Key takeaways
Amazon runs a recruiter screen, sometimes an online assessment, a phone screen that splits coding with Leadership Principles questions, and an onsite loop of four to five interviews. A Bar Raiser from outside the team sits in that loop and runs the written debrief afterward. The decision weighs evidence principle by principle against a bar set at the 50th percentile of current employees in similar roles.
Amazon doesn't publish its internal debrief rubric or scoring scale. The mechanics of how interviewers vote and how the Bar Raiser runs the room come from reputable secondary accounts, including former Bar Raisers and the Working Backwards framework from two longtime Amazon executives.
How does Amazon decide who to hire?
Amazon decides hires in a written debrief that weighs evidence against the Leadership Principles, with a "raise the bar" standard and a Bar Raiser empowered to block. Amazon's own write-up of the Bar Raiser program defines the standard concretely: a new hire should bring "skills and abilities that are better than 50% of their would-be peers in similar roles" and should have growth potential. The bar is a line at the 50th percentile of people already doing the job at that level, and every interview asks whether you clear it.
Two structures enforce that line.
The first is the Bar Raiser, a trained, tenured Amazon employee pulled into your loop from outside the hiring team. They aren't the hiring manager and they don't work on the role you're interviewing for. Amazon describes them as "objective third-party advisers during the interview process." The program runs at scale. Amazon's same write-up counts more than 3,600 Bar Raisers, who are nominated, trained on the Leadership Principles, and shadow experienced Bar Raisers for anywhere from three months to a year before running loops on their own. The Bar Raiser interviews you like everyone else, but their real job comes after, when they run the debrief and drive the decision.
The second is the Leadership Principles scorecard. Amazon has 16 Leadership Principles, from Customer Obsession and Ownership to Earn Trust and Dive Deep. Every interviewer is assigned a subset to probe. They don't ask "are you customer-obsessed." They ask for a specific situation from your past, dig into what you personally did, and write up the evidence against the bar for the principles they own. The debrief assembles those write-ups into a picture of you, principle by principle.
Put those together and the shape of the decision is clear. You aren't scored on whether the room liked you. You're scored on whether the written evidence, read principle by principle, shows someone above the 50th-percentile line, with a Bar Raiser empowered to say no if it doesn't.
What does the Amazon recruiter screen cover?
The recruiter screen is a 30 to 45 minute call that confirms level fit and maps out your loop, with no coding or behavioral grilling. The recruiter forms an early read on the seniority you're a candidate for, and that read shapes which interviewers you draw and what bar they hold you to. Amazon levels are consequential and somewhat sticky, so a vague account of your scope here can set you up to be measured against the wrong line.
This is also where you learn the structure of your specific loop. Amazon says plainly that "our application and interview process differs from role to role," so the number of rounds, whether there's an online assessment, and which exercises you'll face all vary. Ask the recruiter directly. A good one will tell you who's in your loop and what each round covers, the single most useful prep input you can get.
What is the Amazon phone screen like?
The Amazon phone screen for technical roles is rarely pure coding; a typical 60-minute screen splits its time between a technical exercise and Leadership Principles questions. Often that means a coding problem in a shared editor followed by 15 or 20 minutes of "tell me about a time."
That split trips up candidates who prepped like it was a generic FAANG screen. They burn the clock optimizing a solution, then improvise a thin answer when the interviewer pivots to "tell me about a time you disagreed with your manager." Both halves are scored and both go into the packet.
What it screens for. Whether you can produce a correct, reasonably efficient solution while narrating your reasoning, and whether you have real, specific behavioral examples ready. The behavioral half is the first data point on the principles, and a weak one this early colors how the rest of the loop reads you.
For PM and non-engineering roles, the technical half becomes a role-specific exercise (a product sense question, an analytical case, a writing prompt), but the structure holds. Expect Leadership Principles woven into every screen regardless of function.
What happens in the Amazon onsite loop?
The Amazon onsite loop is usually four to five interviews of 45 to 60 minutes each, with Leadership Principles threaded through every round and two seats that matter more than the rest. For a software engineer, expect a couple of coding rounds, a system design round at mid-level and above, and behavioral coverage everywhere. For a PM, expect product design, analytical or execution rounds, and behavioral depth.
