Apple · Primly Community

Apple data scientist interview (SQL, case study, stats): what I saw in the 2026 loop

ds_dmitri · 5 replies

Went through Apple's DS interview for a senior data scientist role on one of the app analytics teams. Wanted to write this up because the Apple DS loop is less documented than the SWE loop and the format surprised me.

Number of rounds: 5 total. Recruiter screen, HM intro, technical (SQL + stats), case study, and a cross-functional panel.

SQL round (60 min): This was legitimately hard. Not FAANG-medium-SQL hard. More like "here's a complex schema with multiple join paths and you need to figure out which one is right" hard. The schema was multi-table (4 tables). Problems included: A window function question with LAG() to find user session gaps A self-join to find users who did event A then event B within 7 days A question about correctly handling NULLs in an aggregate

You write in a shared editor. They're watching you think. Explain your joins before you type them.

Statistics round (part of the same 60-min block): After SQL they pivoted to stats. Questions I got: Explain the difference between a p-value and a confidence interval (they wanted intuitive explanation, not textbook) We ran an A/B test and the result was significant at p=0.03. What else do you want to know before calling it a win? (This is a trap. Say: sample size, test duration, any novelty effect, whether the primary metric moved or we're p-hacking secondary metrics) Walk me through how you'd handle a situation where your experiment has low statistical power

Case study round (60 min): I was given a scenario about a product metric declining. They wanted a full investigation framework: what metrics I'd look at first, how I'd segment the data, what hypotheses I'd form and how I'd test them, what I'd present to the team. This felt very close to the actual work.

What they value: intellectual honesty. I said "I don't know" once on a stats question and then reasoned through it aloud. They seemed fine with that. Faking it would have been worse.

Comp for senior DS in this org: I was told the band was around $195-230k base, with RSUs on top. Offer came in at $205k base. SF-based role.

5 replies

analyst_ana

The A/B test trap question is one I've seen in almost every DS interview at bigger companies now. The 'what else do you want to know' framing is so much better than just asking you to interpret a p-value. Do you remember if they pushed on the novelty effect angle specifically?

ds_dmitri

Yes, when I mentioned novelty effect they asked me to explain it and how I'd detect it. I said I'd look at the time-series of the treatment effect: if it's strong in week 1 and decays in week 2-3, that's a signal. They nodded and moved on, so I think that was the right answer.

de_derek

The self-join to find sequential events within a time window is one of those SQL questions that looks straightforward but trips people up on NULLs and edge cases. Did they care about query performance or was it more about correctness?

ds_dmitri

Mostly correctness, but when I finished they asked if I could make it more efficient. I suggested a CTE to prefilter before the join and they seemed satisfied. They didn't want a full optimization deep dive, just evidence that you think about it.

numbers_only

$205k base for senior DS at Apple tracks with what I've seen. RSUs are usually 4-year vest, 1-year cliff, and the initial grant can vary a lot by level and team. Did you negotiate or take the first number?