Most candidates prepare for "the Amazon interview" as if it were a single event. It is not. It is a pipeline of five or six distinct gates, and each gate filters for something different. A candidate can be a strong coder and still fail at the behavioral phone screen, or tell great stories and still stumble in the work simulation. Understanding what each stage is actually testing is the difference between targeted preparation and generic effort.
This guide walks the pipeline in order — application, recruiter screen, online assessment, phone screen, the loop, and the debrief — and explains what interviewers evaluate at each step and how to prepare for that specific step.
Stage 1: The application and the first filter
Every process starts at amazon.jobs, where postings carry a requisition ID, a job family, and a level (an SDE II posting, for example, maps to an internal level with specific scope expectations). Recruiters and sourcing systems screen applications against that posting's requirements before a human conversation ever happens.
What actually gets evaluated here:
- Keyword alignment with the posting. The posting lists required skills and tools literally. Resumes that mirror that vocabulary — the specific programming languages, the domain tools, the named responsibilities — get noticed; generic resumes do not.
- Scope signals. Amazon levels roles by scope: size of systems owned, size of teams influenced, ambiguity handled. Your bullets should convey scale ("a service handling X requests per day," "a process covering N sites"), not just tasks.
- Evidence of outcomes. Amazon's writing culture values specifics. Bullets that quantify impact read as credible; adjectives do not.
Concrete prep moves:
- Pick the two or three postings that genuinely match your experience instead of mass-applying. Referrals can put your application in front of a recruiter faster, but they do not bypass the interview bar.
- Rewrite your resume in Amazon's native style: data-heavy bullets, no graphics, no tables. Tools like CareerHelp can help you compare your resume against a posting's exact wording, but the substance must be yours and true.
- If you are a student or new graduate, note that Amazon runs dedicated early-career pipelines with their own assessment formats, described on the SDE university-roles prep page.
Stage 2: The recruiter screen
If your application advances, a recruiter calls — typically 15 to 30 minutes. This is a screening conversation, not a technical exam, but it is not a formality either.
What actually gets evaluated:
- Role fit and motivation. Why this team, why Amazon, why now. Vague answers ("it's a big company") are quietly damaging.
- Level calibration. The recruiter probes years of experience, scope, and compensation expectations to confirm you map to the posted level.
- Communication clarity. Can you summarize your background in two minutes without rambling? Amazon interviews reward structured speech.
Concrete prep moves:
- Prepare a 90-second career narrative: where you are, what you have owned, why this role.
- Research the specific team. Amazon's businesses — AWS, retail, advertising, devices, logistics — operate very differently, and saying something accurate about the one you applied to signals genuine interest.
- Be honest about compensation expectations and location constraints; surprises here stall processes late.
Stage 3: The online assessment (where applicable)
Most software development roles, many operations research and data roles, and nearly all new-grad engineering pipelines include an online assessment (OA) before any human interviews. According to Amazon's own prep guidance, the full-time SDE assessment can take up to two hours and may include coding challenges, a Workstyles assessment, a Work Simulation, and a survey; the company explicitly notes that no Amazon-specific knowledge is needed.
What actually gets evaluated:
- Coding challenges: correctness, edge-case handling, and clean code in a language you know well.
- Workstyles and Work Simulation: judgment in scenarios that mirror Amazon's environment — competing priorities, ambiguous requirements, quality-versus-speed tradeoffs. These sections map to the Leadership Principles more than candidates expect.
Concrete prep moves:
- Take the practice test Amazon provides; the interface familiarity alone reduces errors.
- Choose your strongest language and stick to public, unmodified resources — the guidance warns against copied code and screenshots.
- For the behavioral sections, answer consistently rather than strategically. The assessments are designed to detect contradictory responses; candidates who fake a persona often trip over themselves later when a live interviewer probes the same trait.
Stage 4: The phone screen
Next comes a 45- to 60-minute phone or video screen with a hiring manager or senior team member. For engineers this is usually one or two coding problems with live thinking-aloud; for product, program, and business roles it is a structured behavioral conversation, often with a short case discussion.
What actually gets evaluated:
- Technical baseline. Can you solve a realistic problem at the target level, communicate your reasoning, and recover gracefully when nudged?
- Behavioral substance. This is where the Leadership Principles first surface in live form. Amazon currently lists 16 of them, including Customer Obsession, Ownership, Dive Deep, Have Backbone; Disagree and Commit, and the two newer ones — Strive to be Earth's Best Employer and Success and Scale Bring Broad Responsibility. Interviewers typically carry a question list mapped to specific principles.