The two seats to know:
| Seat | Who they are | What they weight |
|---|---|---|
| Hiring manager | Owns the open headcount, manages you if hired | The technical bar and whether you'd be effective on their specific team |
| Bar Raiser | From outside the team, often more senior, identity not disclosed | Leadership Principles and the depth of your evidence, not the surface answer |
A Bar Raiser interview is usually the one where the follow-up questions don't stop, because that's the point. They're testing whether your story holds up under "Dive Deep" pressure or falls apart into vagueness.
What the loop screens for. Each interviewer owns a slice. Amazon deliberately assigns different principles to different interviewers so the loop covers the full scorecard without every round asking the same thing, and interviewers write up their own assessment before comparing notes. The consequence is that you can't coast on one strong round. A weak Earn Trust answer in round two and another in round four don't average out. They compound into a documented concern the Bar Raiser surfaces in the debrief.
How are Amazon's Leadership Principles scored?
The Leadership Principles are the dimensions you get scored on, and the scoring runs on evidence, not familiarity. Each principle has, in effect, a bar. The interviewer assigned to "Ownership" isn't checking whether you can define ownership. They're gathering a real story from your past and asking whether what you personally did clears the Ownership bar for the level you're targeting. Amazon's own interview-prep guidance points candidates to the STAR method (Situation, Task, Action, Result) because that structure surfaces the evidence an interviewer needs to write a real assessment.
Four things follow, and they're where strong candidates lose points.
- Vague stories score as no evidence. "We shipped a big project and it went well" gives the interviewer nothing to write down. The Action and Result have to be specific, and the Action has to be yours. Interviewers are trained to push past "we" to "what did you do."
- The follow-ups are the test. A Bar Raiser keeps digging ("why did you make that call, what did the data say, what would you do differently") because the depth of your answer is the signal. A story that collapses after two follow-ups reads as borrowed.
- Quantify the result. A number you can defend ("cut p99 latency 40%," "grew the segment from 2% to 9%") is far stronger evidence than an adjective. If you can't attach a real metric, pick a different story.
- Some principles carry quiet weight. Earn Trust and Ownership recur across roles, and a clear negative signal on either (an answer that reads as blaming others or dodging accountability) can sink an otherwise strong loop. They function as lines you can't fall below, not points to win.
The takeaway isn't "memorize 16 principles." Prepare 8 to 12 specific, quantified stories from your real experience, each tagged to the principles it demonstrates, and rehearse them until they survive aggressive follow-up. Recognition of the principles is the floor. Defensible evidence clears the bar.
How does the Amazon debrief work?
The debrief is a meeting where the loop's interviewers, each having submitted a written assessment and an initial vote, decide your offer together, and the Bar Raiser runs it. Reported accounts from former Bar Raisers put it around 30 minutes with roughly five interviewers. The Bar Raiser, not the hiring manager, facilitates, because the person driving the decision has no stake in filling the role. In one former Bar Raiser's account, the Bar Raiser walks the room through the written feedback, focuses debate on the principles where interviewers disagree, and pushes everyone back to the evidence rather than the vibe. People can re-vote after hearing the full picture.
"Raising the bar" here has a precise meaning. Not "did we like this person" or "is this person fine," but whether the written evidence, read principle by principle, shows someone better than half the people already doing the job at that level. A candidate the team is excited about can still come up short if the evidence is thin, and that gap between enthusiasm and evidence is what the Bar Raiser is there to catch.
On authority, the documented record and reported practice diverge slightly. Amazon's public materials describe the Bar Raiser as driving the group decision, without the word "veto." Multiple first-hand accounts from people who held the role describe explicit veto power, with one former Bar Raiser writing that "a hire cannot happen without your thumbs up." Those same accounts note that skilled Bar Raisers rarely invoke the veto outright, steering the room to a shared decision instead. Either way, the practical reality is identical. There's a person in that debrief, outside the hiring team, empowered to stop a hire the rest of the room wants. Your evidence has to satisfy them, not just the manager eager to fill the seat.
