Google Growth Marketer Interview Guide

Honestly, the process was way more intense than I expected. It started with a recruiter call that felt pretty relaxed — mostly background, why I’m interested in the role, and whether I understood the business side of data analytics. I felt confident there because I could speak naturally about my experience and projects.

The next round was the HackerRank/SQL assessment, and that’s where reality hit a little. The questions themselves weren’t impossible, but the time pressure made it stressful. There were a couple of joins and aggregation problems where I knew the logic, but I had to really slow down and think through edge cases. I remember one question involving ranking and filtering where I genuinely started sweating because I knew I was close but couldn’t get the exact output at first.

After that came the live technical interview, which surprised me the most. I expected them to only care about writing SQL queries, but they focused heavily on thought process. They kept asking “why this approach?” and “how would this scale with more data?” At one point I blanked on a window function syntax, but I talked through the logic instead of panicking, and that actually helped.

The final round felt much more business-focused. They gave scenarios around retail KPIs, customer behavior, and dashboards, and wanted to see how I’d communicate insights to non-technical teams. That was probably where I felt strongest because it became more about analytical thinking than perfect syntax.

Overall, the biggest surprise was that they cared less about memorizing SQL and more about structured thinking, communication, and how you approach problems when you don’t immediately know the answer.

Questions asked: Question (SQL / Analytics):
You are given a table user_activity(user_id, activity_date). Write a query to find users who were active for at least 3 consecutive days, and return the number of such users.

Answer:

Approach: We need to detect streaks of consecutive dates per user. A common trick is to use ROW_NUMBER() to create a “streak grouping key”.

When you subtract ROW_NUMBER() from the date, consecutive dates will share the same difference value.

sql WITH ranked AS ( SELECT user_id, activity_date, ROW_NUMBER() OVER (PARTITION BY user_id ORDER BY activity_date) AS rn FROM user_activity ), grouped AS ( SELECT user_id, activity_date, DATE_SUB(activity_date, INTERVAL rn DAY) AS grp FROM ranked ), streaks AS ( SELECT user_id, COUNT() AS streak_length FROM grouped GROUP BY user_id, grp ), qualified_users AS ( SELECT DISTINCT user_id FROM streaks WHERE streak_length >= 3 ) SELECT COUNT() AS users_with_3_day_streak FROM qualified_users;

What the interviewer looked for:

• Correct use of window functions (ROW_NUMBER)
• Ability to transform a “sequence problem” into grouping logic
• Handling edge cases (duplicate dates, gaps in activity)
• Clean explanation of why the grouping trick works

Follow-up they asked: “If this dataset had 5 billion rows, how would you optimize it?”

Good answer direction:

• Partition data by user_id
• Pre-aggregate daily activity (remove duplicates)
• Use indexed activity_date
• Possibly compute streaks in a batch pipeline instead of on-the-fly SQL