Jobra
🐞Step 1 of Jobra:AI Debugging Rounds
DSA+AIPreparing you for both.
JOBRA · AI DEBUG ROUND (PREVIEW)sum.cpp
// buggy snippet
int total(vector<int>& a, int n) {
    int sum = 0;
    for (int i = 0; i <= n; i++) { // 🐞        sum += a[i];
    }
    return sum;
}
A real codebase. A real bug. You + AI.

Jobra drops you into a broken repo. Find it, fix it, ship it. AI helps you debug like on the job.

Prompt:Correct the loop condition
Tap 'Let Jobra fix it' below to simulate AI debugging...

Google Data Scientist Interview Guide: Stats, Coding, Product Cases and Googleyness

5 min read

Prepare for Google data scientist interviews with reported assessment patterns, interview questions, and preparation advice.

In this article
On this page

Last reviewed November 10, 2026.

For data scientist and DS research (DSR) candidates.

Built by the Codinzhub team from Google interview experiences that candidates shared, including the ones posted on Codinzhub. Each experience is one person's interview, and Google changes its process by year, level and team, so use this as a map, not a promise.

The 30-second version. A work-style hiring assessment, a recruiter call about your SQL and Python, then a phone screen of one or two sessions (coding, SQL, statistics or a product case), then a virtual onsite of two or three sessions, which reports named "Measurement and Modeling Concepts", "Coding, Applied Analysis and Experiments" and "Googleyness and Leadership". Then team matching. Google DS has separate tracks, including DS research (DSR), so ask which track and team you are interviewing for.

Round-by-round breakdown

Stage 1: Application, where candidates applied

How it works. In several experiences a recruiter reached out first. In one DSR account, the candidate applied online in August, and a recruiter called in October because two teams had picked the resume.

Stage 2: Shortlisting, how candidates moved forward

How it works. The hiring assessment comes first: Likert statements and "most likely / least likely" workplace scenarios, with no coding and no interviewer. One candidate's advice was to be consistent and avoid the middle option. Then a recruiter call about your projects and your SQL and Python experience. In one report the recruiter went quiet for weeks and would not say which team the role was for.

Stage 3: Phone screen

At a glance

  • Length: not stated; 1 or 2 sessions
  • Questions per round: 1 to 3 questions per session
  • What was asked: coding (parsing, streaks, sampling), SQL with window functions, statistics, a product case
  • Most common: no single question repeats

How it works. One or two sessions. Reported formats: one coding session and one behavioral session; one product case and one technical session; or a coding session with three short questions. Passing it got candidates a virtual onsite a few weeks later.

Questions candidates reported:

Question (our summary)What it tests
Count "super streak" segments in an event log; then the rules change partwayCoding, state tracking
Next-word predictor; make predict O(1); then return a probability-weighted wordCoding, hash maps, sampling
Match payments to invoices by memo ID, then fall back to amount and earliest due dateCoding, parsing
Sample uniformly from a 2D square using a 1D uniform generatorProbability, coding
Longest continuous increasing subarrayCoding
Estimate a percentile from bucketed countsStatistics, interpolation
A retention-rate SQL query with many conditions (needed LAG and several CTEs)SQL
Compare the profit of two stores near a school gateProduct case
P-values, simple linear regression, confidence intervals; compare two vendors' modelsStatistics

More questions: Browse Google questions.

Stage 4: The virtual onsite

How it works. Two or three 45-minute sessions, sometimes split over two days. One interviewer cancelled the night before and the loop ran with two sessions. The three session types candidates named:

Measurement and Modeling Concepts

At a glance

  • Length: not stated
  • Questions per round: 4 to 7 questions built around one case
  • What was asked: product metrics, retention modelling, evaluation without an A/B test, statistics
  • Most common: Google Meet adoption or call drops

Questions candidates reported:

  • Google Meet adoption dropped. What do you do?
  • A large customer says Meet calls keep dropping: plan the analysis and estimate the impact.
  • If it affects contract renewal, how would you model retention, and when is logistic regression enough?
  • A new version shipped but you cannot run an A/B test. How do you evaluate it?
  • Fix a bug or build something new: how do you decide?
  • Your favourite model and how it works; problems you hit while training a model.
  • A confidence interval is too wide: what changes it? Then estimate a linear model with more features than data points.

More questions: Browse Google questions.

Coding, Applied Analysis and Experiments

At a glance

  • Length: not stated
  • Questions per round: 2 to 3 questions
  • What was asked: experiment design, success metrics, minimum detectable effect, segmentation, forecasting
  • Most common: A/B test design

Questions candidates reported:

  • Google Maps recommends a best jogging route. How do you measure if it is good? Design the experiment, the success metrics, the trigger logic and the minimum detectable effect.
  • What user segments would you build for Gmail, and how would you use them?
  • A forecasting question made up on the spot, testing modelling intuition.
  • Basic data manipulation, plus how you would choose metrics or methods.

One candidate said no code was written in this session despite its name.

More questions: Browse Google questions.

Googleyness and Leadership

At a glance

  • Length: not stated
  • Questions per round: 3 to 4 general questions
  • What was asked: fit for Google, team problems, missed dates, bad decisions
  • Most common: none repeated

Questions candidates reported: your background and why you fit Google; a problem your team hit and how you found and fixed it; a project that missed its date; a bad decision and what you would do differently.

More questions: Browse Google questions.

Stage 5: Team match and offer

How it works. In one report, the onsite feedback came three days later, team matching took about a week and the offer followed. A DSR candidate heard two days after the onsite that several teams were interested. One rejection came about three weeks after a team-lead conversation that covered only the resume.

The Codinzhub Next-Round Playbook

What data scientists who moved on did. Put together by the Codinzhub team from the DS experiences in this guide.

To get past the phone screen

  • Practise SQL with window functions until LAG and several CTEs are quick.
  • In coding questions, clarify the rules first, then handle the follow-up by changing your state cleanly.

To get past Measurement and Modeling

  • Treat every case as a discussion: assumptions, metric, analysis plan, then impact.
  • Prepare to evaluate a launch without an A/B test.

To get past Applied Analysis and Experiments

  • Know how to set a success metric, trigger logic and minimum detectable effect.
  • Go past the first answer. For a bucketed percentile, interpolate inside the bucket instead of taking the midpoint.

To get past Googleyness and Leadership

  • Prepare stories on team problems, missed dates and bad decisions.

Before you start

  • Ask the recruiter which track and team you are interviewing for, and what each session covers.

Topics to prioritise

Across these experiences, the most common topics are product sense and metrics, behavioral, probability and statistics, experimentation and A/B testing, SQL. A/B test design was the most repeated question type, followed by defining a success metric.

About this guide

This guide is researched and written by the Codinzhub team. We read every Google interview experience we could find, kept only what candidates actually described, and rewrote it in our own words. Questions are our own short summaries; the practice links point to similar problems, not the exact interview questions. Found something out of date? Share your own experience on Codinzhub and we will update the guide.

Ready for the next topic?

Keep track of what you’ve learned.