I applied online. I interviewed at Pinterest (San Francisco, CA) in May 2017
Interview
I applied online, had a 30-minute screen with a recruiter followed by a 45-minute technical screen with the person currently in the position (A/B testing Q&A and SQL challenge) and finally a 2-hour onsite with 4 employees.
The technical screen was pretty easy, just a few questions about how I would set up an A/B test and three questions where I was given a hypothetical table and asked to construct an SQL query to pull data to answer a question.
The onsite was where I completely lost interest in the role. The first interviewer (a lead in a related department) repeatedly told me they were "skeptical" of my answers without asking any follow-up questions or offering any explanation as to why they were skeptical. The interviewer later asked me to solve an equation without giving me the required inputs and frustratedly insisted they had when I asked for clarification. Later, as I was trying to ensure I had asked all of the questions that were important to me they said "you know you don't have to ask any more questions" before taking just one more of my questions and then ending the interview 10 minutes early. The interviewer was generally unfriendly from start to finish.
By contrast, the other employees I spoke with were quite friendly and very collaborative during the interviews. However, a couple of the interviews went deep into machine learning and stats concepts including assumptions of linear regression, impact required sample size from changing relative performance of test/control groups and writing the equation for a t-test. I was disappointed that the interviews focused heavily on details that can easily by Googled and not on whether or not I understood the underlying concepts or how to set up and solve similar problems.
Interview questions [1]
Question 1
What are the assumptions for linear regression?
Which results require a larger sample size to reach statistical significance?
a) test = 7% CVR, control = 5% CVR
b) test = 9% CVR, control = 7% CVR
What statistical test would you use to determine the sample size required to reach statistical significance? What is the formula?
What is the formula for a t-test (inputs alone are not sufficient)?
If Pampers were running an ad campaign on Pinterest, how would you determine which group of Pinterest users the campaign would perform well with?
How would you determine the incremental number of conversions a campaign on Pinterest generates?