Shopify · Primly Community

Shopify data scientist interview (SQL + case + stats), exactly what I was asked

analyst_ana · 5 replies

Went through the Shopify DS interview loop for a mid-senior data scientist role focused on merchant analytics. Here's the breakdown by round.

SQL round (60 min, live coding): Three SQL problems. First one was straightforward aggregations with GROUP BY and window functions. Second was more involved: month-over-month retention for a cohort of merchants. Third was open-ended: given a schema they provided during the interview, write a query to surface merchants who might be at risk of churning. That third one had no single right answer, they wanted to see how I defined risk signals from the available data.

They're not testing syntax. They're testing whether you can translate a business question into a query and whether you explain your reasoning out loud.

Case / product sense round (60 min): Given a scenario: merchant GMV on a segment of Shopify is down 8% QoQ. Walk through how you'd diagnose it. This is standard DS case territory but the Shopify framing means you need to think about merchant-level vs. transaction-level vs. platform-level causes. I walked through a structured breakdown: check if it's one segment, one geography, one merchant size tier, or platform-wide. Then hypothesize causes. Then describe what data I'd pull.

They pushed on 'what if you can't get that data in time, what's your fallback' which is a good signal that they care about being pragmatic.

Stats / experimentation round (45 min): Conceptual, not a coding round. Questions on A/B test design, sample size calculation, how to handle novelty effects, and one question specifically about switchback experiments (relevant because some of their experiments can't be randomized at the user level). If you haven't read about switchback or cluster-level experiment design, do that.

Overall: harder than most DS loops I've done, specifically because the SQL is live and the case requires real domain intuition about merchants. The stats round is above average but not PhD-level. Good prep if you're also targeting other commerce or marketplace companies.

5 replies

de_derek

The switchback experiment question is a great signal that they're asking real questions from their actual work. That's specific to platforms where you can't isolate user groups cleanly. Appreciate you calling it out.

analyst_ana

Was the SQL done in a specific editor or plain HackerRank? Also did they care about query optimization / indexes or just getting the right answer?

hardware_hugo

A shared Google doc for the live SQL, not HackerRank. They didn't care about indexes explicitly, but I mentioned one indexing consideration on the retention query and it generated a short interesting side conversation. Wouldn't go deep on it unless they open the door.

contractor_kai

How technical was the hiring manager in the loop? I've had DS loops where the HM is a business stakeholder with no stats background and the calibration is all over the place.

consultant_cam

My HM had a strong data background, had been a DS themselves before moving into management. The calibration felt very consistent across the loop. I don't think that's universal across all Shopify DS teams though.