Data analyst interview questions

The rounds

The case study, and a structure for it

A typical prompt: "Daily active users dropped 8% last week. Find out why." There is no correct answer, and interviewers are scoring your approach. A structure that works:

• Check the data first. Is this real, or is it a tracking bug, a pipeline failure or a definition change? Candidates who jump straight to product explanations lose points here, and in real life this is the answer surprisingly often.
• Segment. Platform, geography, new versus returning, app version. A drop concentrated in one segment is a different problem from a uniform one.
• Check the timeline. Did it drop on one day or decline gradually? Did anything ship?
• Consider the outside world. Seasonality, holidays, a competitor launch, a marketing campaign ending.
• State what you would need to confirm the leading hypothesis, and what you would do about it.

Metrics questions

• How would you measure the success of a new feature? Expect to defend one primary metric rather than list ten.
• What is the difference between a good metric and a vanity metric?
• This metric went up and the business got worse. How is that possible?
• Define DAU for a product where users leave the tab open all day. Definitions are the question.
• What guardrail metrics would you watch alongside the one you are trying to move?

Statistics, at the level actually asked

• How would you design an A/B test for this feature, and how long would you run it?
• What is a p-value, in words a product manager would understand?
• What is the problem with checking the test every day and stopping when it is significant?
• Correlation and causation, with an example from something you have worked on.
• When would you not run an A/B test?

The depth expected is applied rather than theoretical. Being able to explain significance clearly matters more than deriving anything.

The SQL round

This is the gate. Analyst SQL leans on window functions, cohort and retention queries, and conditional aggregation. The pattern set is in SQL interview questions, and the retention query - users active in week N after signup - is worth being able to write from memory, because it comes up constantly.

Where CoPilot Interview fits

Disclosure: we build an interview assistant, so treat this as a disclosed pitch. CoPilot Interview is a native Windows and macOS app that puts suggested answers on your own screen in about two seconds, across coding, system design and behavioral rounds, with a permanent free tier that needs no credit card.

There is a related role page at data science interview help.

Frequently asked questions

What questions are asked in a data analyst interview?

A live SQL round, a case study turning a vague business question into a measurable one, metrics and product sense questions, applied statistics around A/B testing, and a communication exercise explaining a result to a non-technical stakeholder.

How do you answer a 'metric dropped, find out why' question?

Check whether the data is real before explaining it - tracking bugs, pipeline failures and definition changes are common causes. Then segment by platform, geography and user type, look at whether the drop was sudden or gradual and what shipped, and consider seasonality and external events.

How much statistics do data analysts need for interviews?

Applied rather than theoretical. You need to design an A/B test, explain a p-value in plain language, know why peeking at results daily is a problem, and distinguish correlation from causation with a real example. Derivations are rarely asked.

Is SQL the most important skill for a data analyst interview?

It is the most common gate. Analyst SQL leans on window functions, cohort and retention queries and conditional aggregation, and the retention query in particular is worth being able to write from memory.