A dashboard builder and a decision-changing analyst have the same job title, and the difference costs real money. These questions test SQL depth, statistical honesty, and the judgment to push back on bad questions — each with notes on what a convincing answer includes.
What a strong answer covers
The mechanics are table stakes; the story is the test. Good answers involve rows silently dropped (inner join filtering out customers with no orders) or duplicated (joining on a non-unique key inflating revenue). Analysts who have been burned check join cardinality by habit.
What a strong answer covers
Running totals, rank within group, period-over-period comparisons using LAG or LEAD, deduplication with ROW_NUMBER. They should articulate the difference from GROUP BY: keeping row-level detail while adding aggregate context. A concrete business example beats textbook syntax.
What a strong answer covers
Interrogate the definition: churn of what population, over what window, voluntary versus involuntary, logo versus revenue churn. Strong analysts negotiate a written metric definition and check whether one already exists before creating a competing version.
What a strong answer covers
A systematic reconciliation: compare grain and filters, time-zone and date-boundary handling, refunds and credits treatment, currency conversion, and pipeline freshness. The habit of finding which system is right rather than assuming the warehouse wins shows maturity.
What a strong answer covers
Sample size awareness, variance, confidence intervals or significance testing at an appropriate level, and honesty about multiple comparisons. You want someone who says 'this difference is within noise' out loud even when it is unpopular.
What a strong answer covers
Recognition that daily peeking inflates false positives, questions about pre-registered sample size and stopping rules, and interest in the effect on guardrail metrics. They should recommend a proper fixed-horizon or sequential design next time, not just accept the number.
What a strong answer covers
Centralized definitions — dbt metrics, a semantic layer, or at minimum a governed source-of-truth model — instead of every dashboard re-deriving revenue its own way. Version control and code review for analytics logic is a strong modern signal.
What a strong answer covers
A concrete story with stakes: how they validated the result before escalating, how they communicated it, and what happened. You are hiring for the spine to deliver unwelcome numbers with evidence, and this question reveals it.
What a strong answer covers
Quantify the damage first: how much is missing and whether it is random or systematic. Then explicit handling — exclusion with documented impact, imputation with stated assumptions, or bounding the result both ways — and a caveat in the deliverable. Silent cleaning is the anti-pattern.
What a strong answer covers
Executives get the answer first, the recommendation, and the risk, in one page or three slides; the appendix holds the method. Product teams get more granularity and interactivity. Tailoring by audience, not one giant dashboard for everyone, is the mark of someone who has done this.
What a strong answer covers
Aggregates hide mix shifts — new-user growth can mask collapsing retention. They should describe defining cohorts by acquisition period or behavior, reading retention curves, and at least one decision the cohort view changed that the aggregate view would have missed.
What a strong answer covers
Fluency in SQL plus a warehouse, a BI tool, and ideally dbt and a scripting language. The 'replace tomorrow' half tests critical thinking about tooling rather than attachment — and tells you how they will react to your stack.
Skip the interviews entirely — get matched with pre-vetted Data Analyst developers in 48 hours, $0 until you hire.
Need a custom question set?
Our free interview question generator builds a tailored list for any role, seniority, and focus area.
Try the interview question generator →Yes, a short one — 20 to 30 minutes on a realistic schema with a join, an aggregation, and one window-function problem. Watching someone check their join cardinality unprompted tells you more than an hour of conversation.
Analyst. If your need is predictive modeling in production, add machine-learning questions from our data science set. If the honest need is decision support and metrics, these questions target exactly that and an analyst is the cheaper, faster hire.
Hiring the best tool operator instead of the best question-asker. Candidates who interrogate the churn question before answering will save you from months of decisions made on a confidently wrong number.
Hire directly
Hire vetted Data Analyst developers in the USA →Other interview guides
Vetted talent ready for US teams. No recruitment fees. Zero risk.
🇺🇸 Trusted by companies across the United States