One key to hiring your next data scientist
For years we’ve been told that data science is the future; that artificial intelligence (AI) and machine learning (ML) will enable us to automate everything. And yet most (85%) data science projects fail, according to a 2019 report, though such scare statistics might not reflect reality. Still, there are plenty of reasons why a data science project might not work as advertised, but one reason stands out: Talent. Or, rather, the lack thereof, as Gartner has highlighted.
If you’re thinking, “Well, I’ll just send my recruiters to LinkedIn to scour for talent,” I have news for you: It’s not going to work. Running the numbers, Vicki Boykis has persuasively argued that there simply aren’t enough people to fill all the data science jobs on the market. And if you’re hoping for a qualified person (rather than digging into the “oversupply of junior data scientists hoping to enter the industry [with] mismatched expectations on what they can hope to find once they do get that coveted title of ‘data scientist'”), well, good luck with that.
And yet, there just might be a way to fill those data science needs.
Years ago, Gartner analyst Svetlana Sicular pointed out, “Organizations already have people who know their own data better than mystical data scientists….Learning Hadoop is easier than learning the company’s business.” Her suggestion? Train existing employees on the data science tools and techniques that will help them unlock that data.
More recently, Gartner analyst Nick Heudecker has reiterated this basic theme, but with a twist: “The data science talent market is hyper-competitive. Look for role adjacencies to find and develop talent.” The difference in Heudecker’s (and Gartner’s) suggestion is that this search for talent is both within your enterprise and beyond, but involves an expansion of the experience that could fit the data science role.
As a reminder, there simply aren’t enough qualified candidates to go around. Maybe you’ll win the data scientist lottery and hire all the talent you need, but probably not. After all, as Gartner has shown in its report, there are a few key things that a would-be hire needs to be successful with data science, and it’s pretty easy to disqualify 98% of the candidate pool simply by seeking four desirable (and, really, necessary) traits (Figure A).
Gartner suggests we look to adjacent roles to find potential data science candidates, but what are those roles? Or, rather, what are the requisite skills?
Years ago Mitchell Sanders called out a few key attributes of the ideal data scientist: Domain knowledge (i.e., they know the ins and outs of whichever industry in which they operate), math (multivariable calculus and linear algebra) and statistics expertise, and programming skills (especially those with R or Python experience). When trying to hire a data scientist (who may not have that current title), also look for people who understand how to model complex economic or growth systems, said William Chen, as well as those good with communication, product intuition, and other non-technical skills.
Would it be great to find someone with Sanders’ triple play? Sure. But given the competition for talent, and the dearth of data scientists with more than a year or two of experience, it’s perhaps wise to start looking for people who are deep in one or two key areas, and helping them expand their skillsets to round out their abilities.