A data scientist at Instagram, Yuan Jiang, and a quantitative researcher at Two Sigma, William Chen, shared an insightful and useful three-part framework on Quora.com for how to answer open-ended data science interview questions.
Identify
Depending on how clearly the question is asked by your interviewer this could be as simple as just restating the problem or objective. Make sure you clarify the connection of the problem or objective to the company’s products or business issues. In doing so, Chen say, you make “it clear that you are a data scientist who is able to solve actual problems, and not just get stuck in brainstorming or exploratory analysis.”
Propose
Give a simple solution to the problem, if possible. Companies typically prefer a solution that is easy to understand, explain and implement. “It is much more important for something in industry to work and be implementable than to be theoretically perfect,” Chen posits. Don’t get caught up in what type of data or software is available, assume you will have what you need. Focusing on these issues takes emphasis away from the problem and the solution. Process isn’t as important as proving you have clarity on the problem and a clear solution. Providing answers to open-ended data science interview questions is critical for a successful interview.
Critique
“Critiquing and improving your simple approach lets you drive the interview forward as an interviewee,” says Chen. Chen is correct that you want any answer to any question to be part of a conversation versus a strict question and answer session. Chances are your interviewer will have expertise in the data science area and will appreciate that you can recognize and recommend multiple approaches without drowning in the morass of options. You want to do the critique after providing a specific solution in step two because it separates your other ideas from the choice you made. Making a choice of a solution gets you out of the brainstorming trap, where you give a long list of different ideas without actually solving the problem.
For additional help, go to the original post where Chen gives an example of how to follow the three-part framework for a specific open-ended data science interview questions on churn prediction.
