Data Analyst Interview Experience
Interview process
After a live coding round, I had a "Data Analysis" interview: 5 case studies built around one product scenario — an in-app ads widget for partner offers. Cases covered probability (joint + conditional/Bayes), funnel and metrics design with limited tracking, dashboard design, and A/B testing including quasi-experiment alternatives. It felt more like a structured conversation than an exam: we ran over time on some cases and the interviewer was fine with it, asked guiding questions when I got stuck, and cared more about reasoning than polished delivery. What didn't go well: explaining specific concepts precisely under time pressure was harder than expected, especially as a non-native speaker — but the interviewer worked with that rather than against it.
- Technical interview
Interview tips
Think about how to reason about a whole feature end-to-end: what data you'd need, what's missing, what matters most — the cases reward big-picture product thinking, not memorized frameworks. Definitely refresh probability (Bayes, sequences of events, "at least one") and practice explaining it out loud — it's clear why they ask it, and even if you stumble, showing you understand what you're doing and why counts more than a perfect answer. Treat it as a conversation: ask clarifying questions, state assumptions, think out loud.
Company culture
No pressure tactics despite Revolut's intense reputation. The interviewer genuinely tried to understand my thinking, gave hints when needed, and didn't punish imprecise wording as long as the logic was sound. It felt like being evaluated by a future colleague, not tested by a gatekeeper.
Questions asked
Overview
All cases revolved around one theme: an in-app advertising widget promoting partner offers.
Question types asked
Specific questions asked
Metrics & funnel design: what metrics to track for the widget; how to build a funnel when we only see the final transaction outcome — user is redirected to the partner's site, we don't observe intermediate drop-off steps.
Dashboard: design a dashboard for this feature (audience, metric tree, guardrails).
Experimentation: design an A/B test for the widget
then — if an A/B test is not possible, how would you measure impact instead (quasi-experiments: diff-in-diff, holdout markets, interrupted time series, etc.).
Probability (joint + conditional): widget open rate 20%; from widget, user opens one of three tabs — products (50%), hotels (30%), travel (20%) — with purchase conversion 60%/20%/10%. Q1: overall purchase probability. Q2: given a purchase happened, probability it came from products (Bayes).
Get full access with a membership, or share your experience to try it free.
