strategy for interviewing data scientists

Posted

In our current, dramatically low unemployment environment, positions with high demand for candidates, like data science, require an effective strategy for pursuing and for interviewing. Improve your process and your success in hiring by following this strategy for interviewing data scientists.

Carefully Select Your Interview Panel

Leverage the knowledge of your existing team as part of your strategy for interviewing data scientists. Organize each member to focus on a certain skill set, not exclusively, but predominantly. Use your best statistician to interview more closely on the candidate’s math and statistics skills. Your staffer with the best programming skills should address the candidate’s coding skill set. If your staff is large enough, have a more junior member “shadow” the primary interviewer. You will be preparing all of your staff to do a better job recruiting, and get one more set of insights and feedback. Is this your first data science hire? Use your consultant or an academic contact to assist you.

Develop Recommended Interview Questions

You want to measure the skills most important for your opening. In a strategy for interviewing data scientists these typically address four primary areas: statistics, databases, programming and interpreting data. Questions that test on the basics of these primary areas, if well prepared, can be as revealing as more advanced questions. For instance, if they understand and can discuss projects they’ve done in linear and logistic regression, chances are they will be able to pick up Random Forests or boosting on the job.

The best interview questions are multilayer. They allow the answers to go in a number of different directions. Sebastian de Larquier, Data Scientist at Netflix, suggest you “follow the candidate, go where they are going.” Don’t waste time with brain teasers. Don’t look for a single right answer, discussions about alternative results or strategies are often far more interesting and informative about the candidate’s thought processes.

Plan for Each Interview to Last One Hour

Interviews should feel like conversations, but use your time and the candidate’s time wisely. This may mean parceling questions across multiple interviewers. Test technical questions on your existing team to see how long the answers might take. If the candidate gets stuck on a technical question, provide an alternate way of thinking about the problem. Remember these candidates are in high demand and making them feel failure in the interview is not an effective way of recruiting them. Often their reaction when they can’t answer a question more clearly highlights their ability to shift gears and search for an answer versus becoming defensive and give up. Always leave time for candidate’s questions as they are often the most insightful about what they are looking for and what they understand about your company and your opening.

Be Careful with Data Challenges

No candidate wants to spend the weekend working on a project for your firm. Target four hours for the amount of time a candidate should have to commit to your data challenge. Include messy data in the challenge. It is very revealing to see how they clean the data, or note interesting findings. Consider the possibility of allowing a more senior candidate to present a project or sample code they have previously done in place of doing your data challenge. Remember, you aren’t the only firm pursuing this candidate and your strategy for interviewing data scientists shouldn’t be so onerous as to lose them in the process.

Standardize Your Evaluations

Create a scorecard with room for open-ended comments. Requiring your staff to commit to a score, and to writing something down often clarifies their feelings about a candidate. Along with the primary skill categories of statistics, databases, programming and interpreting data include less tangible skills like energy level, how logical, how articulate, level of interest in our firm and our job and fit in the team/company. Ivu Vukicevic, Data Scientist at Macy’s, advises if you find a candidate with strong logic and communication skills even with weaker technical skills, “consider investing in them. A smart worker who meshes well with your existing team can quickly be brought up to speed.” Make these scorecards due within 24 hours of the interview. Prompt feedback to the candidate shows level of interest and level of organization in a company.

Interested in interviewing data scientists and in hiring them? Contact Smith Hanley Associates’ Data Science and Analytics Recruiters.


Leave a Reply

Your email address will not be published. Required fields are marked *