Nike · Primly Community

Nike data scientist interview (SQL + case + stats): what actually came up in 2026

analyst_ana · 5 replies

just finished a Nike DS loop, consumer analytics team (they do a lot of membership and consumer lifetime value work). this is 2026 data.

Loop structure for DS: OA, technical phone screen, then 4-round virtual onsite. no ML take-home which i honestly appreciated.

OA: SQL-heavy. HackerRank, 3 SQL problems and 1 Python problem. SQL ranged from medium to medium-hard. window functions showed up. nothing exotic but you need to be fluent. the Python problem was a data cleaning/transformation task, no algorithms.

Phone screen: 30 min with a senior DS. one SQL problem (joins + aggregation), then discussion of a project from my resume. they wanted to know what metrics i cared about and why, not just what the model did.

Virtual onsite rounds: SQL round. 45 min. 2-3 problems, progressively harder. one involved a self-join to compare user behavior across sessions. another involved building a retention cohort table. if you're used to LeetCode SQL problems these are similar but more business-contextualized. Stats + product sense. they gave a scenario: a new feature in the SNKRS app showed a 5% lift in conversion in an A/B test. what questions do you ask before calling it a win? this is where they tested statistical intuition: sample size, segment effects, novelty effect, whether the metric actually matters. good discussion, not a gotcha. ML case. describe a model you'd build to predict which members are at risk of churning. open-ended. they wanted feature thinking (what signals predict churn for a consumer brand), model selection rationale, and how you'd evaluate it. i went with gradient boosted trees and they pushed back to ask if i'd considered simpler baselines first. correct answer: yes always. Behavioral / values. same format as the rest of the company.

What mattered most: business framing. they're not impressed by model sophistication for its own sake. every technical answer should connect to a consumer or business outcome.

comp i got: my offer was around $155k base for a mid-level DS role in Beaverton. RSUs on top of that but vesting schedule is 4 years.

5 replies

analyst_ana

the retention cohort SQL question shows up at a LOT of companies. is it true that Nike's DS team skews more toward analytics/BI or do they do real ML work?

hardware_hugo

from what i saw in the loop it's genuinely split. the consumer analytics side is more BI-adjacent but there's a real ML team doing recommendation and personalization for the app. depends a lot on which sub-team you're targeting. ask the recruiter before the loop which team it's for.

ml_mike

the pushback on gradient boosted trees to check if you considered baselines is a standard probe in good DS interviews. shows they actually care about model selection discipline rather than just CV scores.

de_derek

the SNKRS A/B test question is such a good one because that scenario is genuinely hard. product drops introduce so many confounds. novelty effect is real, the user pool during a drop is self-selected, traffic spikes change infrastructure behavior. you could talk about that for an hour.

alex_design

$155k base for mid-level DS in Beaverton, 2026. logged. RSU total compensation breakdown would be useful if you have it.