PayPal · Primly Community

PayPal data scientist interview (SQL + case + stats), what the loop looks like

ds_dmitri · 5 replies

went through PayPal's DS interview in early 2026 for a mid-senior data scientist role on their risk/fraud analytics team. this loop is harder to find info on so writing it up.

total rounds: 4, over about 3.5 weeks from OA to offer.

OA (take-home, 2 hours): SQL heavy. about 5-6 SQL questions of varying difficulty. topics: window functions, self-joins, CASE statements, aggregations, handling NULLs. one question was data cleaning focused. platform was HackerRank. if you've done SQL interview prep, medium difficulty on average with one or two harder problems.

round 1: SQL deep dive (45 min). this surprised me. they gave me a schema representing a payment transaction table and asked me to write queries live. one was straightforward (monthly active users), one was tricky (identify users who made a payment, then had a failure, within the same session). know window functions cold: LAG, LEAD, RANK, PARTITION BY. they care a lot about whether you can write performant queries, not just correct ones.

round 2: analytics case study (60 min). no SQL. a business problem: "PayPal is seeing a drop in repeat payment rates among a specific merchant segment. how would you approach this?" pure analytical thinking, structuring the problem, defining metrics, thinking about potential causes, what data you'd look at. a bit like PM analytical rounds but with more emphasis on statistics and segmentation.

round 3: statistics + ML (45 min). A/B testing was front and center. how do you design an experiment to test a new fraud alert feature? how do you handle novelty effects? what's statistical power and why does sample size matter? also one question on model evaluation: given a fraud detection model with high precision but low recall, how do you think about the tradeoff? have real opinions, not just textbook answers.

round 4: behavioral (30 min). lighter than the SWE behavioral. two STAR questions, emphasis on working with stakeholders and communicating data findings to non-technical audiences.

comp: mid-senior DS, San Jose, offer was $155k base + some equity. a bit lower than I hoped but risk/fraud analytics roles can sometimes negotiate on equity.

5 replies

analyst_ana

window functions question in the live round is common across fintech DS interviews. LAG to look at previous transactions per user is exactly the pattern you need for fraud and churn analysis. good to know PayPal leans into this.

ml_mike

the precision/recall tradeoff question on fraud models is genuinely interesting. in fraud you almost always want higher recall (catch the fraud, even if you have false positives) but false positives have real cost: declined legitimate transactions, customer calls, merchant churn. did they want you to quantify the business cost or just explain the concept?

brand_ben

they wanted me to tie it to business impact. quantify: if we increase recall by X%, how many more fraudulent transactions do we block, and what's the value? versus: how many legitimate transactions get declined, and what does that cost in churn or support tickets? they appreciated when I asked 'what does a false positive actually cost the business' before answering.

de_derek

does the DS role at PayPal require Python coding or mostly SQL? I do more data engineering work but curious about the overlap.

ae_andre

no Python coding in the live interviews I went through. they mentioned Python/pandas in job description but all interview questions were SQL or conceptual. your mileage may vary by team though, risk/fraud is very SQL-heavy by nature.