Engineering Ideas
Published in

Engineering Ideas

Summary of “Working Backwards”, a book about Amazon’s culture

Principles, causal models, the role of intuition in business, PR FAQ, people-centric approach to invention, critical rationalism

Working Backwards by Colin Bryar and Bill Carr is a book about how Amazon company functions. The authors describe concrete processes and tactics used at Amazon to support their leadership principles. The authors also recount many stories from Amazon to illustrate and convey these principles.

A large corporation can be thought of as an organism. Throughout the summary, I’ll draw analogies between what Amazon is doing and some functions of biological systems and features of evolution.

Capitalistic ethics

Important note: the book is written in the frame of a capitalistic ethic where the very existence and the growth of the company (or the larger economy) are considered overarching goals. I personally don’t endorse this view. Think about almost every section in the review as having a disclaimer: “this is a good practice as long as keeping the company alive and growing is your goal that supersedes any other goals in a given context”.

Active inference

Active inference is a concept that unifies maintenance of boundaries, keeping homeostasis, learning, achieving goals, and more things that organisms do. The following Amazon’s principles and methods fall in this category as well:

Customer obsession instead of competitor obsession

Focus on input metrics (i. e., the metrics of product or service quality) to ensure that teams work on activities that propel the business.

This is related to Elon Musk’s idea that companies should better focus on improving their product than anything else.

Customer obsession, or improving the product could be seen as learning customers’ preferences in order to predict their satisfaction, thus minimising the “predictive gap” between the customers, the products, and the company. Ideally, the product should become a “natural” extension of the organisms or the minds of customers. The criterion of “naturalness” is that customers think (learn, predict, and act, in the active inference framework) as if these products were their parts.

Grow a new org capability in response to a market need rather than “do what the company already knows how to do”

Building a new function in response to an environmental need is fundamentally predictive, keeping doing what an organism already doing will eventually lead to decay and death because the environment always changes.

This rule of thumb also epitomises one of the main principles of systems thinking: first, looking “up” the system hierarchy, i. e. looking for needs of the higher-level system, then looking “down” at the architecture necessary to support these needs, the architecture of the organisation in this case.

A related idea is the institutional “no”: the tendency of well-meaning people within a large org to say “no” to new ideas. It offers managers short-term stability even if leads to value destruction later on. This is an example of the principal — agent problem. At Amazon, only Bezos himself pushed for the launch of Amazon Prime. This required intuitive thinking (more on this below).

Dive deep

Leaders operate at all levels, stay connected to the details, audit frequently, and are skeptical when metrics and anecdote differ. No task is beneath them.

Will Larson recently wrote a post on this topic: “Inspection and the limits of trust”.

“Diving deep” is a form of quick, non-local information transfer between the layers of an organisation. Essentially all species with more than one level of cellular hierarchy (i. e., apart from single-cell organisms and most primitive cell colonies) have some forms or such information transfer, such as a nervous system.

Gregor Hohpe touches on the theme of quick information transmission between the company ranks in The Architect Elevator as well.

Write down the organisation’s principles and put them to use

In the early days of Amazon, Jeff Bezos often worked in pairs with almost everyone in the company. He was never shy of saying how he wanted things to get done. He was also solely responsible for hiring for a long time. This way he could shape the working (entrepreneurial, managerial, engineering) principles before the company grew.

Several years later, many tenured employees were surveyed and the leadership principles were written down. Then these principles were integrated into the hiring and the regular performance review processes so that employees were reminded about these principles often.

The ideas that all people who work at the company should participate in developing the work principles and the mission statement of the company, and that then these principles should be put to use also appear in Stephen Covey’s The 7 Habits of Highly Effective People:

One of the fundamental problems in organizations, including families, is that people are not committed to the determinations of other people for their lives. They simply don’t buy into them.

Many times as I work with organizations, I find people whose goals are totally different from the goals of the enterprise. I commonly find reward systems completely out of alignment with stated value systems.

When I begin work with companies that have already developed some kind .of mission statement, I ask them, “How many of the people here know that you have a mission statement? How many of you know what it contains? How many were involved in creating it? How many really buy into it and use it as your frame of reference in making decisions?”

Without involvement, there is no commitment.

