Amazon Bar Raiser Interview: 4 Interviewers Voted Hire. You Got Rejected Anyway.

Four interviewers can vote hire and you can still be rejected. That single fact is why the Amazon Bar Raiser interview generates more anxiety than any other round in tech hiring, and why almost everything written about it repeats the same unchecked claim: that the Bar Raiser holds a veto.

Amazon's own published description of the program never uses the word veto. It says something narrower, and the difference changes how you should prepare.

This guide covers what the Bar Raiser round actually tests, how the decision gets made in a room you are not in, the questions Bar Raisers are trained to ask, and a two-week prep plan sized to a loop where you are not allowed to repeat yourself. It assumes you already know the 16 Leadership Principles. If you do not, start with our Amazon Leadership Principles guide and come back.

What the Amazon Bar Raiser actually is

A Bar Raiser is an Amazon employee from outside the hiring team, trained as an objective third party, who sits in your loop as one of the interviewers and then chairs the meeting where your hire decision is made. Amazon says there are more than 3,600 of them. Counting people still in training, AWS put the number above 10,000 when the program turned 25.

It is a volunteer role stacked on top of a full-time job. Training is nomination-only and takes three months to a year. Amazon's Bar Raiser program manager, Katie Anderson, describes graduation as variable rather than fixed: "May take fewer than a dozen interviews for one person to graduate and 40 or more for another."

The program started in 1999 under then-CIO Rick Dalzell, co-built by John Vlastelica, Amazon's first technical recruiting manager. Its original name was the barkeeper programme. It was renamed because keeping the bar was the wrong goal. The whole design traces back to a line in Jeff Bezos's 1998 shareholder letter: setting the bar high in hiring was, he wrote, the single most important element of Amazon's success.

Does the Bar Raiser have veto power?

Here is the claim you will find on every prep site, coaching page, and Blind thread: the Bar Raiser can veto your offer, so they are the only person you have to convince.

Amazon does not say that. Its description of the role lists three duties, and the decision-making one reads: "With hiring manager, makes final hiring decision." Shared authority. The word veto appears nowhere on Amazon's public pages about the program.

The veto language comes from ex-Amazon leaders, not from Amazon. Colin Bryar and Bill Carr, both former Amazon executives, describe Bar Raisers as holding veto power over hiring managers in their account of the process. And in the same breath they add the sentence that resolves the contradiction: a good Bar Raiser never explicitly exercises their veto power.

Both things are true because the mechanism is not an override button. It is a deadlock rule. The operating principle in the debrief is that if the hiring manager or the Bar Raiser is unsure, the right decision is not to hire. Two people hold the decision, and either one of them being unconvinced ends it. That reads as a veto from the outside, but it works differently from one, and the difference matters:

• A veto implies you must impress one person most. You must not. You must give five people evidence that survives being read out loud by someone else.
• A veto implies the Bar Raiser overrules the room. They do not. They aggregate the room, then decide with the hiring manager.
• A veto implies your Bar Raiser round is the round that counts. It is not. The written feedback from all five rounds is the input. Your Bar Raiser round is one fifth of it, plus the person reading the other four.

If you prepare for a veto, you optimize for charming one interviewer. If you prepare for a deadlock rule, you optimize for leaving quotable evidence with every interviewer. The second one is what gets offers.

What "raising the bar" actually means

Amazon publishes the standard, and it is more specific than the folklore suggests. Every person hired should be better than 50 percent of those currently in similar roles. Amazon's own phrasing is that a new hire brings skills and abilities better than 50 percent of their would-be peers in similar roles, with growth potential on top.

Read that carefully, because it reframes the whole round.

You are not being compared to the other candidates in the pipeline. You are being compared to the median person already doing that job at that level at Amazon. Beating the applicant pool is irrelevant. Beating the median incumbent is the entire test.

Three consequences follow, and senior candidates get all three wrong:

The bar moves with the level, not with the market. An L6 candidate is measured against the median L6, who has been operating at Amazon scope for years. Your outside-Amazon seniority does not transfer at face value.

"No weaknesses" is a losing profile. Being adequate everywhere puts you at the median by definition. The bar requires being above it, which means at least one dimension where you are clearly better than the person already in the seat.

Growth potential is scored, not assumed. The standard includes future trajectory. A candidate who is exactly at the bar today and flat is a no hire. A candidate slightly above it and climbing is a hire.

Why this round feels different: they are collecting data, not listening

The Bar Raiser round feels colder than the others. Candidates read it as hostility. It is not. It is training.

Carlos Arguelles was an Amazon Bar Raiser from 2009 to 2020 and conducted 813 interviews there. He failed his own Bar Raiser certification on the first attempt. The critique his trainers gave him explains the entire interviewing style: "Hey I noticed you spent 2 minutes on question X, exactly what concrete data did you get from these 2 minutes?"

