Interview Prep
Common interview questions for data roles
Data interviews combine technical depth with business communication. The SQL questions test your craft. The case questions test whether you can frame a problem before you start writing queries. These cover both.
PrepVault organizes your prep and keeps your research by company. Free to start.
Common interview questions for data roles
How does the multiple testing problem affect A/B experiment interpretation?
many tests inflate false positives. primary metric predefinition. Bonferroni correction. reduced power tradeoff. exploratory vs confirmatory
How would you design an A/B test for a real product change to a checkout flow?
clear hypothesis. stable randomization. primary metric. guardrails. predefined stopping rule
How should dashboards be designed to support decision-making rather than just reporting?
start from decisions. avoid metric overload. context and exceptions. drill-down paths. freshness and definitions
How would you design a production data pipeline for daily revenue reporting?
design from SLA. idempotent daily load. revenue reconciliation. late adjustments. monitoring and ownership
How do you design a multi-touch attribution model that accounts for the full customer journey rather than a single channel?
choose a model that reflects the business's theory of how buyers decide. position-based is a practical default for most companies. data-driven requires sufficient conversion volume. identity resolution across devices is a prerequisite
How would you align cross-functional teams on metric ownership and documentation?
business and technical owners. full metric definition. catalog or semantic layer. cross-functional review. change management
What is a holdout group and how does it differ from a standard A/B test control group?
holdout is long-running vs. test control which is temporary. measures cumulative effect of all shipped changes. catches interaction effects between features. typically 2-10% of users maintained for a full quarter or more
How do you design a metrics review process that prevents metric rot - where dashboards accumulate stale or unused numbers?
every metric needs an owner. quarterly audit with three questions. removal should be easy and feel like progress. forcing function prevents passive accumulation
How do you use cohort analysis to evaluate whether a product change improved long-term retention?
align cohorts by lifecycle week, not calendar date. retention curve shift that persists beyond novelty window is evidence of improvement. check for acquisition mix changes as a confounder. sample size must be sufficient to trust the comparison
How would you define a product activation metric?
first value moment. clear numerator and denominator. time window. validate against retention. avoid signup-only metric
Common mistakes
- Writing a query before framing the question
- Confusing correlation with causation
- Not stating the assumptions in your analysis
What interviewers weigh
- How you frame a vague business problem
- Whether you can explain findings to non-technical people
- Rigor in your SQL and statistics