The Capital One data scientist loop is a case interview as much as a technical one. Of 245 reported questions, 127 come from technical screens and 81 from onsites, with 25 from HR screens and 12 from online assessments; candidates report a Power Day onsite and case-style rounds. Analytics and experimentation is the largest category at 76 questions, ahead of statistics and math at 57 and behavioral and leadership at 52. SQL and Python data manipulation adds 23, machine learning 20, and coding and algorithms only 15. Difficulty sits mostly in the middle, at 174 medium against 36 hard and 35 easy. The recurring shape is a business scenario you have to structure yourself. Reported cases include evaluating growth and pricing for a grocery delivery startup, choosing between shows, computing gala versus online break-even donors, and modeling network-service unit economics and breakeven. Modeling questions stay applied, such as designing delay and watchlist models or building and evaluating a donation propensity model, and SQL appears as revenue work like computing campaign net revenue. Experiment design comes up as fundraising experiments with guardrails. Behavioral questions ask about resolving ambiguity under pressure. 34 first-hand interview experiences cover the loop, including screens on streaming show renewal, ride-share profitability and a Groupon case.