That is the job. Bar Raisers are trained to convert minutes into data points, and they are graded on the conversion rate. Which means:

Being interrupted is normal, and usually good. They got what they needed and are buying time for another data point. Candidates read the interruption as failure and speed up, which degrades the next answer.

Long context is expensive to you. Ninety seconds of setup is ninety seconds that produced nothing they can write down. Situation and Task should take two sentences combined.

Vagueness reads as absence, not modesty. If they cannot extract a concrete fact, the honest thing for them to write is that no evidence was obtained. That is scored as a miss, not a neutral.

"We" answers produce no data. Every time you say we shipped it, the interviewer still has nothing attributable to you. Say I, and name the decision you personally made.

The debrief decides your offer, not your interview

This is the part almost nobody prepares for, and it is where offers are actually made.

After the loop, every interviewer writes up their feedback with a hire or no hire vote and supporting evidence, and publishes it before they can see anyone else's. That sequencing is deliberate: nobody's memory gets contaminated by the loudest person in the room.

Then the debrief runs, chaired by the Bar Raiser rather than the hiring manager:

1. Silent reading. Everyone reads every assessment, 10 to 15 minutes, no discussion.
2. Re-vote. Interviewers revise their votes now that they have seen the full picture. Votes move at this step, in both directions.
3. Discussion. The vote usually splits. The room debates which Leadership Principles were demonstrated and which were not, using the written evidence.
4. Decision. The Bar Raiser frames which principles clear the bar and which fall short, and decides with the hiring manager. Unsure means no.

Now look at what that process rewards. Your advocate in that room is not you. It is four other people reading from notes they wrote days ago. The candidates who get offers are the ones whose stories survive being retold badly by a tired stranger.

So the real preparation question is not "what will I say." It is "what will they write." For every story in your bank, write the one sentence you want to appear in that person's feedback. If you cannot write that sentence, the story is not ready. If the sentence has no number in it, the story is not ready.

The five things a Bar Raiser writes down about you

Bar Raisers do not score charisma. Across Amazon's stated criteria and the accounts of people who held the role, five things end up in the file:

1. Attributable action. What you personally did, separable from your team. This is the single most common gap in senior candidates' answers.
2. Evidence with numbers. Revenue, cost, latency, headcount, error rate, time. A result without a number is an opinion, and opinions do not survive the silent-reading step.
3. Judgment under incomplete information. Not whether you were right. How you decided, what you would have accepted as disconfirming evidence, and whether you went looking for it.
4. Behavior when the answer was expensive. Disagreement with a superior, a launch you delayed, a metric you traded away. Comfortable stories are unscoreable.
5. What changed afterward. The one-line "since then, I" closer. It feeds Learn and Be Curious for free and signals the growth potential the bar explicitly requires.

What Bar Raisers ask, by probe type

Question lists are the least useful thing published about this round, because Bar Raisers do not work from a script. They work from probes. Learn the six probes and any question is recognizable.

The boundary probe. Tell me about something you fixed that your team did not own. Tell me about taking on work outside your job description. They are testing whether Ownership means crossing a line or just doing your job well.

The disconfirmation probe. Who disagreed with you, and what exactly did they say? Tell me about a time you changed your mind after new evidence. Being right is not the score. Having gone looking for the counterargument is.

The counterfactual probe. What would you do differently? This is the third-level follow-up, and it is where thin stories collapse. A candidate who lived the story has a specific answer. A candidate who assembled it has a generic one.

The data probe. Where is the number? How did you measure that? How did you know it worked? Expect it on every story, and expect a follow-up on how the metric was defined.

The failure probe. Your biggest wrong call, and how you caught it. Amazon is not looking for a disguised strength. It is looking for whether the detection mechanism was yours.

The scope probe, at L6 and above. How many people, whose budget, who else was affected, who did you have to convince without authority. Senior loops weight this heavily, and it is the probe most often failed by strong operators from smaller companies.

Real loops confirm the pattern. A candidate hired at L5 in 2025 reported a 60-minute Bar Raiser round that was entirely Leadership Principles, testing consistency, depth, and range across examples. A 2026 account described the Bar Raiser opening with why Amazon, then working through Dive Deep, Learn and Be Curious, and Earn Trust. Different loops, same shape: no technical rescue, no small talk, all probes.

Three ways strong candidates fail this round

They reuse a story. This is the most preventable rejection in the Amazon loop. Any single interviewer would never notice. The Bar Raiser reads all five write-ups and sees the duplicate immediately, and it reads as a shallow career rather than a preparation slip. Candidates are routinely told this by their recruiter before the loop, and still do it, because their story bank ran dry in round four.

