Reading Time: 3 minutes

by: Zank, CEO of Bennett Data Science

Why Most Data Science Interview Questions are Rubbish

Here’s why I think a lot of the tough technical questions asked of young applicants are unnecessary, and why project-based conversations are my preferred method of assessing whether or not an applicant will thrive.

The job interview can strike fear into the bodies of professionals at any level. I remember interviewing once at Qualcomm, and, after clearly stating that I was not a C/C++ programmer, being drilled on a six line snippet designed to confuse even the compiler itself! I was exasperated. Clearly I haven’t forgotten it! I get some solace in knowing that The Q is quite well known for these sorts of interviews. You’d have thought they’d be easier on me. After all, I spent the better part of six years walking in the engineering building at UCSD named after the founder and his wife 😛

I think we can all agree that job interviews can be harrowing, never mind the possibility of being asked questions like, “how much dirt is in a 4x4x4 hole?”! (There’s no dirt in a hole.) Thankfully, we don’t hear about many interview questions like those anymore. And for good reason, they were shown to predict nothing.

So what are companies asking? What’s fair to ask? What’s best to ask? How do I go about this process?

There are so many articles out there, for hiring managers and job seekers, offering up the Ultimate [your job title here] Interview Question List.

The problem with these is multifold:

  • They paralyze both sides by giving too many options to study up on
  • There’s no guarantee they’ll be relevant to the industry the candidate is applying for
  • A candidate who aces these technical questions can still be a poor fit, (while an intelligent, experienced candidate who fits well, hardly ever is)

Let’s look at what Google has to say. Here’s the top 10 from the Google search: top data science interview questions

  • 100 Data Science Interview Questions and Answers (General) for 2018
  • 109 Data Science Interview Questions and Answers for 2019 …
  • 21 Must-Know Data Science Interview Questions and Answers
  • Data Science Interview Questions and Answers for 2019 – Intellipaat
  • Top 45 Data Science Interview Questions and Answers For 2019 …
  • Comprehensive Data Science, Machine Learning Interview Guide
  • Top Data Science Interview Questions & Answers — Part 2 – Medium
  • Top 100 Data science interview questions (article) – DataCamp
  • 20 Most Popular Data Science Interview Questions – Simplilearn
  • 7 Data Scientist Interview Questions and Answers |

I was intrigued by the second one, so I took a peek inside.

Here are the headers:

  • A Curated List of Data Science Interview Questions and Answers
  • 1. Statistics Interview Questions
  • 2. Programming
  • 3. Modeling
  • 4. Past Behavior
  • 5. Culture Fit
  • 6. Problem-Solving
  • Conclusion

They broke the questions down into six categories. And did a damn good job giving really good technical questions and answers.

Here are a few:

  • How do you deal with sparsity?
  • What is sampling? How many sampling methods do you know?
  • How would you clean a data set in [Python]?

Those some great questions. But just imagine yourself hearing a perfect answer. Do you hire based on that? Is it just a piece of a more holistic picture? Who’s assembling the pieces, and is that even possible?

My argument is this: there’s a huge difference between a well-crafted answer to a technical question and a story about how a candidate actually DID something. I tend to favor the higher-level project details over the lower level operational details any time.

To that end, I always recommend taking a step back, and up, during interviews. I like to ask what was exciting about the last project they worked on. Or, what side projects do you work on in your spare time? I try to get them to tell me about the moments when they stood up from the keyboard and clapped aloud or raised their hands in the air. We’ve all done it! I want to know about those moments and what drove them at that time.

Knowing how to define lift in an FPGrowth algorithm is just useless, unless that’s all the person will be doing…and even then, something that simple can be taught, instantly to the right person.

Ask what makes people go, DAMN! Find out what really excites your candidates. And if you can provide that at your company, I believe you’ll have a match you’ll enjoy working with for a long time.

Zank Bennett is CEO of Bennett Data Science, a group that works with companies from early-stage startups to the Fortune 500. BDS specializes in working with large volumes of data to solve complex business problems, finding novel ways for companies to grow their products and revenue using data, and maximizing the effectiveness of existing data science personnel.

Signup for our Newsletter