What is an Amazon Bar Raiser?

A Bar Raiser is an Amazon employee, trained and certified for the role, who sits on interview loops outside their own team. Amazon describes the programme publicly. Because they do not report to the hiring manager and gain nothing from the seat being filled, their job is to judge whether you would raise the average of the people already at that level.

Does the Bar Raiser have veto power over the hire?

Yes. The Bar Raiser can block an offer even when the hiring manager and the rest of the loop want to hire. That is the whole point of putting an outsider in the room. In practice they also run the debrief, so they set the order in which evidence is read out and decide when the group has enough to call it.

How do I know which interviewer is the Bar Raiser?

Often you cannot, and you should not plan around guessing. The usual tells are an interviewer who does not work on the team you would join, a round that is almost entirely behavioural, and heavy note-taking with long typing pauses. Some recruiters name the Bar Raiser in the schedule and some do not. Treat every behavioural round as if it is the one.

Why do Amazon interviewers keep asking what my specific contribution was?

Because the note they write has to be attributable to you. Amazon interviewers capture near-verbatim detail and tag it to the Leadership Principles they were assigned to probe. A line that says the team migrated a service is evidence about a team. The follow-up is not scepticism, it is the interviewer trying to salvage a scoreable note from an unscoreable answer.

Should I say I or we in an Amazon interview?

Say we once to set the scene, then switch to I for every action. Use first-person verbs: I wrote, I escalated, I rolled back, I said no. This fails harder at Amazon than elsewhere because the debrief reads written notes rather than impressions, and an ambiguous note resolves against the candidate when a Bar Raiser is the one reading it.

How many stories should I prepare for a Bar Raiser round?

Twelve to fourteen, tagged by Leadership Principle rather than by theme, with at least two that can be re-aimed at Ownership and Dive Deep because those come up most. Each story needs one number you can defend two follow-ups deep, plus a prepared answer to what you would do differently. Prompts are allocated across the loop, so repetition across interviewers is expected.

Amazon Bar Raiser Round: What That Interviewer Is Actually Doing

One person on your Amazon loop does not work on the team you would join, was trained and certified separately for this job, and can block the offer on their own even when everyone else says hire. That is the Bar Raiser. Their round is behavioural almost end to end, every question is aimed at a specific Leadership Principle they were assigned to probe, and they will drill each answer three or four levels down until they have a note they can attribute to you personally. Which is why answers built out of the word "we" fail in this round harder than anywhere else in the loop.

Who is sitting opposite you, and why they behave differently

Amazon runs the Bar Raiser programme openly, and the mechanics that matter to you as a candidate are these. Bar Raisers volunteer, train, shadow experienced Bar Raisers, and get certified before they run loops alone. They are deliberately placed outside the hiring org, so the team's headcount pressure is not their problem. And they hold a veto on the hire.

Everyone else in your loop has some reason to want the seat filled. The hiring manager has a roadmap slipping. The engineer interviewing you is on the rota because their team is short. The Bar Raiser has none of that, and the question they are answering is not "can this person do the job", it is "would this person be better than the median person we already have at this level". That comparison is why an interview can go pleasantly, feel like a good conversation, and still come back as no hire. Adequate is a no.

Two more mechanics change how you should behave in the room. Amazon interviewers take near-verbatim notes while you talk, so expect typing, expect the interviewer to look down, and do not read that as boredom. Written feedback is normally submitted before the debrief opens, which means the note taken during your sentence is what gets argued over later, not a remembered impression of you. The Bar Raiser usually chairs that debrief.

The decoder: question phrasing on the left, the principle being scored on the right

Interviewers on a loop are allocated Leadership Principles so the panel does not probe the same three all day. The Bar Raiser tends to pick up whatever nobody else was assigned, plus the couple that are hardest to fake. Nothing you are asked is small talk. Here is the mapping, and the specific thing that turns an answer into a scoreable note.

What they ask	Principle being probed	What earns the note
"Tell me about a time you had to decide without all the data."	Bias for Action	Name the cost of waiting and whether the decision was reversible. "I shipped it behind a flag because backing it out took an hour" scores. "I gathered more data first" answers the opposite question.
"Walk me through the hardest problem you have debugged. How did you find root cause?"	Dive Deep	Raw evidence you looked at yourself: the query you ran, the log line, the graph. Relaying someone else's summary is the standard way to fail this one.
"Tell me about a time you disagreed with your manager."	Have Backbone; Disagree and Commit	Both halves. You pushed, out loud, with a reason. Then you executed the decision that went against you without hedging or slow-walking it.
"Tell me about something you owned that was not your job."	Ownership	No stopping at the team boundary. The guaranteed follow-up is what happened after you handed it over, so know the ending.
"Tell me about a time you removed something instead of building something."	Invent and Simplify	A deleted service, a killed config flag, three steps taken out of a runbook. Adding a framework is not simplification.
"Tell me about a time you were wrong."	Earn Trust, and Are Right, A Lot	The behaviour you changed afterwards and who you told at the time. A failure that cost you nothing reads as a rehearsed non-answer.
"Tell me about delivering with fewer resources than you needed."	Frugality	The constraint used as a design input. Cheaper architecture scores. Longer hours do not.
"Tell me about a time you would not accept a teammate's work."	Insist on the Highest Standards	The exact standard, the conversation you actually had, and what holding it cost you in time or goodwill.
"Tell me about choosing the customer over an internal target."	Customer Obsession	A named customer harm, and a trade you made against your own team's interest. A satisfaction score is not a story.
"What is the most aggressive deadline you have hit?"	Deliver Results	The scope you traded away plus the number at the end. Effort with no outcome gets written down as a miss.
"Tell me about something you taught yourself for a project."	Learn and Be Curious	It should have been outside your job description and necessary to the result. Both halves, or it is just a course you took.
"Tell me about feedback you gave that someone did not want to hear."	Hire and Develop the Best	What that person was doing six months later. Your delivery is not the signal, their trajectory is.
"Tell me about a proposal that was rejected as too ambitious."	Think Big	The size of the bet, and what you did with the no. Quietly shrinking it is a fail.
"Tell me about work of yours that affected people outside your team."	Success and Scale Bring Broad Responsibility	Second-order effects you spotted yourself, before somebody else complained about them.