They answer in the plural. Senior candidates do this more than junior ones, because at their level the work genuinely was collective and claiming it feels dishonest. The interviewer does not read humility. They read an absent data point, and they write that down.

They are a solid C-plus everywhere. Dave Anderson, an ex-Amazon tech director and GM who served as a Bar Raiser, describes blocking a candidate a team was desperate to hire after weeks of searching. Nothing was wrong with him. That was the problem: "absolutely nothing impressed me. His energy levels were passable, but not exciting." Against a 50th-percentile bar, passable is the definition of the bar, not a clearance of it. This is the failure mode nobody prepares for, because it produces no visible mistake.

A two-week plan sized to a no-repeat loop

Start with the arithmetic, because it determines everything else. Five interviewers, 2 to 3 behavioral questions each, no repeats allowed: that is 10 to 15 distinct story slots. You need 8 to 10 real stories, each mapped to 2 or 3 Leadership Principles so it can flex to whichever probe arrives.

Week one, build the bank. Write 10 stories in bullet form, not prose. Each one gets: the decision that was yours, the number that proves the outcome, and the thing that changed afterward. Kill any story where you cannot name the number. Kill any story where the hardest moment was easy. Then map each to its principles and check the coverage grid for holes.

Week two, break them. Practice out loud, and only against follow-ups. First answers are assumed rehearsed; the round is decided at level three. For each story run the sequence: what did you do, why that way, what would you do differently. If level three produces a generic answer, the story is decorative and should be cut.

Two habits do most of the work. Compress Situation and Task to two sentences, and give Action 60 percent of your airtime in the first person. And after every practice answer, write the one sentence you want in the interviewer's file. That sentence is the actual deliverable of the round.

If you want the probes to come at you unscripted, run a mock Amazon behavioral interview and practice the follow-ups rather than the openers.

What changed for 2026

The reference population got stronger, so the bar went up. Amazon cut about 14,000 corporate roles in October 2025 and roughly 16,000 more in January 2026, and said reductions would continue while it keeps hiring in strategic areas. The bar is defined as the median of people currently in the role. Cut the bottom of a distribution and the median rises. Candidates benchmarking against what Amazon hired in 2022 are benchmarking against a population that no longer exists.

Using AI in the interview can disqualify you. Amazon requires candidates to acknowledge they will not use unauthorized tools including generative AI during interviews or assessments, and spokesperson Margaret Callahan confirmed the policy publicly in March 2025. Recruiters have internal guidance on spotting it, including unnatural reading patterns and typing during questions. Practice with AI as much as you like. Do not bring it into the room.

Behavioral rounds got heavier, not lighter. As technical screens become harder to trust, the round that is hardest to fake carries more weight. That is this one.

The one-line version

The Bar Raiser is not a gatekeeper you have to charm. They are an auditor with a stated benchmark: better than half the people already doing the job. They spend your hour converting your answers into written evidence, then chair a meeting where four other people's written evidence is read out loud and re-voted on. Prepare for the file, not the conversation.

If you are running other loops alongside this one, do not reuse the prep. Netflix has no rubric, no central bar, and weights system design above everything, which rewards the opposite behavior. We covered it in our Netflix senior software engineer interview guide.

Sources

• What is an Amazon bar raiser?, Amazon (3,600+ Bar Raisers; the 50 percent standard; "With hiring manager, makes final hiring decision"; Katie Anderson on training)
• What is a 'Bar Raiser' at Amazon?, Amazon EU (the 1999 barkeeper programme and the rename)
• Exploring Amazon's unique interview process, Amazon (Bar Raiser duties, written feedback)
• Amazon's Bar Raiser program, AWS Careers (Rick Dalzell, 1999 founding, 10,000+ Bar Raisers and trainees at 25 years)
• Bar Raiser Hiring, Colin Bryar and Bill Carr, former Amazon executives (debrief steps, silent reading, re-vote, veto language, "if the hiring manager or the bar raiser is unsure, then the right decision is not to hire")
• Memoirs of an Amazon Bar Raiser, Carlos Arguelles, Amazon Bar Raiser 2009 to 2020, 813 interviews
• Three Tales from an Amazon Bar Raiser, Dave Anderson, ex-Amazon tech director and GM
• Podcast: the origin story of Amazon's Bar Raiser program, John Vlastelica, Amazon's first technical recruiting manager
• Is it cheating? AI use during job interviews sparks debate, GeekWire, March 2025 (Amazon spokesperson Margaret Callahan on the GenAI policy)
• Amazon laying off about 14,000 corporate workers, CNBC, October 2025, and about 16,000 more, CNBC, January 2026