Organisation’s principles should be aligned with the moral matrices of the employees

A moral matrix is a proportion of different moral foundations in the “moral taste” of a person or a group of people. I think that the working principles and the mission statement should be aligned with the prevailing moral sentiments among the people who work at the company. For example, if the company makes some product that alleviates human suffering in some way, then it makes the most sense to hire people who emphasise the Care/harm foundation and also make sure that people in the workplace also feel that they are cared about.

Make a causal model of the important aspects of your organisation’s functioning

Such aspects could be growth, customer acquisition, customer satisfaction (added value), operational excellence, innovation, etc.

The famous “Amazon Flywheel” which Jeff Bezos drew is such a causal diagram of Amazon’s growth:

It’s interesting to note that similarly to the leadership principles, it was made sure that managers are frequently reminded of this causal model: the picture was put on the front page of the Weekly Business Review deck.

Bryar and Carr recommend thinking about the implications of the causal diagram for the organisational structure, the culture, and the strategy of the organisation.

A causal diagram of a system can be considered one of the types of system views. A causal diagram can be less formal than a data flow diagram or a money flow diagram or even a “value” flow diagram, as one can see in the example of “Amazon’s flywheel”. But covering this informal, intuitive spectrum of thinking about systems might be as important as more formal thinking. More on this below.

People should have a single goal they are responsible for

So-called “single-threaded leadership” helps Amazon to achieve their goals more reliably. Note that with “single-threaded leadership”, multiple people may be responsible for a single goal: what’s prohibited is that a single person is responsible for multiple goals.

The side benefit of single-threaded leadership is reduced work-in-progress:

Unfortunately, you’ve likely noticed that when you start a bunch of projects, you actually finish fewer of them. Instead you jump from task to task, answering questions and so on, slowing down more as you have more open tasks.

Reprioritization is very expensive in scenarios where we’re heavily constrained on execution, but fairly inexpensive in scenarios where we’re already finishing a great deal of work.

Applications of intuitive thinking and deconcentration in management and entrepreneurship

I’ve already covered how intuitive thinking should be used in decision-making in a recent post: “Formal reasoning, intuition, and embodied cognition in decision making”.

Deconcentration of attention is used in a regular operation report meeting to notice early signs of problem onset (this deconcentrated attention might need to span several such meetings). Both metrics and anecdotes should be discussed during such meetings to gauge how things are going by using more modalities of thinking: formalised reasoning about metrics as well as more intuitive thinking about anecdotes.

Entrepreneurs should develop a sense for important, pivotal decisions, even if they don’t seem so at the moment, such as was the case at Amazon when they were considering the launch of Amazon Prime. Spotting such pivotal moments for the enterprise requires deconcentration and intuitive thinking.

Relatedly, entrepreneurs (or even enterprises as a whole, if we think about the intelligence of the whole enterprise) should develop a taste for new high-potential product ideas and future market trends. Amazon did this with AWS: they realised the huge potential of providing storage and computing services for developers much earlier than other market players. Developing such a taste is similar to developing a taste for interesting and important problems in science, both amount to intuitive thinking. Michael Nielsen writes about how to train such intuition:

What do you think are the characteristics of important science? What makes one area thrive, while another dies away? What sorts of unifying ideas are the most useful? What have been the most important developments in your field? Why are they important? What were the apparently promising ideas that didn’t pan out? Why didn’t they pan out? You need to be thinking constantly about these issues, both in concrete terms, and also in the abstract, developing both a general feeling for what is important (and what is not), and also some specific beliefs about what is important and what is not in your fields of interest. Richard Hamming describes setting aside time each week for “Great Thoughts”, time in which he would focus on and discuss with others only things that he believed were of the highest importance. Systematically setting aside time to think (and talk with colleagues) about where the important problems are is an excellent way of developing as a problem-creator.

Amazon’s PR FAQ document is essentially the project alphas schema

At Amazon, people write and iterate on a so-called PR FAQ document (a press release and two FAQs: an internal and an external one) of the future product ahead of starting to build that product.

