What I Learned from My (Failed) Industry Interviews
My very first industry interview was in college, interviewing for summer intern at Deloitte Beijing. About 10 students are assigned to a big group, given a case, and we had to “fight” to shine. I did poorly. While everyone else was actively discussing and brainstorming, I was off-topic, and trying so hard to think of something to say but failed. Through the corner of my eyes, I saw the partner filtered out my resume, and discarded it like a worthless scratch paper. That was somehow devastating for me as a college student stepping awkwardly and timidly into the adult world, and I was afraid of interviews since then until last year. However, the fact that these interviews are crappy themselves doesn’t bother me. Being more like a techie nerd now, I won’t go interview at any company that assesses one’s ability and “leadership skills” in such a rough way.
After graduating from college I started my master’s and PhD’s study. During the last year of my PhD, I had interviews at different companies, and failed most of them (sadly). Every time I fail something I keep think back about it, and here are some of my thoughts.
Ask clarifying questions before jumping to any conclusions. For example, suppose you are given two groups of patients who use different drugs, and then you are asked to compare the effect of these drugs on survival times. Myself a year ago might just say: why not do a two sample t-test? Now I realize the importance of clarification in a hard way…Examples of useful questions to ask here:
- Are the two groups of patients balanced in terms of demographics and many other aspects such as income, education?
- If this is a historical dataset, which means there’s no way to control and balance patients during enrollment, was there any special criteria to meet before being give any medication?
- What measure of survival do we focus on here? Is it the Kaplan-Meier type survival curve, or simply the final “time” we observe? What if observations are censored?
Think out loud. Although it is very nice that you answer each question clearly and without much tough thinking, it is not generally expected. I think they would allow you time to think. It is not desirable to remain silent as they want to know your mindset. It’s ok to tell them what you are thinking, and where you get stuck, as they will sometimes give you hints. Utilizing hints and quickly turn to the right direction is also appreciated.
Being able to abstract a problem into a statistical framework is super important. I didn’t get enough training for this part as in school I deal with cleaned data everyday. This, however, will be the most frequently used technique in industry as an analyst or data scientist.
Be prepared for writing code on a whiteboard. Without an IDE, there is no auto-completion, no help manual, no nothing. This can be practiced at home with a board, or just using white papers.
Do not drink soda during breaks. It makes you burp.
Hopefully these are helpful tips. If you have any other suggestions, please add in comments, and I’ll update this post.