Read that table backwards as your prep exercise. For each principle: can you name a story, one number inside it, and the person who disagreed with you? A principle with no story attached is a round you are hoping does not come up.

The drill-down, and how to survive four levels of it

Bar Raisers do not collect answers, they excavate one. Budget roughly ten to twelve minutes per question, of which your story is the first ninety seconds. The rest is this.

"What was your specific contribution?"

The most predictable probe in the round, and it is not scepticism. They need a sentence they can write down with your name on it. Answer in verbs: I wrote the migration script, I paged the on-call, I made the call to roll back.

"What was the number?"

Latency, cost, error rate, hours, tickets, revenue. Bring one you can defend two questions past this one, because the next probe is how you measured it. If you cannot defend it, give a smaller true number rather than a larger unsafe one.

"Why that option and not the other one?"

Name the option you rejected and the reason, in one sentence. An answer with no rejected alternative describes a task, not a decision, and tasks do not evidence judgement.

"How did the other person see it?"

State their case fairly, in their words, before you say why you disagreed. Caricaturing the counterparty reads as weak Earn Trust and it gets written down that way.

"What would you do differently?"

Not a humility test. "Nothing" and "I would communicate more" both score zero. Name one decision you would reverse and the earlier signal you now know to look for. This probe closes most Bar Raiser questions, so prepare it as a separate twenty-second answer for every story you carry.

One more pattern worth knowing. If you answer in the general present tense, "I usually make sure the team is aligned", they will stop you and ask for one specific instance. Hypotheticals cannot be scored, so they get handed back. Open every answer with a date, a team and a project, and you never trigger it.

Why "we" fails here specifically

Plenty of interviewers let a collective answer slide. This one cannot, for a mechanical reason rather than a stylistic one.

The output of the round is written notes tagged to Leadership Principles, and the debrief argues from those notes. A note reading "the team migrated the service off the legacy queue" is evidence about a team. There is no candidate in it. So when you say "we", one of two things happens. Either the interviewer spends three of your twelve minutes trying to extract who did what, or they write the ambiguous version down and move on. Both are bad outcomes. The first burns the clock you needed for the second half of the story. The second leaves the Bar Raiser holding weak evidence, and the person holding weak evidence is the one human in that debrief with no incentive to round it upward.

Ambiguity resolves against the hire. That is the design of the role, not a mood the interviewer is in.

The fix is small and worth drilling out loud, because it is a speech habit rather than a knowledge gap. Say "we" once, for context, then hand off: "The team owned the migration. My part was the dual-write layer and the cutover plan." After that, first-person verbs only. If a rehearsed sentence starts with "we decided", rewrite it as "I proposed X, my tech lead pushed for Y, and we went with mine after I showed the p99 numbers." That version still credits the team, and it still gives the interviewer something to type.

Worth naming plainly: this hits some candidates much harder than others. If you come from a team where claiming individual credit is bad manners, or from a culture where "we" is simply the polite default, you will sound arrogant to yourself long before you sound clear to the interviewer. You are not bragging, you are supplying the unit of measurement the format runs on. Record yourself and count the pronouns, or use the company Q&A generator to build Amazon-shaped prompts and rehearse against them. The company question sets are useful for seeing the shape of a full loop before you sit one.

How to spot the Bar Raiser, and why it barely matters

Often you cannot, and it is not worth spending attention on. Some recruiters name the Bar Raiser in the schedule, some do not. The usual tells are an interviewer whose team is not the team you would join, a round with no coding or system design in it, and the heaviest note-taking of the day. Practice varies between orgs and levels, so treat every behavioural round in the loop as though it is the one that counts. Behaving differently for one interviewer is a strategy that fails silently when you guess wrong, and it costs you nothing to give the same first-person, numbered answers to all five.

Where preparation tools, including ours, stop helping

This is the round where live assistance is least useful, and it is worth being straight about that. OfferPilot AI drafts a speakable answer from the interviewer's audio in about a second, grounded in your resume, which genuinely helps when you blank on a definition or a system design trade-off. A Bar Raiser drill-down is a different animal. The interviewer is asking for the number from an incident only you attended, then watching your face while you retrieve it. No tool holds that number. Reading a drafted paragraph about your own life, four probes deep, to someone taking verbatim notes, sounds exactly like what it is.

The limits of the software matter too. The overlay is excluded from screen capture at the operating-system level, so it does not show up in Zoom, Teams, Meet or a recording, but that does nothing about a webcam watching your eyes, an interviewer across a table, or a second-device camera. Amazon loops are frequently onsite, and in a room nothing on your laptop helps. It runs on Windows 10 and 11 only, with no macOS, Linux, mobile or browser version. For this particular round the useful work happens days earlier: write twelve to fourteen stories, tag each to a principle, put one defensible number in each, and rehearse the "what would you do differently" answer separately from the story it belongs to.

The Bar Raiser is not there to trip you up. They are there because Amazon does not trust a team that needs a body to judge its own bar, and they will keep digging until they have a sentence with your name on it. Give them that sentence early and the round gets much shorter.

The Amazon Bar Raiser round: what that interviewer is doing