In a previous post, we explained that data science is about solving problems, which requires both domain knowledge and modeling concepts expertise. Specific modeling techniques and IT skills, although very important and useful, are just tools. This is extremely important, because it explains why focusing mostly on assessing knowledge of modeling techniques and IT skills while hiring data scientists will not help you select the best ones. Let’s see what matters when hiring a data scientist.

Interviews are not the right way to assess technical skills

There is a multitude of reasons why technical interviews just don’t work. First, candidates are evaluated by different interviewers who are bringing their own biases and ask different questions. This results in incomparable evaluations. Second, conflicts of interest cloud evaluations. How many times have you been interviewed by your potential next boss, who is younger and/or less experienced than yourself? The incentives for the interviewer to look for reasons to not hire you are high. And the list goes on.

The best way to assess technical skills is through an objective process, with rigorous testing and case studies. Based on the candidates’ profile, his skills on different subjects are assessed through online questionnaires. The problem solving skills are assessed though case studies, where a candidate is given a dataset and a particular goal. He has to define the question(s) he is going to answer in order to reach that goal and also determine if he wants to use additional data. During this step we can also assess communication skills by requiring the production of a report, a slideshow…

Focus on hiring a team

When building your data science capabilities, you should focus on building a team of problem solvers, with experience in different domains: a football team is not made only of quarterbacks or running backs, you need to fill all the positions. There are several reasons for this. First, you want to be able to solve different types of problems which usually requires different experiences and modeling approaches. The more industries and modeling techniques your team has been exposed to, the better. Second, you want to avoid biases in the choice of modeling methods. People tend to use techniques they used in the past and are comfortable with. That’s the infamous saying: “if all you have is a hammer, everything looks like a nail”.

An additional way to determine the best mix of people is to assess the cognitive and psychological profiles of the candidates. This type of objective assessments will help employers understand how candidates process information and identify potential social emotional concerns. They are a good complement to technical evaluations. In a field where ego problems abound, they help you weed out some candidates who lack team spirit for example. An interview with a psychologist to confirm the findings is a recommended option.

Interview for soft skills

An important step of the process is to interview for soft skills like:

  • curiosity
  • creativity
  • passion for the scientific detective work to identify the right data and modeling approaches
  • generalist vs. specialist
  • learning potential vs. existing skills

Here we should look, for example, at past experiences that demonstrate those skills: changing industry, applying many different modeling approaches, taking online classes regularly in different fields… As always, fact-checking your evaluation is very important.


Hiring the right people is crucial to the success of your data science initiatives. It requires you to stop hiring implementers with existing technical skills, and to start hiring for the ability to quickly learn new skills and for curiosity and creativity, using an evidence-based hiring process. Data Science provides a unique opportunity to revolutionize the way your company operates. Don’t miss out on it by hiring the wrong people!

Leave a Reply