given an in-house test in which they are teamed up with existing employees and asked to solve a business problem using available data. At the end of that exercise, they have to give a presentation explaining their findings.
It's an involved process, but Newman said it assess all the core functions that a new hire would have to handle. "It doesn't just come from hiring smart people," he said. "We really have to start by thinking about the problems we want to solve and how people map to those areas."
Focus on tech skills might miss mark
As in Lo's case at Wayfair, the question of what technical skills data scientists need always comes up. Inevitably, they need to know how to obtain, manipulate and analyze data, often in very large volumes. But businesses might miss out on good employees when they hire data scientists if they insist that job candidates have deep experience in all of their existing systems and analytics tools.
"You hire the right smart people, [and] they can learn most of your stack in three months," said conference speaker Scott Sokoloff, chief data officer at OrderUp Inc., an online food delivery service based in Baltimore.
Sokoloff said that job descriptions for data scientists often read like technical spec sheets. Many companies want to know that their new hires can hit the ground running in using the tools they already have in place. But for Sokoloff, focusing too much on specific tools creates a risk of disqualifying potentially strong candidates.
"It's not about the technology," he said. "It's not about using this software or that solution. It's about having a business problem you want to solve."
Is it better to hire a data scientist or train a data analyst?
Ask these questions when looking to hire data scientists
New analytics need in some organizations: Data storytelling skills
Expert tips on building a data science team for business