Amazon's Leadership Principles, and the Questions They Produce

All 16 principles

The current list, in Amazon's own order. Each one is a thing they will ask you to evidence from your own history.

1. Customer Obsession. Start from the customer and work backwards, even when it costs you internally.
2. Ownership. Act on behalf of the whole company, not your team. Never say that is not my job.
3. Invent and Simplify. Find new approaches, and cut steps out of existing ones.
4. Are Right, A Lot. Strong judgment, and a habit of seeking views that disconfirm your own.
5. Learn and Be Curious. Keep improving yourself and go looking at things nobody asked you to look at.
6. Hire and Develop the Best. Raise the bar with every hire and move people on deliberately.
7. Insist on the Highest Standards. Standards most people think are unreasonably high, held continuously.
8. Think Big. Small thinking is self-fulfilling. Set direction that others want to follow.
9. Bias for Action. Speed matters. Many decisions are reversible and do not need long study.
10. Frugality. Do more with less. Constraints produce invention.
11. Earn Trust. Listen properly, speak candidly, treat yourself and your team against the best.
12. Dive Deep. Stay connected to the detail, audit often, and distrust anecdote that conflicts with data.
13. Have Backbone; Disagree and Commit. Challenge decisions you disagree with, then commit fully once made.
14. Deliver Results. Focus on the key inputs, deliver on time, and rise to setbacks rather than settling.
15. Strive to be Earth's Best Employer. Build a safer, more productive, more just working environment.
16. Success and Scale Bring Broad Responsibility. Amazon's size has consequences, and they must be handled with humility.

How Amazon actually interviews on them

Three things make this loop different from a normal behavioral interview.

Your interviewer is taking near-verbatim notes. They are typing what you say while you say it, and afterwards they write up a structured account with evidence tagged against specific principles. Speaking in vague summary is expensive here in a way it is not elsewhere, because there is nothing to write down. A sentence like we improved the process significantly produces an empty line in somebody's notes, and empty lines are what the debrief argues over. Give them numbers, names and dates, and let them type.

There is a bar raiser in the loop. One interviewer, from outside the hiring team, whose job is to hold the standard against Amazon's overall bar rather than against the team's immediate need to fill a role. They have veto power. They are usually the person who asks the fourth follow-up.

You will be interrupted. Most interviewers let you finish. Amazon interviewers cut in to redirect you toward the detail they need for their write-up, which candidates often read as hostility. It is not. It means they have decided the part you are on does not contain evidence and they are steering you toward the part that does.

There is a fourth thing worth knowing, which is that the loop is designed to be inconsistent on purpose. Different interviewers are assigned different principles, they do not compare notes until the debrief, and each writes an independent assessment. That means a story that landed well in your second interview may get a much harder time in your fourth, because the person asking has been told to probe a different competency with it. Candidates read this as one bad interviewer. It is the system working as designed.

The practical consequence is that a rehearsed two-minute answer rarely survives intact, and you should not try to make it. Bring the structure, expect to be steered, and have the specifics ready.

The question each principle produces

You will rarely be told which principle is being tested, and asking is not usually productive. The mapping below is the one most interviewers work from. Read it in both directions: down the left when you are preparing, and up from the right when you are in the room trying to work out what is being assessed.

If you can identify the principle from the question, you know which details to lead with. A question about taking on work outside your remit is Ownership, so the detail that matters is the boundary. A question about a decision made with limited data is Are Right, A Lot, so the detail that matters is your confidence at the time and what you did to close the gap.

Six worked answers

The six that come up most across Amazon loops, answered in STAR format. Each is ninety seconds out loud.

Customer Obsession

What they are listening for is a cost you absorbed internally. Doing right by a customer when it was also cheap and popular evidences nothing.

We had a billing bug overcharging about four hundred small business accounts by an average of eleven pounds a month, running for five months. My task was deciding what we refunded. Finance wanted to refund on request, which would have cost us roughly nine thousand pounds because most people never check. I pulled the full affected list, refunded all four hundred proactively, and emailed each one explaining what had happened before they found out from us reactively. That cost about twenty six thousand pounds. Forty of them replied to say thank you, and our churn in that cohort over the following two quarters ran below the book average.

Ownership

They want something that was demonstrably not yours, and they want the boundary between what you did and what the owning team did drawn before they ask.

Our checkout error rate climbed from under one percent to about four percent over a fortnight. It belonged to the payments team and I was on growth, and they were mid-migration and not looking at it. Rather than escalate and wait, I reproduced it myself, found it was a specific card issuer failing on a retry path, and wrote up a reproduction with logs. I took it to their lead with the fix already scoped. It shipped in three days. I do not own payments and the four percent was coming out of my funnel, so it was mine to chase.

Invent and Simplify

Simplify scores as well as invent, and it is easier to evidence. Removing steps from something that already works counts fully.

