Data Scientist interview
at Amazon.

Tell-me-about-a-time stories on ownership, conflict, and impact — scored against this company's real values.

1. 01 Tell me about a time you dove deep into a messy dataset and uncovered a root cause that contradicted what the metrics or stakeholders initially believed (Dive Deep).
2. 02 Describe a situation where you disagreed with a product or business stakeholder about the interpretation of an experiment's results, voiced your concern, and then committed to the final decision (Have Backbone; Disagree and Commit).
3. 03 Give me an example of when you used data to challenge an assumption on behalf of customers and changed the direction of a feature (Customer Obsession).
4. 04 Tell me about a time you delivered an analysis under a tight deadline with incomplete data and still drove a measurable business outcome (Bias for Action / Deliver Results).

The hard screen for the craft itself — talked through out loud, not whiteboarded in your head.

1. 01 How would you design and interpret an A/B test to evaluate a new Prime feature, including how you would choose sample size, handle novelty effects, and decide when to stop the test?
2. 02 Given a SQL table of customer orders and returns, write a query using window functions to find the products with the highest return rate among customers in their first 30 days.
3. 03 Customers are returning a particular product category at an unusually high rate; how would you investigate the cause statistically and what metrics would you build to monitor it going forward?
4. 04 Explain how you would detect and correct for selection bias when measuring the impact of a personalized recommendation widget on the Amazon homepage.

Open-ended design of the systems this company actually runs, with the interviewer probing your tradeoffs.

1. 01 Design the experimentation and metrics platform that lets Amazon teams run thousands of concurrent A/B tests on the retail website without contaminating each other.
2. 02 Design a data pipeline and model to forecast next-week demand for individual SKUs across Amazon fulfillment centers.
3. 03 Design a metric and anomaly-detection system to catch sudden drops in Buy Box conversion for third-party sellers.
4. 04 Design an offline-to-online feature store that powers churn-prediction models for Amazon Prime subscribers.