behavioral · hard
Amazon Leadership Principles interview: the complete PM guide
Tell me about a time you demonstrated [Leadership Principle].
25% of candidates who clear Amazon’s technical bar are cut on behavioral. Stories use “we,” carry no metric, and read too clean. Bar Raisers are trained to break polished answers: in 2026 they ask unscripted follow-ups specifically designed to collapse scripted STAR arcs. The candidate with 10 polished scripts fails. The candidate with 8 real stories with honest texture passes.
How the loop works
Five 55-minute interviews. Each interviewer is assigned 1-3 LPs and scores independently before debrief. The Bar Raiser comes from outside the hiring team and holds veto over a unanimous hire: one “no hire” blocks the offer regardless of every other vote. Amazon’s jobs site is explicit: “We won’t ask brain teasers. Instead, we’ll focus on the what and how of your experiences, as well as the why of your decisions.”
Which LPs dominate the PM loop
Not all 16 get equal airtime. The five most frequently tested for PM candidates: Customer Obsession (almost always tested twice), Ownership, Have Backbone; Disagree and Commit, Bias for Action, and Deliver Results. The two 2021 additions (Strive to be Earth’s Best Employer, Success and Scale Bring Broad Responsibility) rarely surface for L5s but almost always appear at L6+.
The underlying throughline: all 16 funnel back to Customer Obsession. LP stories without a customer or user dimension feel incomplete to experienced Bar Raisers.
L5 vs. L6 calibration
Same LP, different bar. For Have Backbone; Disagree and Commit: an L5 answer covers a peer roadmap disagreement settled by data. An L6 answer must show both a win and a loss: senior stakeholder, cross-functional stakes (design, engineering, and business simultaneously), and full commitment after the decision went against you. Showing only wins reads as inexperienced.
Build 5-7 stories, not 16
One story can cover 3-4 LPs if framed correctly. A story where you expanded scope outside your charter, shipped under uncertainty, and hit a measurable result covers Ownership, Bias for Action, and Deliver Results simultaneously. Tag each story with every LP it can serve. Stories with the most LP overlap include: a decision under incomplete data, a concrete metric, a moment of real conflict, and something you would do differently.
The 2026 angle
For PMs, Think Big and Invent and Simplify have shifted. “We could build X with AI” is no longer a bold claim. The ambitious insight now is viability: was this a problem worth solving at the cost of building it? Think Big stories should center on market sizing, willingness to pay, and defensibility. Feasibility is the baseline.
Success and Scale Bring Broad Responsibility is the LP most commonly skipped in prep. A PM-relevant answer names a harm vector you anticipated (agentic capability, personalization surface, data handling), what you built to catch it, and how you balanced that against velocity.
Strong vs. weak
strong
"I was PM for the seller invoicing tool. We needed to deprecate a legacy API endpoint. Engineering asked for three more months of usage data; our SLA window gave us six weeks. I owned the call. I pulled usage logs: 94% of traffic came from three internal teams, all confirmed migrated. The remaining 6% were external developers on no support contract. I deprecated on schedule, wrote the migration guide myself, and gave 30 days' notice. Two external teams complained; one escalated. I handled both directly, resolved them within a week. Zero P1 incidents post-deprecation. If I did it again I'd have set automated alerts on the tail 6% ninety days earlier so engineering had the confidence data before the crunch."
weak
"Our team needed to launch a feature and we didn't have all the metrics we wanted. I worked with the team to gather as much data as possible and we made the best decision we could. It went well and the feature launched successfully." Uses "we" throughout. No specifics on what data was missing, what the decision was, or what was at stake. "It went well" is not a result. A Bar Raiser will probe immediately and have nothing to work with.
Amazon’s own interviewer guidance: “It is better to share no example than a bar-lowering example.” The two most common debrief notes: “Most words should be I not we” and “It would have been better if they described more impact.” Never name the LP in your answer (“This really shows my customer obsession”). Show it, don’t label it.
For how to structure STAR stories without sounding scripted, see the STAR framework guide. For the full Amazon process including the written exercise and technical rounds, see the Amazon process page. For how interviewers detect AI-rehearsed answers, see how interviewers catch AI answers.