AI data analyst interview with hands-on SQL
Two questions, and the first is written, not described: you build a SQL query in a shared editor while the AI interviewer reads along. The second asks how you would tell whether a report nobody trusts has a data problem or a logic problem.
Free to take with a work email. No account needed.
The 2 questions the AI interviewer asks
Word for word from the demo. In the demo and in a real interview, the AI also asks its own follow-ups from what the candidate says.
- 1In the code editor
“In the code editor, write a SQL query against a table orders with columns customer_id, order_date and amount that returns the top 3 customers by total spend over the last 90 days. Explain your choices as you go.”
What a strong answer shows
- Filters the 90-day window before grouping, using their dialect's date arithmetic
- Groups by customer_id, sums amount, sorts descending and limits to three
- Raises ties, null amounts and refunds, and how each would change the result
- 2
“You inherit a report whose numbers nobody trusts. How do you go about finding out whether the data or the logic is wrong?”
What a strong answer shows
- Reconciles a few totals with the source system before reading any logic
- Checks common culprits: join fan-out, filter mistakes, time zone cutoffs, stale data
- Confirms metric definitions with the report's owners and writes them down
What this demo shows
A data analyst's SQL is easiest to judge while it is being written. In this demo you type the query in the shared code editor, and the AI interviewer sees it take shape and asks about it, so the reasoning behind the query becomes part of the answer. The second question turns from syntax to judgment. When an employer runs an AI interview for data analysts, it uses their own questions and scorecard, can include more coding exercises, and saves the work to the report.
The demo and a real interview
- Aspect
- Demo
- Real interview
- Questions
- 2 sample questions
- Your own, with follow-ups your team writes plus the AI's own
- Length
- About 10 minutes
- 30 to 45 minutes is normal
- Scoring
- A scored report in your inbox
- Your scorecard and grading guides, with a scoring trace
Common questions
Can data analysts answer in Python instead of SQL?
In a real interview, yes, if you write the question that way. The shared code editor supports SQL, Python, Java, C/C++, JavaScript/TypeScript, Go and Rust, so the exercise can match the tools your analysts use every day. This demo sticks to SQL.
How do I know the person interviewing is the applicant?
Switch on the identity check for the role. Before the first question, the candidate's government ID and a selfie are checked through Stripe Identity. Identity verification is an optional add-on at $2.50. Phone interviews do not include it.
How is a data analyst interview scored?
Against your scorecard: weighted skills, for example SQL and data validation, each with a written guide that describes a Novice, Competent, Proficient and Expert answer. The same guide applies to every candidate. Scores are recomputed in code, and the scoring trace quotes the evidence for each skill and shows the arithmetic.
Interview your own candidates this way
Your questions, your scorecard, the same guide for every candidate.