Data AnalystInterview Answers

I start by clarifying the business question — a vague request like "analyse our sales" is too broad. I probe to understand what decision the analysis will inform. Then I assess data availability and quality, design the analytical approach, and validate assumptions with stakeholders before building. I present findings with the "so what" upfront, then support with evidence. The goal is actionable insight, not a data dump.

I build data validation into every pipeline: checking for nulls, duplicates, outliers, and referential integrity. I profile data before analysis — understanding distributions, row counts, and expected value ranges. I document data lineage and transformations so others can audit my work. When I find anomalies, I investigate root cause before proceeding. Good analysis is only as good as the underlying data, so I'm never shy about raising data quality issues with engineering teams.

I analysed customer churn patterns for a SaaS product and found that customers who didn't complete onboarding within 14 days had a 3x higher churn rate. I presented this to the product team with cohort analysis and survival curves. They redesigned the onboarding flow and added human check-ins at day 7. Three months later, 90-day retention improved by 18%. That's the kind of analysis that changes how a business operates.

SQL is my primary tool for data extraction and transformation — I'm comfortable with window functions, CTEs, and query optimisation. I use Python (pandas, NumPy, matplotlib, seaborn) for more complex analysis and automation. For visualisation and reporting, I use Tableau, Power BI, or Looker depending on the stack. I've worked with BigQuery, Snowflake, and Redshift as cloud data warehouses. I pick the right tool for the question — not the most impressive one.

I translate numbers into business language. Instead of "the R-squared is 0.87," I say "our model explains 87% of the variation in revenue." I lead with the insight and recommendation, then provide evidence for those who want to dig deeper. I use clear, well-labelled visuals — avoiding chart junk — and build narratives that tell a story. I also prepare for questions: I know my data well enough to answer "what if" questions in the room.

I welcome contradictions — they often reveal the most valuable insights. First, I double-check my analysis for errors. If the data holds up, I investigate why it contradicts expectations. Sometimes it's a data quality issue upstream; other times the data is revealing a business truth that assumptions haven't caught up to. I present contradictions transparently, with context and a recommendation for further investigation. I never adjust the data to fit the expected answer.

I've designed and analysed dozens of A/B tests — from landing page optimisation to pricing experiments. I calculate required sample sizes upfront to avoid underpowered tests, determine run duration based on traffic, and check for pre-existing differences between groups (A/A testing). I use appropriate statistical tests — t-tests, chi-squared, or Bayesian methods depending on context — and always consider practical significance alongside statistical significance. I also know when not to run an A/B test.

I score requests by potential business impact, data availability, and stakeholder urgency. I maintain a transparent backlog and discuss priorities with my manager and key stakeholders regularly. I'm not afraid to push back on low-value requests or challenge the framing of a question if a different analysis would be more useful. I also proactively propose analyses I think will be valuable — waiting for requests misses opportunities to add value.