Engineering and data
A first round for data analysts should answer three things: do they start from the business question, do they distrust numbers until they’ve checked them, and can they explain a finding so someone acts on it. It complements a technical round on SQL or a take-home. It doesn’t replace one.
Follow-ups ask how they checked a number, which is where real analysis and a rehearsed story come apart.
What to screen for
The questions
Ask them in order. Listen for specifics: what happened, what they did, how it ended.
Strong answer
States the business question, the data used, the finding, and what the team did differently afterward.
Weak answer
Describes a dashboard they built with no decision attached to it.
Strong answer
Checks the drop is real first, like a tracking change or late data, then splits by channel, device and region, looks for releases or ended campaigns, and reports early with caveats.
Weak answer
Jumps to a cause straight away, or starts building a model.
Strong answer
Names the real cause, like duplicate rows from a join or a changed definition, traced it to the source, and told the people who’d already used the numbers.
Weak answer
Removed outliers until the result looked reasonable.
Strong answer
Explains the logic in plain words, like the grain of each table or how they handled duplicates, and how they checked the output against a number they trusted.
Weak answer
Can’t explain it without the code in front of them, or never checked the result.
Strong answer
Uses a concrete example, names the likely hidden factor, and suggests a test or a cheaper way to check before committing.
Weak answer
Gives a lecture on theory, or goes along with it to avoid the conversation.
Strong answer
Delivers what was asked, explains the gap plainly, and proposes a better measure with reasons.
Weak answer
Builds it without comment, or refuses until it’s changed.
Strong answer
Starts from the decisions the viewer makes, keeps it to a few metrics with visible definitions, and removes what nobody uses.
Weak answer
Includes everything available in case someone needs it.
Strong answer
Owned it quickly, corrected it with everyone who’d seen it, and added a check so it wouldn’t happen again.
Weak answer
Can’t think of one, or fixed it quietly and hoped nobody noticed.
Before the interview
Quick facts to read alongside the interview. They aren’t scored.
Listen for
Avoid
Use this in LessRounds
Paste the questions into a scripted interview, add the scorecard skills, and share one link. Every data analyst candidate gets the same questions, and you get a score for each skill with the quote behind it.
Partly. Ask candidates to talk through a query they wrote and how they checked it. That shows whether they understand joins and grain. Test writing SQL in a later technical round.
Senior analysts push back on the question before answering it, and they know which findings are worth a manager’s time. Junior analysts tend to answer exactly what they were asked.
Start with 10 free interviews a month. No card, no sales call.