More and more companies are looking to fill open Data Scientist roles for their technical team. In fact, 8.4% of respondents who took our Annual Developer Survey identified as Data Scientists, up 6.8% from last year's results. Looking at Google Trends, we also see that interest in the term "Data Scientist" has steadily increased over the past 5 years.
My fellow Data Scientist David Robinson and I have noticed several priorities in our own job searches, as well as our peers. While the role of a Data Scientist (and how to hire them) is all still in flux, we've found the following points important and broadly applicable.
Most Data Scientists (vs. Data Engineers) thrive best and can have the most impact in organizations with already mature data engineering. Data needs to already be in decent shape and accessible for a Data Scientist to be able to do analysis, visualization, and modeling.
During a job search, this is an important way to evaluate companies. Even the kind of data a company sends a candidate for a project tells the candidate something about this.
Data Scientists want to know that they will have a stake in the company's decision-making process. For companies hiring data scientists, are they actually going to listen to them? Does leadership have a history of paying attention to and trusting results of modeling? What happens with a null result, or a result that isn’t what leadership was hoping for? This is about company culture when it comes to leadership and decision making, and is something we have seen play into peers’ decisions about data science jobs.
Data Scientists typically come from highly academic backgrounds. This means that we are used to some academic norms, such as having a lot of autonomy about the kinds of questions we work on and working with open data. Data Scientists understand that moving to an industry means some adjustment here, but most of us still like a lot about working in academia and find workplaces more appealing when they make room for some of those aspects when possible.
Data Scientists shouldn’t be interviewed using the same process and questions as an Engineer, since they have an entirely different set of skills and job requirements. Common programming interview questions about sorting lists, recursion, or binary arithmetic won’t help evaluate a Data Scientist’s skills and may send a signal that the company doesn’t understand their role.
At Stack Overflow, we recently implemented an interview process for Data Scientists where the candidate goes through three steps:
The steps above aren't the only approach to interview, but they show how the emphasis should be on practical analysis skills rather than abstract programming problems.
If this is the first time they're hiring data scientists, they should ask this first. It’s a common question since some companies hire a Data Scientist before they need one, simply because “everyone’s doing it.” Examples of reasons a company may need data scientists include to implement or improve machine-learning driven features, or to answer specific and business-relevant questions about their datasets.