This practice coincides with the practice of systems engineering and systems management of writing and maintaining the project alphas schema in the beginning and throughout the project’s life. Here’re how the questions from the PR FAQ template correspond to the project alphas:

  • Problem Paragraph — alpha “Opportunity”, sub-alpha “External stakeholders’ needs”. As noted above, in systems thinking, the first move is always to think about the higher-level system’s needs.
  • Solution Paragraph(s) — alpha “System Description”, sub-alpha “Architecture”, functional view of the system.
  • “Quotes & Getting Started: Add one quote from you or your company’s spokesperson and a second quote from a hypothetical customer in which they describe the benefit they are getting from using your new product.” — alpha “Stakeholders”.
  • “Q: What is the price?” — alpha “Opportunity”, sub-alpha “Economic viability”.
  • “Q: How do I get it?” — alpha “System”, sub-alpha “Delivery/Distribution”.
  • “Q: How does it work? (You are likely to have multiple versions of this question that cover different aspects of the customer experience.)” — alpha “System Description”, sub-alpha “Architecture”, various views of the system, e. g. constructive, spatial, and informational, in addition to functional. See types of system views.
  • “Q: How large is the estimated consumer demand for XXX? What is the TAM (total addressable market)?” — alpha “Opportunity”, sub-alpha “Economic viability”.
  • “Q: What happens if a customer encounters x? How does the product deal with use case x? (there are likely to be several such questions).” — alpha “System Description”, sub-alpha “Requirements”.
  • “Q: What are the challenging product engineering problems we will need to solve?” — alpha “Opportunity”.
  • “Q: What are the challenging product engineering problems we will need to solve?” — alpha “Opportunity”, sub-alpha “Innovation”. Best product pitches set stretch goals for price or functionality or other characteristic and discuss how to achieve them.

Project alphas schema is more comprehensive than the PR FAQ template so those who follow the former are less likely to forget to think about something important before starting to work on the product. The fact that in the systems thinking approach, the important objects of attention are structured also permits tracking the statuses of these objects, which the PR FAQ process doesn’t support. For example, each external stakeholder may pass through the following statuses: recognised, represented, involved, in agreement, satisfied for deployment, satisfied in use. The list of statuses is different for each alpha and sub-alpha and is tailored to the specifics of the particular project.

Goal discovery

Eagerness to invent is one of the cornerstone pieces of Amazon’s culture. Here’s how Amazon achieves a good invention/discovery rate:

People-centric approach to invention

When Amazon started digital media division and AWS division, the main questions that the higher-up executive team has decided on were whom to put in the leadership of this new team and how to place this new division within the existing organisation structure (which also implies what dependencies and relationships the new division will have with the existing teams), not what they should do and how they should do it. Then the leaders of the new team were tasked to write PR FAQ documents and submit them for review by the executive team.

First hires on a project team have an outsized influence on the eventual success of the project.

Autonomous teams

Both digital media and AWS divisions were hooked “high” in the organisation structure and the decision was made that they should use little (if any) of the existing infrastructure, but rather build it up on their own. This is because both projects were recognised to have a high potential early on. If these projects depended on the existing organisational and system structures, this could have limited their “search space” and prevented them from finding and realising some interesting product ideas.

When teams have to drag through technical and organisational dependencies to complete their task, they are disempowered, frustrated and demotivated. People should feel agency to be able to take pride in their work.

If a company is organised into autonomous teams, this makes decisions about launching or discontinuing products less consequential for the whole company. This allows increasing the risk appetite which means higher overall returns for the company, even though a bigger proportion of products will fail.

Generate more ideas than you have resources to realise and filter them

Only about 10% of product or service ideas that have had a PR FAQ document have been eventually realised at Amazon.

The executive team often decided what not to do rather than what to do. Making clear why you aren’t doing something is often as important as having clarity about what you are doing.

This idea is similar to Popper’s approach to science, critical rationalism: putting forward many bold hypotheses and then testing them rigorously until only the fittest survive. I touched upon this topic in the recent post about causality.

This summary was first published on Substack.




Writings about software, systems, reliability, and data engineering, software operations, peopleware, philosophy, etc. by Roman Leventov. Originally published at

Recommended from Medium

This Is the Only Way to Reward Your Employees

Meet a Member: Nilay Aydogan

BDD Part 2: You are not as efficient as you think you are

Connecting with the great people around you

My management lessons from a man who missed 2 Nobel Prizes

James River Capital and the Importance of Feedback

Making better decisions as Product Owner

How to Challenge Your Mental Models and Think Differently

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Roman Leventov

Roman Leventov

Writing about systems, technology, philosophy.

More from Medium

“Real-time”is becoming the default expectation. What is holding it back?

The Basics of any Consulting or Services Business

The Ivory Tower Problem

Startup Growing Pains — Lessons & Questions — Part 2— Organisational Structure