Interview Questions for Data Science – Three Case Interview Examples

Part two in this series of useful posts for aspiring data scientists focuses on case interviews and how you can best go about answering them.

By Kaiser Fung, Founder, Principal Analytics Prep

In Part 1, I described two aspects of “critical thinking,” which according to many data science hiring managers, is the most desired skill they’re looking for. To test critical thinking, hiring managers use case interviews. Case interviews have long been popular with management consulting companies; they involve a free-ranging dialogue between the job-seeker and the hiring manager about a business problem – in the case of a data science interview, the business problem is expected to be solved using data-driven insights.

Data Science Job

Case interviews scare many graduates who have never seen a test that does not have an answer key. In industry, any problem worth solving does not come with an answer key. Critical thinking is the antithesis of formulaic thinking. The best way to master case interviews is to practice, practice, practice.

I set three sample case interviews for you to practice. Find some friends, and collaborate on your answers. With five people, your answers should cover most possibilities. Then find a hiring manager who’s willing to give you feedback. Remember: you are judged not only on the contents but the presentation.

  1. College Admissions Scandal
    • Imagine that you are a data scientist who’s been recruited to help detect fraud during the college admissions process. For this conversation, we shall narrow the focus to fraudulent information submitted in the college application forms, whether it is an inflated GPA, an invented sports achievement, or a fake community service achievement, or other types of forgeries.
    • You will be building a set of fraud detection models. Tell me what your first model will do, and why you choose that as your first model.
    • Imagine I am the director of admissions. Tell me why I should pay for your model.
    • What training data will you need to run that model? Where and how will you obtain the data?
  1. Lyft IPO
    • You are advising a friend who is considering either (a) driving for Lyft or (b) driving a yellow cab. Explain how you’d compare the profitability of these two options.
  1. Blue Apron post-IPO
    • In Q2 2018, Blue Apron, the meal-kit delivery business, reported that about 700,000 customers, 24 percent lower year over year while revenue per customer was $250, slightly down by $1 year over year.

Senior management is desperate to stem the customer churn. You are tasked with finding the reasons for the customer churn.

  1. Come up with 5 hypotheses for why the number of customers dropped drastically.
  2. Pick one of those hypotheses, and describe how you’d validate it.
  3. Assume you are able to validate the hypothesis, explain what you’d recommend to reverse the customer churn.

Let me address three questions that inevitably come up.

Where is the answer key?

If you are asking for an answer key, you’ve missed the point about case interviews. Case interviews are open-ended by design. If these questions have answer keys, then they’d be useless to assess critical thinking. The hiring manager is listening for how you approach problems, and structure the analysis.

How do I know I have a good answer?

Try your answer out on a few friends, or better, a hiring manager. Then, try again. If you’re doing it right, these different attempts should move along different paths, because the interview questions are designed to be open-ended.

There is not enough information to come to a conclusion.

That is exactly what every real-world data problem is like. You never have enough data, or all the right data, which means you need to make sensible assumptions, and keep moving along. A good case interview is a dialogue – you gather more information by asking your interviewer questions.

Bio: Kaiser Fung is the founder of Principal Analytics Prep, a leading data analytics bootcamp; best-selling author of Numbers Rule Your World; and the author of Junk Charts (, the popular data visualization blog.

Twitter: @junkcharts