Amazon's official interview guidance recommends the STAR format — Situation, Task, Action, Result — for behavioral answers, and it remains the house style. But the format is the floor, not the ceiling. Interviewers probe until they reach what you personally did. "We decided" answers get interrupted with "and what did you decide?"
Illustrative scenario: a candidate describes a failed launch in two vague sentences, then pivots to a success story. The interviewer asks three levels deeper — who raised the risk, what data was available, what would you change — and the candidate cannot go deeper because the story was a team summary, not a personal account. That is the failure mode this stage is built to catch.
Concrete prep moves:
- Build a story bank of eight to ten real episodes, each usable against two or three principles. Ownership, Customer Obsession, Deliver Results, and Dive Deep appear constantly; Earn Trust and Have Backbone surface in the harder follow-ups.
- For coding, practice speaking while you solve. Silent thinking reads as struggling, even when it is not.
- Prepare questions for your interviewer about the team's systems and on-call reality; interviewers notice when candidates ask nothing.
Stage 5: The loop
The loop is Amazon's defining interview structure: roughly four to six back-to-back interviews (in person or virtual), each interviewer assigned specific focus areas — for an SDE, typically two coding rounds, a system design round, and behavioral rounds spread across the panel. Every interviewer writes up detailed notes afterward.
Three loop dynamics are worth understanding:
1. Each interviewer owns a lane. Interviewers deconflict before the loop so the panel collectively covers technical skill, design, behavioral principles, and role specifics. If one interviewer goes deep on your background, another will probe principles. Expect the full range.
2. The Bar Raiser. One interviewer is a certified Bar Raiser, drawn from outside the hiring team. Their job is to ask: would this person raise the average effectiveness of the team they join? They are deliberately insulated from the hiring manager's urgency, and their read carries heavy weight in the debrief. Bar Raisers tend to dig hardest into ownership, judgment under ambiguity, and standards — the moments where candidates inflate their role are exactly where Bar Raisers probe.
3. Written evidence beats memory. Because decisions are made from written feedback, what can be documented wins. Precise, specific answers — numbers, constraints, tradeoffs — give interviewers material to defend your candidacy in writing.
What actually gets evaluated, by lane:
- Coding: correctness under time pressure, clean abstractions, test cases, response to hints.
- System design: requirements first, then high-level architecture, then deep dives on data model, API design, and scaling tradeoffs. Interviewers care about judgment more than diagrams.
- Behavioral: depth of personal contribution, quality of decision-making, and honest retrospection on failures.
Concrete prep moves:
- Do at least two full mock loops out loud. The stamina problem — staying sharp in hour five — is real and underrated.
- Prepare your two best failure stories. Almost every loop includes a "tell me about a time you failed" probe, and the winning answer shows genuine accountability plus a concrete fix that outlived the incident.
- For senior levels, expect scope questions: how you influenced beyond your team, handled disagreement with leadership, or made calls with incomplete data.
Stage 6: The debrief and the offer
After the loop, interviewers submit written feedback and the group meets — the debrief — with the hiring manager and Bar Raiser present. The discussion is evidence-based: interviewers cite what you said, not impressions. The Bar Raiser typically facilitates toward a hire/no-hire standard rather than a vote.
What candidates should know:
- Silence is normal. Delays of one to two weeks after a loop are common and usually mean nothing; the debrief is scheduled around five busy people.
- Outcomes are not binary. Candidates sometimes get steered to a different level or team rather than rejected outright, especially when feedback is strong but level-fit is off.
- Offers come from a recruiter, with base salary, sign-on, and equity discussed separately from the loop. Recruiters can and do share the team's enthusiasm level; asking for it is fair.
- A no-hire is not permanent. Amazon candidates can reapply, and earlier feedback is sometimes reviewed, so a genuine change in readiness — new scope, new skills — is worth communicating on a second attempt.
The thread running through every stage
Zoom out and the pipeline has one consistent logic: Amazon is hunting for evidence, not claims. Every stage is engineered to convert a candidate's self-description into verifiable specifics — a code artifact, a simulated decision, a drilled-down story, a written interview note. Candidates who prepare by accumulating specific, truthful, deeply remembered episodes of their own work find each stage far less hostile than candidates who prepare by memorizing model answers.
The Leadership Principles are the shared vocabulary across all of it. Learn them, map your real experience to them, and let the specifics do the persuading.