INTERVIEW GUIDE

Google Data Scientist Interview: Questions & Process

Google's Data Scientist interview leans heavily on statistics, experiment design (A/B testing), and product sense — not just coding. You'll write SQL, reason about metrics for a Google product, and defend statistical choices out loud. Expect a recruiter screen, a technical phone screen, then a 4-5 round virtual onsite over several weeks.

The interview process

1. Recruiter screen — ~30 min call

Tests: background, the specific DS track (Product Analytics vs. research-leaning), and logistics

2. Technical phone screen — 45-60 min

Tests: SQL on a shared doc, plus probability/statistics fundamentals and a quick product-metric reasoning question

3. Statistics & experimentation — 45-60 min onsite round

Tests: hypothesis testing, p-values vs. confidence intervals, A/B test design, power, and common pitfalls (peeking, novelty effects)

4. Analytics / SQL coding — 45-60 min onsite round

Tests: writing correct SQL against a realistic schema — joins, window functions, cohort and funnel logic — and some light algorithmic coding in Python

5. Product sense & behavioral (Googleyness) — 45-60 min

Tests: choosing and defining metrics for a Google product, diagnosing a metric drop, and collaboration / ambiguity (Googleyness)

Questions you're likely to get

Technical

Daily active users dropped 5% week-over-week. Walk me through how you'd investigate whether it's real.
How would you design an A/B test for a new YouTube recommendation feature? What's your metric and how long do you run it?
Explain the difference between a p-value and a confidence interval to a non-technical PM.
Write SQL to find each user's first purchase date and the revenue from that first order.
Given an events table, write SQL for 7-day retention by signup cohort.
What is statistical power, and how do you choose a sample size for an experiment?
A coin is flipped until the first heads. What's the expected number of flips? Now derive it.

Role-specific

Pick a single north-star metric for Google Maps and justify it. What are its blind spots?
How would you detect whether an experiment result is a novelty effect versus a durable change?
Your A/B test is significant on a guardrail metric but not the primary metric. What do you tell the PM?

Behavioral

Tell me about a time your analysis changed a decision. How did you make it land with stakeholders?
Describe a time you had to explain a nuanced statistical result to a non-technical audience.
Tell me about a project where the data was messy or incomplete. How did you handle the uncertainty?
Describe a time you disagreed with a stakeholder about what the data showed.

Practice these problems live

Relevant LeetCode problems for the Google Data Scientist loop. Start a live, AI-run coding interview on any of them — or open the problem on LeetCode.

▶ Combine Two Tables· Easy · LeetCode #175

Combine Two Tables — foundational JOIN logic every DS SQL round assumes.

▶ Article Views I· Easy · LeetCode #1148

Article Views I — DISTINCT + filtering, a quick product-analytics warm-up.

▶ Game Play Analysis IV· Medium · LeetCode #550

Game Play Analysis IV — retention/cohort logic with window functions, core to product DS.

▶ Department Top Three Salaries· Hard · LeetCode #185

Department Top Three Salaries — advanced window/ranking SQL Google may push toward.

▶ Two Sum· Easy · LeetCode #1

Two Sum — light Python coding to confirm basic algorithmic fluency.

▶ Maximum Subarray· Medium · LeetCode #53

Maximum Subarray — light DSA round; clean O(n) reasoning under time pressure.

How to answer (worked examples)

Daily active users dropped 5% week-over-week — how do you investigate?

Structure before you speculate. First ask: is it real (instrumentation bug, logging change, holiday/seasonality) or a true drop? Then segment — by platform, geography, new vs. returning, app version — to localize it. Form a hypothesis, check it against the data, and quantify the contribution of each segment. Close with what you'd recommend and how you'd confirm. Google rewards the candidate who isolates the cause methodically over one who jumps to a guess.

Design an A/B test for a new recommendation feature.

Name the hypothesis and the primary metric (e.g., watch time per user) plus guardrails (latency, opt-outs). Define the unit of randomization, sample size from a power calculation, and run length to cover weekly seasonality. Then volunteer the pitfalls yourself — peeking, novelty effects, network interference, multiple comparisons — because surfacing them is the signal. End with the decision rule before you ever look at results.

Tell me about a time your analysis changed a decision.

STAR. Situation: the decision on the table and the assumption being made. Task: the question you owned. Action: the analysis and, crucially, how you communicated it (a clear visual, a one-line takeaway). Result: the decision that changed and the measurable impact. Googleyness shows up in how you made a rigorous result usable, not just correct.

What Google looks for

Statistical rigor: you reason about power, significance, and bias rather than reciting definitions
Clean, correct SQL on realistic schemas, including window functions and cohort logic
Product sense: you pick metrics that map to user value and name their blind spots
Structured problem-solving on ambiguous, open-ended questions
Clear communication of nuanced results to non-technical stakeholders (Googleyness)
Red flag: jumping to a model or a conclusion before clarifying the question or checking the data

FAQ

Is the Google DS interview more stats or more coding?

More stats and product reasoning than a SWE loop, but SQL is non-negotiable and you'll do light Python. The differentiator is experiment design and metrics thinking, not heavy algorithms.

How much SQL do I need?

Solid intermediate-to-advanced: joins, aggregation, window functions, and cohort/funnel queries written quickly and correctly. Practice writing it by hand, not just running it in an IDE.

Are there different Data Scientist tracks?

Yes. Google has historically split DS roles between product/analytics-focused and more research/ML-focused. Ask your recruiter which track you're in — the round mix and emphasis differ.

What is 'Googleyness'?

Google's behavioral bar: comfort with ambiguity, collaboration, intellectual humility, and bias to act on imperfect information. It's assessed alongside your technical rounds, so prepare real stories.

How long does the process take?

Often 4-8 weeks, partly because of Google's hiring-committee review after your onsite. Pace can vary widely by team and time of year.

Rehearse the stats round out loud

Knowing what a p-value is and explaining it cleanly under pressure are different skills. Practice Google's experiment-design and metric-diagnosis questions out loud with OfferLoop's realtime voice coach so your reasoning is structured before you're on the spot.

Practice this interview out loud →

All Google interviews →Google Software Engineer interview →Google Data Analyst interview →Google Data Engineer interview →Google Machine Learning Engineer interview →

OfferLoop is an independent interview-practice tool and is not affiliated with, endorsed by, or sponsored by Google. All company names and trademarks are the property of their respective owners.

Interview formats vary by team, level and year, and this guide is compiled from general knowledge of publicly discussed hiring processes — treat it as preparation material, not an official description of Google's current process.