Capital One Data Scientist Interview Experience — A Four-Round Power Day Onsite

1. Timeline & Full Process

Online application / referral -> Recruiter Screen (15-20 min) -> OA (Assessment): you can check other posts on the forum for what this looks like -> Power Day: four rounds back to back, 45-60 minutes each.

2. Power Day Breakdown

Round 1: Case Interview (Business Analyst Case)

Scenario: you're given a concrete credit business scenario (for example, a credit-limit-increase strategy for a specific customer segment, or a profitability analysis for a new product entering the market).

Core logic: whatever you do, don't jump straight to modeling!

• Step 1: Define the goal. Are you trying to increase revenue, or lower the default rate?
• Step 2: Break down the metrics. Write out the P&L formula: Profit = (Revenue - Loss - Operational Cost).
• Step 3: Find the levers. What tools (interest rate, credit limit, risk models) can you use to move these metrics?

Note: there's some simple mental math involved, testing how comfortable you are with numbers.

Round 2: Technical / Machine Learning

Deep follow-ups: it's not just asking about algorithm theory. If you mention using XGBoost, the interviewer will follow up: how do you handle missing values? How do you prevent overfitting?

Coding/SQL: the SQL portion tested window functions (like RANK()) and complex joins; the Python portion leaned toward data wrangling.

Round 3: Role Play (Communication)

Simulated scenario: you're a DS and you have to explain to a Product Manager with zero technical background why a model performed great on the validation set but got worse after going live.

Round 4: Job Fit / Behavioral

Standard STAR format. Come prepared with stories about "handling conflict," "overcoming a tough technical problem," and "dealing with ambiguous requirements."

Wishing everyone lots of offers!