Expedia: Data Science Interview Questions

Real Interview Questions You Could Be Asked

Image by Jason Goodman on Unsplash

I interviewed for a Data Scientist role at Expedia in Feb 2022. I made it to the final round, but did not get an offer. I’ll share the questions I was asked with you here.

You will probably be asked similar questions. It is company policy to ask standardised questions to be “fair.” You will have an edge over the competition. You’re welcome.

Take home test

There will be a small take home test. It is a rip off of this kaggle challenge. Use the add_datepart function from fastai and run it through a LightGBM model. You should be fine. This is what I did. They don’t care about EDA, fancy graphs and annotations. They have too many applicants. They just want to see that your model is above a certain benchmark.

Second Round

Technical SQL Interview.

You will be asked to live code SQL in a google doc. There are a series of questions that are fairly simple to answer (easy level on hackerrank). But there are a few tricks:

1. Remember to use having instead of where when you are doing a filter on something that you calculate in the query itself.
2. There will be a question on nested queries. They want you to use a CTE rather than creating a long nested query. Make sure you don’t fall for this trap.
3. Use a left join instead of an inner join. Pretty elementary. Just make sure you don’t make this mistake.

If the interviewer has to help you then you will be marked down. You may not make the next round.

Third Round

4 interviews with Data Scientists and Marketing Directors.

If you make it this far, be prepared for 4 x 1 hour intense interviews. The interviews with marketing directors are going to be very non-technical. They may ask you high level technical questions such as:

1. How would you go about trying to predict which SEM keywords were generating the most conversions? Tip: Pepper your answer with business acronyms such as CPC, CPM, ROAS, etc. I find this usually makes you look competent and knowledgable.
2. How would you estimate the mean of infinity? Would the answer be finite or infinite? (I was seriously asked this question)

The Data Scientists will ask Technical Questions. Be prepared. Here are a few:

1. If Expedia decided to change its logo, how would you AB test the change? What would make you recommend changing the logo? Tip: Make sure you mention some kind of threshold here.
2. Expedia gift cards: How would you assess whether they were working or not? If the gift cards were running at a loss, why would the company choose the keep them?
3. What are your steps for AB testing?
4. What features do you think would be most useful in terms of predicting whether a customer would cancel their booking? (list of datasets given, you have to pick relevant features from each dataset)
5. Are you more Frequentist or Bayesian in your approach to AB testing? Why?
6. How do you Explain DS Models to non-technical stakeholders?
7. How would you explain the results of an AB test to non technical stakeholders in a powerpoint presentation?
8. In what cases does it make sense to use the t-distribution rather than the normal distribution?

Tip: AB Testing is almost always the answer with questions that include “how would you test this?” Make sure you don’t miss out on it.

If you made it this far — thank you! Please let me know in the comments if you enjoyed this. Consider subscribing to my newsletter. Maybe I’ll continue to make more of these posts.