What gets strong technical candidates rejected at Amazon?
Strong technical candidates get rejected at Amazon for thin behavioral evidence far more often than for code, and the failures cluster in four predictable places.
- Strong code, thin Leadership Principle evidence. The classic Amazon no-hire. Every problem solved, but vague unquantified behavioral answers leave the debrief nothing to write under half the scorecard. Empty rows read as risk.
- A specific negative signal. One answer that reads as blaming a teammate (a failed Earn Trust signal) or never owning an outcome (a failed Ownership signal) can override otherwise strong rounds. These principles act as floors.
- Stories that don't survive Dive Deep. A polished narrative that falls apart under follow-up suggests the example wasn't yours, and interviewers probe exactly there.
- "We" instead of "I." Candidates who can't separate their contribution from their team's give the interviewer nothing to score, because the principles are about individual behavior.
The thread through all four: the technical bar is necessary but not sufficient. The decision turns on documented, specific, defensible evidence against the Leadership Principles, and candidates who treat the behavioral rounds as filler are the ones who get held.
How long does the Amazon interview process take?
Amazon doesn't publish a standard timeline, since the process "differs from role to role," so the useful thing to know is where loops stall. Three points. The first is between the recruiter screen and the loop, while scheduling comes together. The second, often the longest, is the debrief itself: it can't happen until every interviewer submits written feedback, and because the Bar Raiser is shared across many loops company-wide, getting them in a room is a frequent bottleneck. The third is after a hire decision, during reference and background checks. If you go quiet for a week or two after a strong onsite, it usually reflects this scheduling reality, not a verdict on you.
What to prep in the last week
Weight your prep toward the behavioral side, where the risk actually sits at Amazon. Four moves.
- Build your story bank. Write 8 to 12 real STAR stories, each tagged to the Leadership Principles it demonstrates. Cover Ownership, Earn Trust, Dive Deep, Bias for Action, and Deliver Results, because they recur across loops. Attach a defensible number to every result.
- Rehearse the follow-ups, not just the stories. The Bar Raiser's depth questions are where stories break. For each one, pre-answer "why did you do it that way," "what did the data say," and "what would you change." Reading your stories builds recognition, not the fluency to defend a decision while someone digs.
- Practice the split. In a technical role, rehearse pivoting from a coding problem straight into a clean behavioral answer in the same hour, because the phone screen will ask you to.
- Drill the principle floors. Have at least one strong story for Earn Trust and one for Ownership that show you taking accountability, not deflecting it.
Knowing your Leadership Principles stories and delivering them under follow-up are two different skills, and the loop only tests the second. Four-Leaf's voice mock interviews are built for that one. You answer real Leadership Principles questions out loud, get scored on the depth of your evidence, and drill the follow-ups that make rehearsed stories fall apart. The delivery half is measured rather than judged, computed from your own recording. Your rate of um and uh, your words per minute, and how often you hedged. The hedging count earns its keep in a bar raiser round, where "we kind of decided" and "I guess it worked" are the phrases that quietly turn an ownership story into a bystander story, and where nobody hears themselves say them. Run a full behavioral mock free for three days with every feature included, or with a $5 one-time 5 Day Pass if you just have the one Amazon onsite coming up. If you would rather survey the field first, our roundup of the best AI interview prep tools runs to ten of them, free through $150 a month.
The one thing to remember
Amazon's loop looks like a coding gauntlet and isn't one. The coding bar is real, but the decision is made in a debrief run by someone from outside the team, against a 16-principle scorecard, with a standard that asks whether your written evidence puts you above the 50th-percentile line for the role.
Prepare like the behavioral rounds are the main event, because at Amazon they are. Build specific, quantified stories. Rehearse them until they survive the follow-ups. Cover the principle floors. The candidates who understand that the Bar Raiser is reading evidence, not vibes, are the ones who give that debrief an easy yes.
The closest comparison is the Microsoft interview process, which has a senior final interviewer often likened to the Bar Raiser but no published program, training, or debrief behind it. The Netflix interview process goes further still and documents no loop at all.
For how this compares against sixteen other employers, including which of them publish their process at all, see our guide to how tech companies run their interview processes.