Amazon Data Analyst Interview Loop
For an Amazon data analyst interview, know the title first. At Amazon, this job is called Business Intelligence Engineer. That's the title on the posting, the one the recruiter uses, and the one in every report. The loop starts with a recruiter call and a SQL gate. The gate is an online assessment or a phone screen, sometimes both, and each has a Leadership Principles question attached. Then there are three to five rounds: SQL and analysis on a live notepad, visualization and metrics, a business case, the hiring manager, and the Bar Raiser. Candidates all agree that the manager and Bar Raiser rounds decide the outcome.
What decides it
Leadership Principles, told well. Expect around ten behavioral questions, and don't repeat a story. The technical rounds are a gate. The behavioral rounds are the real interview.
Practice by round
Joins, aggregations, date and string functions, and a ranking problem. There are about seven questions on the assessment, or two or three live. It ends with one Leadership Principle question, and even here they expect a story with a number in it.
Three problems on a data model you're given, using CTEs, window functions and ranking. They cover best-selling items, the most expensive products per category, month-over-month percentage difference, rolling monthly sales and shipments per month.
The business case, set as mini-analyses on Amazon's own operations. They cover delivery promise misses, split shipment rate, a Buy Again rollout and duplicate transactions, and each one ends in a number a business leader would act on. In the real loop, the visualization-and-metrics interview asks what a weekly business review needs on its first page. Practice that as well. Both rounds score whether you think like an operator.
| Case study | Difficulty |
|---|---|
| DCF for a Fulfillment Center | Hard |
| Data project | Difficulty |
|---|---|
| Resolving Duplicate Transactions | Medium |
| Delivery Promise Misses | Medium |
| Split Shipment Rate | Medium |
| Buy Again Rollout | Hard |
Statistics the way a BIE uses it. That covers reading a p-value, random bucketing, a hypothesis about broken packages, and when a difference means nothing.
| Question | Difficulty |
|---|---|
| Linear Regression & T-test | Medium |
| Broken Packages | Medium |
| Multi-dimensional Data | Medium |
| Statistical Hypothesis | Medium |
| An Instrument From a Rollout Glitch | Hard |