Amazon's operating principles: how AWS builds at scale with frugality and rigour

Executive overview

Most companies slow down as they grow because decision-making centralises and consensus replaces clarity. Amazon runs on 14 leadership principles — an operating system that lets people close to the customer make high-velocity, high-quality decisions without escalating.

Frugality is not cost-cutting. Constrained resources force creative problem-solving. Banning PowerPoint and replacing it with six-page narratives removes communication bias and surfaces better decisions.

The core insight: structure and constraint are not bureaucracy — they are the forcing functions that produce speed and invention.

The 14 leadership principles

• Act as an operating system: consistent, decentralised decision-making at every level.
• Bookended by two: customer obsession (start with the customer, work backwards) and deliver results.
• Every person inside Amazon is considered a leader — not just people managers.
• Principles are used in recruitment, meetings, investment decisions, and performance.

Frugality as an innovation lever

• Amazon invests heavily (over $20B annually in R&D) but deliberately focuses spend on customer value.
• Active resource constraints force teams to find non-linear, creative solutions.
• When teams ask for investment, they are routinely asked: how would you do this with half the budget?
• The constraint is culturally accepted — it is not a frustration but a forcing function.

Two-pizza teams and scaling horizontally

• A two-pizza team is a small (10-12 person) cross-functional unit that can operate fully autonomously.
• Small teams avoid the coordination overhead that kills speed at scale.
• Teams scale horizontally — spin up new two-pizza teams rather than growing existing ones.
• Amazon accepts that two teams may work on the same thing; doing nothing is worse than some duplication.

One-way doors vs two-way doors

• Two-way door decisions are reversible — move fast, use judgment, don't wait for perfect information.
• One-way door decisions are irreversible — dive deep, use rigorous data, senior leaders get closely involved.
• The distinction is not always obvious, which is why good judgment is a core expectation.
• Bias for action and deep dive are deliberately in tension; that tension is intentional.

Disagree and commit

• Consensus produces grey answers when the answer is clearly white or black.
• Robust, data-driven debate is expected — having an opinion and not sharing it counts against you.
• Once a decision is made, everyone commits fully and stops second-guessing.
• The senior-most person in a meeting speaks last, so they don't inadvertently anchor the room.

Six-page narratives instead of PowerPoint

• All internal meetings are governed by a written, long-form, six-page narrative — no bullet points, no slides.
• The first 15-20 minutes of every meeting are spent reading the document in silence.
• Writing forces rigour: you cannot make vague assertions — every claim must be substantiated with data.
• Presentational skill can hide weak thinking in PowerPoint; narrative removes that cover.
• Fixed format rules: 10-point minimum font, strict margin and line-spacing guidelines.
• A six-pager typically takes 3-4 hours for a routine doc; 50-60 hours for an annual business review.
• Appendices are allowed but not expected to be read — the six pages must stand alone.

Working backwards: the PRFAQ process

• Everything starts with the customer. Five questions anchor every initiative: who is the customer, what is their problem, what is the key benefit, how do you know what they want, what does their experience look like?
• A PRFAQ begins with a fictitious press release written as if the product already exists and succeeded.
• Fictitious customer quotes force the team to test whether the value proposition is real.
• FAQs — both external and internal — are used to surface the hardest questions before build begins.
• PRFAQs iterate; they become a reference point throughout the product lifecycle.
• There is no investment committee — the goal is to find a path to yes, not to gatekeep ideas.

Hiring: the bar raiser mechanism

• Every hire must be better than 50% of people already in that role — an empirical bar, not a vibe.
• Interviewers in the structured loop process are each assigned one or two leadership principles to probe.
• A bar raiser — a certified, role-model Amazonian with no stake in the outcome — holds a veto in the debrief.
• The bar raiser's job is to protect culture, not fill headcount.
• The process self-selects: candidates who don't fit the culture often opt out during the process itself.

Managing inputs, not outputs

• Leaders focus on inputs (leading indicators) rather than outputs (revenue, P&L).
• Inputs are measured rigorously and causally linked to business outcomes.
• The goal is to give teams maximum freedom to move fast in front of customers.
• Coaching replaces management; the principles replace constant oversight.

The organisational yes

• Default orientation for leaders: find a way to yes, not a reason to say no.
• A no is easy and makes you sound smart; a yes requires real work.
• Ideas are iterated toward viability rather than killed at the gate.