Our merchant onboarding took eleven steps and around nine days, and drop-off was concentrated in steps four to six, which collected tax details we did not need until the first payout. I moved those three steps behind the first successful transaction. Onboarding went to eight steps and about two days, and completion rose from fifty four percent to seventy one. The invention was noticing that a step being required eventually is not the same as being required now.

Dive Deep

The test is whether you go past the dashboard to the thing underneath it, and whether you distrust a number that looks too well behaved.

A weekly revenue report had been flat for six weeks and everyone had accepted it as a plateau. I did not believe a number could be that stable, so I went into the query rather than the dashboard. A join was silently dropping any order with more than one shipment, which was about nine percent of volume and growing. The plateau was an artefact. Actual revenue had grown eleven percent over those six weeks and we had been making roadmap decisions against a broken number for a month and a half.

Have Backbone; Disagree and Commit

Both halves are being scored. Disagreeing is the half candidates prepare. Committing after losing is the half that decides the answer.

My director wanted to launch a referral programme before we had instrumented attribution. I disagreed twice, once in writing, because I thought we would not be able to tell whether it worked and would end up defending it on anecdote. He decided to launch anyway. I built it, launched on his date, and did not relitigate it in the team. I was partly wrong: it drove measurable signups. I was partly right: nine months later we still could not say what it cost per acquisition, and that argument came back at budget time.

Deliver Results

They want a commitment you held when the conditions changed underneath it, and an honest account of what the holding cost.

Six weeks before a compliance filing deadline, one of the two people who did it left. I had not done a filing before. I cut my own scope to nothing else, worked through the previous submission line by line, and brought the new starter onto reconciliation from her second day. We filed two days early with nothing outstanding. What it cost was her onboarding: she learned that one process and almost none of the rest of her role for two months, and I made that trade without discussing it with her.

Where these answers break

Amazon interviewers dig harder than most, and each principle has a characteristic place where a prepared answer gives out.

Have Backbone is the one that collapses most often. Candidates tell the half where they disagreed, because it sounds like courage, and stop before the half where they lost. The follow-up is: what happened after you disagreed and lost. If the answer is that you were eventually proved right, you have described being right rather than committing, and the principle has two halves for a reason.

The worked answer above passes because it contains a loss the candidate did not recover from. He disagreed twice, lost, built the thing on the original date, and did not carry the argument back into the team. He also concedes he was partly wrong on the substance. That combination is hard to fake and it is exactly what the principle describes.

Ownership breaks on the plural pronoun. This is the single most common failure across the whole loop. The story is about doing something outside your remit, which by definition involved other teams, so the Action drifts plural. Then: which part did you personally do, and what did the payments team do. If you cannot separate it cleanly you have described noticing a problem, which is not ownership.

Customer Obsession breaks on cost. Anyone can describe doing right by a customer. The follow-up asks what it cost internally and who objected to paying it. If nothing was traded and nobody pushed back, the decision was easy, and easy decisions do not evidence a principle that exists precisely to describe hard ones. Name the person who disagreed and what their argument was, because their argument was usually reasonable.

Dive Deep breaks on what you did next. Finding a broken join is good. The follow-up is what changed so it would be caught next time, and if the answer is that you fixed the query, you have described debugging rather than a systemic fix.

Bias for Action breaks on the reversibility question. Amazon distinguishes decisions that can be undone from decisions that cannot, and moving fast on the second kind is a failure rather than a strength. Expect to be asked whether the decision you made quickly was reversible, and to be asked what you would have done had it not been. An answer that treats speed as good in all cases lands badly with people who think about this distinction constantly.

Frugality breaks on whether the constraint was real. Delivering under budget is only interesting if the budget was genuinely tight. The follow-up asks what you would have done with more, and if the honest answer is the same thing, the constraint was never binding.

These four account for most of the rejections we see described in Amazon loops. For a fuller treatment of the disagreement question, see disagreeing with your manager. For decisions where somebody paid a price, see a difficult decision you made. For team decisions people disagreed with, see showing leadership.

Preparing the loop

You need roughly six to eight stories, not sixteen. Most stories evidence three or four principles depending on which details you lead with, and Amazon expects reuse across interviewers as long as the substance holds.

Build each story with the numbers attached and the boundary between your work and your team's already drawn. Then, for each principle, know which of your stories you would reach for and which detail you would lead with. That is far more useful preparation than writing sixteen separate answers, most of which you will never use.

One practical note on coverage. Look at your six stories against the sixteen principles and find the gaps. Most candidates discover they have nothing for Frugality, nothing for Hire and Develop the Best, and nothing for the two newest principles about employment and scale. Those gaps are worth filling with a real example even if it is a small one, because a thin answer to a principle you were unprepared for reads considerably worse than a modest answer you had ready.

For the wider set of behavioral questions outside Amazon's format, see behavioral interview questions.