Interview Questions for a Data Analyst

Write a query to find the second-highest salary per department. Walk me through your approach.

What to look for: Correct use of window functions or a subquery, handles ties, and reasons about NULLs and empty departments.

Explain the difference between an INNER JOIN and a LEFT JOIN with a concrete example of when the wrong one causes a bug.

What to look for: Understands how a LEFT JOIN preserves unmatched rows and how an INNER JOIN silently drops data, a common source of wrong metrics.

How do you define a metric like 'active user,' and why does the definition matter?

What to look for: Recognizes that ambiguous definitions break comparisons, sets explicit time windows and criteria, and documents them.

You're given a CSV with duplicates, inconsistent dates, and nulls. How do you clean it?

What to look for: Systematic profiling, deduping on a key, standardizing formats, deciding how to treat nulls, and validating row counts before and after.

When would you use a median instead of a mean, and why does it matter for reporting?

What to look for: Understands skew and outliers, uses median for skewed distributions like income or latency, and reports the right summary.

How do you choose between a line chart, bar chart, and table for a given question?

What to look for: Matches chart to intent: trends over time, category comparison, or precise values, and avoids misleading visuals like truncated axes.

A dashboard number looks wrong to a stakeholder. How do you verify whether it's a data issue or a real change?

What to look for: Traces the number to source, checks the pipeline and filters, reconciles against a known total, and rules out data quality before alarm.

Explain how you'd calculate week-over-week retention or a cohort metric in SQL.

What to look for: Correctly anchors cohorts by signup period, uses date logic and self-joins or window functions, and avoids double-counting users.

Tell me about a time your analysis changed a business decision.

What to look for: Connects analysis to outcome, communicated clearly to decision-makers, and understood the business context not just the data.

Describe a time you found a data-quality problem others had missed.

What to look for: Healthy skepticism, attention to detail, and the habit of sanity-checking results before publishing them.

How do you handle a request for a report when the underlying data isn't trustworthy?

What to look for: Transparency about limitations, fixing or flagging the data issue, and not delivering a confident-looking but wrong report.

Tell me about a time you simplified a confusing dashboard or report for non-technical users.

What to look for: Empathy for the audience, ruthless focus on the decision the report drives, and removing noise rather than adding charts.

A stakeholder asks for 'all the data' on a topic. How do you turn that into a useful deliverable?

What to look for: Clarifies the decision behind the request, scopes the question, and delivers a focused answer rather than a data dump.

Two reports show different numbers for the same metric. How do you reconcile them?

What to look for: Compares definitions, filters, time zones, and refresh times, and finds the source of divergence systematically.

You have one hour to answer an urgent question with imperfect data. What do you do?

What to look for: Prioritizes a defensible directional answer, states assumptions and caveats, and avoids false precision under time pressure.

A weekly metric spiked sharply. How do you tell a real trend from a tracking or pipeline artifact?

What to look for: Checks for instrumentation changes, deploys, seasonality, and bot traffic before declaring it a genuine business shift.