Bridging the Talent Gap: How to Find the Right Data Scientist

Claudia Perlich, chief scientist at Media6Degrees, is teaching a graduate MBA course at New York University called “Data Mining and Business Intelligence.”

Google Research senior statistician Rachel Schutt is teaching “Introduction to Data Science” at Columbia University.

It’s this trend in educating a new cadre of data scientists — hiring faculty with both practical and theoretical skills — that will help solve some of the talent shortages organizations are facing in the era of big data. But what should organizations do now, before those newly minted data scientists graduate?

In a recent day-long conference, Big Data: The Management Revolution, hosted by the MIT Center for Digital Business, the practice of finding and hiring data scientists was pondered by three panelists: Perlich, Schutt and Thomas H. Davenport, visiting professor at Harvard Business School and author of the upcoming book Keeping Up With the Quants (Harvard Business Review Press, June 2013). What can companies do in the next thirty days, or the next six months, to help fill their data scientist gap? MIT’s Erik Brynjolfsson, Director at the Center, posed this question to the panel.

Here is their advice:

• Claudia Perlich: “Take inventory of the people you have already. Make sure they are not marginalized. Make them feel more important. You probably have someone with that skill set. See how you can push them a little bit more into the business side, or someone in business, to the data side.”
• Rachel Schutt: “Really value data and examine your data strategy. How are you currently using data? Longer term, believe in the notion of teams that can collaborate. Put people with business savvy skills along with those that have data skills. Put data people in executive level positions so they are interacting with decision makers.”
• Tom Davenport: “What key decisions do you need to make; what product areas do you need to build in? Is it big data or is it small? That will drive the big data and analytics strategy. EMC is creating an internal training program. That’s a smart thing to do. Identify the people in Rachel’s class at Columbia, in Claudia’s class at NYU. Identify relationships with schools and get those people, quickly.”

For those organizations at the interview stage, Davenport has an additional suggestion: Ask candidates what business problems they have solved, or that they are interested in. “That eliminates about half the interviewees. You either have that or you don’t. Filter that out in the interview process.”

For Perlich and her New York startup m6d, the demand for data scientists is no small matter. Her company serves up targeted ads to consumers based on a data-driven technology platform that has to make 15-millisecond decisions about which adverts to display, when. The company runs two thousands models a week. Over the past few years it has grown its team of data scientists from two to five.

Perlich is looking to hire again in 2013. “I need an all-a-rounder; I can’t have a statistician,” she said. “I am looking for someone who is deeply skeptical, someone who is very good at tricking data, eking out results and [questioning] themselves. That’s a personality trait. I’m also looking for someone who can use new technology and use it quickly, who can really manipulate data.”

In addition, Perlich is seeking candidates with domain expertise. The reason: So they know how to ask the right questions. “They need to know how the data was generated and that’s a confluence of technology and domain expertise,” she said. “I can write you a predictive model about anything. So what? It’s the question that is important.”

Google’s Schutt offered additional insights into what to look for. “I still believe having an actual foundation of math — that’s still important,” she said. “Educating people in coding and programming is important.” They should also be educated about the ethics of data use and the ethical implications of data models. “They can’t just predict and cause things to happen. Data scientists need to be cautious people who care about their impact on the world.”

One caveat to the pursuit of talent: In these heady days of big data and data science, Perlich is worried about people relabeling themselves data scientists, but not having the skills to back up their claims — and doing poor work as a result. This, she said, could wreak havoc on the field. “You can fool yourself with data like you can’t with anything else. I fear a Big Data bubble.”