How not to hire a data scientist
4 pitfalls to avoid
In June Glassdoor crowned data scientist as the best job in America for the 4th year in a row.
But the profession that the Harvard Business Review famously described as “the sexiest job of the 21st century” is also one of the most ambiguous. As companies of all sizes figure out how to filter the ever-increasing firehose of data they’re being hit with, they’re also dealing with the fact that there simply aren’t enough data scientists around who are skilled enough to glean important business insights from it all.
For smaller businesses, hiring a data scientist is a big investment that must be made carefully—a candidate with too few skills, or the wrong ones for your company, can throw a major wrench into your data-driven projects. Here are a few mistakes to avoid.
Don’t count on finding a unicorn
There are many subdisciplines of data science, but the easiest way to explain the differences to the less technically-inclined people on your staff is to divide them into two categories: data science for people and data science for machines. Decision scientists (the people ones) draw conclusions from data that help other team members make critical business decisions—everything from sales to marketing to product design. Modeling scientists (the machine ones) develop models, algorithms and training data that gets fed directly back into computers, either by the data scientists themselves or by software engineers.
Smaller companies that don’t have the budget for a team of data scientists may try to find a “jack of all trades” candidate who can solve both these sets of problems. These candidates are exceedingly rare—if you can find one, great! But don’t bank on it. Instead, think about which category is the best fit for your needs and narrow your search accordingly.
Don’t bank on AI expertise
Let’s talk more about modeling scientists for a minute. The rise in AI and machine learning is a major reason companies are seeking them out, even though they only play a supporting role in that kind of analysis. If your business is serious about developing a machine learning-based product or an overarching AI strategy, then yes, you’ll need your first data scientist to have a machine learning background. But you’ll also need data engineers who know where data lives and what it contains, as well as dev ops pros who can operationalize a machine learning model at scale. That requires a big budget—possibly bigger than what your c-suite is expecting—and they should understand those parameters from the get-go.
Don’t rely on nanodegrees
Considering how scarce experienced data scientists are, it may seem tempting to invest in data science nanodegrees for your current staff. Nanodegrees have a mixed reputation. They are cheaper and less involved than a master’s degree, only focusing on a specific set of skills tailored to the needs of different tech industry jobs. They also vary in quality and may not be as useful as they seem. Data scientists are an educated bunch: 88 percent have at least a master’s degree, and 46 percent have PhDs. It’s hard to substitute those years of study with a few online classes, even if they’re good ones.
Don’t hire candidates fresh out of school.
Some universities are offering their undergraduates a digital technology credential (sort of like the nanodegrees we just mentioned) that is neither a major or minor, but a sequence of courses the universities develop with business leaders that focus on specific data science skills. But four or five courses—even highly specialized ones—isn’t enough for a recent grad to hit the ground running as your team’s lead data scientist. Larger companies have the budgets to hire both senior data scientists and junior data analysts; a small company may not. Choosing an experienced candidate who won’t make novice mistakes will pay off in the long run.