Working Backwards

The problem it starts from

Every growing company meets the same difficulty. The founders’ instincts made the early decisions good, and those instincts do not scale. Thousands of people cannot each sit in a room with the founder. What usually replaces judgement is bureaucracy: approvals, slides and metrics that look like rigour but let weak ideas through and stop strong ones.

Colin Bryar and Bill Carr spent long careers at Amazon. Bryar worked for two years as Jeff Bezos’s technical adviser, a role that meant shadowing him through his meetings; Carr led the company’s digital media businesses. Their book, published in 2021, describes the practices Amazon built to solve the scaling problem, and the reasoning behind each one. It is written from inside, by people who used the practices and helped design some of them, and it should be read with that in mind. It is the company’s mechanisms as their users understood them.

The practices

The authors make a distinction that carries the whole book. Good intentions do not work; mechanisms do. Telling people to care about customers changes little. A process that forces them to describe the customer’s experience before they build anything changes a great deal.

Write the press release first. The book’s title practice is the PR/FAQ. Before a team builds a product, it writes the announcement it would issue on launch day, in plain language, saying who the customer is, what problem the product solves and why it is better than what exists. Behind the release sits a set of frequently asked questions, both those customers would ask and those the company’s own leaders would ask about cost, feasibility and risk. Most PR/FAQs never become products, and that is the point: it is far cheaper to discover in a document that an idea is weak than after a year of engineering. The authors describe the method at work across the company, including in the digital media teams Carr led.

Replace slides with narratives. In 2004 Amazon’s senior team stopped using PowerPoint in its meetings. Instead, the person presenting writes a narrative memo of up to six pages, and meetings begin with everyone reading it in silence. Bezos later explained in a shareholder letter that writing in full sentences forces the writer to think through the connections between ideas that bullet points let them skip, and that the reading period ensures everyone has actually absorbed the argument before discussing it.

Hire with a bar raiser. Every interview loop includes a trained interviewer from outside the hiring team, who has a veto. The bar raiser’s job is to keep the standard of each new hire above the median of the people already in that role, protecting the company from the pressure that a team under strain feels to fill a vacancy quickly.

Give each initiative a single-threaded leader. Amazon’s early experiment with small “two-pizza teams” evolved into the principle that a major project should be led by one person whose only job is that project, with a team organised around it. Shared ownership, the authors argue, means that nobody is fully accountable and every dependency becomes a negotiation.

Manage the inputs. Output metrics such as revenue and profit tell you what happened but not what to do. Amazon’s weekly business reviews concentrate on controllable input metrics, such as selection, price and delivery speed, which drive the outputs and which teams can actually move. Finding the right input metric is itself iterative; the book describes how a measure of how many product pages existed gave way, after several revisions, to a measure of whether those products were actually in stock and ready to ship quickly.

The idea underneath

The book’s theme is that culture lives in practice rather than in statements. Amazon’s leadership principles are famous, but the authors argue that they mean nothing without the mechanisms that make them operational. A principle like “customer obsession” turns into behaviour only when a process requires a team to describe the customer in writing before getting resources.

Writing is at the centre of nearly every mechanism, and that is not an accident. Writing is slower than talking, and its slowness is the value. A fluent speaker can carry a room on confidence; a written argument exposes its gaps to every reader. The narrative memo and the PR/FAQ are, among other things, defences against the persuasive power of charisma, which psychologists have long found to be a poor guide to the quality of an idea.

There is also an idea about decision speed. Bezos’s 2016 shareholder letter, which the book draws on, distinguished one-way doors, decisions that are hard to reverse and deserve slow, careful thought, from two-way doors, which can be reversed and should be made quickly by individuals or small groups. Most decisions are two-way doors, and companies that treat them all as one-way become slow. The site’s idea page Irreversibility and the Value of Waiting follows that distinction into economics and philosophy.

Since it was published

Since the book appeared, the PR/FAQ and narrative memo have been widely copied, and many companies have discovered that the documents are the easy part. They work at Amazon because leaders read them closely, argue with them and say no. The book’s subject, the company itself, has also faced intense scrutiny for its warehouse working conditions and for competition questions about its marketplace, which the site’s essay on online marketplaces, The Bazaar Without Walls, examines. Those debates are about what the company optimised for, and the mechanisms described here are neutral tools that serve whatever aim is set.

A first step

Before your next project starts, write a one-page press release for its launch day. Name the customer, the problem and why your answer is better than the alternatives. Then write the five questions a sceptical executive would ask, and answer them honestly. If the release is hard to write, or the answers are weak, you have learned something important for the price of an afternoon.