Coinbase Senior Data Scientist Interview Experience — SQL, A/B Test Design, and a BTC Volatility Case

Background:
Bachelor's degree in China + Master's in the US, Statistics / Data track
3+ years of data analysis / risk-control experience (mostly finance-related)
Applied for: Data Scientist – Risk / Product Analytics (remote)

1. Overall Process

HR Screen → SQL & Stats → Product Case → Hiring Manager → Team Fit

The whole process took about 4 weeks. Not super fast, but the feedback at each stage was pretty clear.

2. HR Phone Screen (30 min)

Mostly background and motivation, no technical questions.

Questions:

• Why Coinbase?
• What data projects have you worked on before?
• Have you dealt with problems that were vague and had no standard answer?

I got the feeling Coinbase really cares about whether you can work self-directed in an uncertain environment — the remote culture came through clearly.

3. SQL + Data Analysis Interview (the most important round)

Wrote SQL in CoderPad, 2 questions total.

Question 1 (time window + active users)

There's a table trades(user_id, trade_time, volume, asset).

Find the percentage of users with 3 or more trades in the past 30 days.

What it's testing: time filtering, group by, count distinct, and how you define the denominator.

Question 2 (retention)

Users table users(user_id, signup_date), trades table trades.

Find 7-day retention.

The interviewer followed up: how do you define retention — first trade or first login?

The SQL itself wasn't that hard, but they really cared about whether your thinking was clear.

4. Statistics / Experiment Interview (mostly A/B testing)

This round felt more like a conversation, but the questions were tough.

Classic question: Coinbase wants to increase the first-trade rate for new users — how would you design an experiment?

My answer structure:

• Pick the metric (first trade within 7 days)
• Treatment group vs control group
• Potential bias (selection bias, user learning effect)
• What to do if the result isn't significant

The interviewer didn't really care about the formulas — they cared more about why you designed it that way.

5. Product / Business Case (very much a Coinbase-style round)

The case was something like: BTC has been very volatile lately, but retail trading volume has dropped — how would you analyze this?

I broke it into 4 layers:

• Market layer (volatility vs sentiment)
• User layer (new vs existing users, retail vs institutional)
• Product layer (trade failure rate, fees)
• Compliance / risk constraints

The focus was on structured thinking plus how the data supports a decision.

6. Hiring Manager Interview (mostly behavioral)

The questions leaned heavily toward ownership:

• Tell me about a time data changed a decision
• When data quality was bad, what did you do?
• How do you prioritize ambiguous requests?

You could clearly feel that Coinbase values independence, decision-making judgment, and the ability to work with PM / Eng / Legal.

7. Overall Impression

The SQL and stats weren't brutally hard, but the bar for business understanding was high. They put a lot of weight on data → decision → business impact. You don't need to be a crypto expert, but you can't be totally clueless about it either.

8. My Advice

• Get really comfortable with SQL (time windows, retention, rank)
• Don't just memorize the t-test for A/B testing
• Prepare more ambiguous business cases
• A finance / risk-control background is a plus