LinkedIn · Primly Community

LinkedIn data scientist interview (SQL + case + stats): what the loop actually looks like in 2026

ds_dmitri · 5 replies

Went through the LinkedIn DS loop a few weeks ago for a mid-level role on the monetization team. Five rounds total. Sharing the breakdown because I spent forever hunting for current info and found mostly 2022 posts.

Phone screen (recruiter): 20 minutes, background + timeline. They wanted to know if I'd worked on products that monetized at scale. Straightforward.

Technical phone screen: 45 min with a DS on the team. One medium SQL problem, one product metric question. The SQL was a self-join to find users active in two consecutive months. Not hard, but you need to write it fast and explain your logic while typing. The metric question was basically: "LinkedIn adds a new 'Share' button to job postings. How do you measure success?" Open-ended. They want to see if you think about guardrail metrics not just north-star.

Onsite (4 rounds): SQL deep dive: two problems back to back, one involving window functions (rank users by connection growth over 30-day rolling windows). The second was a self-referential query on a connections table. Know your window functions cold. Stats / probability: one A/B testing scenario (calculating sample size, discussing novelty effect), one probability question I hadn't seen before involving conditional probability on user behavior. Not leetcode-style, more like graduate stats. Product case: given a metric (feed engagement CTR dropped 10% week over week), walk through your investigation. This round felt most like the job. They really want to see a structured framework: data quality first, then segment, then hypothesize. Cross-functional / behavioral: stories about working with PMs and engineers. How do you communicate a result when stakeholders don't want to hear it.

Overall about 3 weeks from first recruiter contact to offer. Leveling conversations happened after. They're careful about L3 vs L4 and will push back on your self-assessment if they think you're reaching.

Happy to answer questions on any specific round.

5 replies

analyst_ana

The rolling 30-day window function question is really common there apparently. Did they ask you to write it completely from scratch or fill in a skeleton? Also which SQL dialect, BigQuery?

ds_dmitri

Scratch, no skeleton. They used a generic SQL editor, not BigQuery-specific. Standard window function syntax. OVER (PARTITION BY user_id ORDER BY date ROWS BETWEEN 29 PRECEDING AND CURRENT ROW) type of thing. Make sure you can write that without pausing.

pivot_pat

What level was the role and what was the offer ballpark? Trying to calibrate whether to pursue. I'm at L5 equivalent elsewhere.

ds_dmitri

L4 (their DS4). Offer came in around $195k base + RSUs vesting over 4 years, modest signing. Not top of market but competitive for San Jose location. Didn't get into extended negotiation.

returner_ren

Super helpful. How strict were they about the product metric frameworks? I have the stats background but I've been out for 2 years and am worried my product sense framing is rusty.