Jeff Bezos's Decision Principles: The Amazon Framework
Who it's for: Founders, product leaders, managers, and strategists
Primary sources: Amazon shareholder letters (1997β2021), Invent and Wander (2021)
Core principles: 8 documented frameworks
Jeff Bezos built Amazon from an online bookstore in a garage (1994) into one of the world's most valuable companies β across e-commerce, cloud computing, logistics, entertainment, and hardware. What's unusual is how explicitly and consistently he documented his decision-making philosophy across 25 years of shareholder letters.
Unlike most CEOs who describe strategy in hindsight, Bezos articulated frameworks in advance and then demonstrated them through repeated decisions. The result is one of the most coherent and well-evidenced decision philosophies in business history.
"We are stubborn on vision. We are flexible on details."
β Jeff Bezos
Principle 1: Two-Way Door vs. One-Way Door Decisionsβ
This is Bezos's most operationally important framework β and the one most applicable to any organisation at any scale.
The Frameworkβ
Decisions differ fundamentally in their reversibility:
Type 1 decisions (One-Way Doors): Consequential, largely irreversible. Walking through and discovering you don't like what you find, you can't easily get back. These require careful deliberation, senior involvement, and conservative analysis. Examples: major acquisitions, fundamental product architecture, entry into a new market.
Type 2 decisions (Two-Way Doors): Reversible, changeable. If you make a suboptimal Type 2 decision, you can undo it quickly at low cost. These should be made quickly, at a low level in the organisation, by a small team. Examples: most product features, pricing tests, hiring decisions, marketing experiments.
The Insight Most Organisations Missβ
Most large organisations apply Type 1 process to Type 2 decisions. They require broad consensus, multiple sign-offs, and extensive analysis for decisions that could simply be tried and reversed if wrong. The result: institutional slowness that provides no corresponding risk reduction.
Bezos described this as one of the primary ways large organisations become ineffective: "Decisions are made slowly but with the same mediocrity as fast decisions β you get the cost of deliberation without the benefit of better outcomes."
The diagnostic question for any decision: "If this turns out to be wrong, how costly and difficult is it to undo?" If the answer is "not very" β it's a Type 2 decision, and slow process is waste.
Principle 2: Working Backwardsβ
What It Isβ
Before building a product, Amazon writes:
- A press release announcing the finished product as if it has already launched β written in customer-facing language, describing the benefit to the customer
- A FAQ answering the hardest questions customers and skeptics will ask
- A user manual describing how the product actually works
If the team can't write a compelling press release, the product isn't ready to build. The discipline forces clarity about customer value before any engineering begins.
Why It Worksβ
Traditional product development starts with what we can build and works forward to what customers might want. Working Backwards inverts this: it starts from what customers would find genuinely valuable and works backward to what needs to be built.
The press release serves as a quality filter. A press release that no one would want to read reveals that the product no one would want to use. Discovering this at the press-release stage costs a few hours. Discovering it after 18 months of development costs much more.
The Amazon Kindle origin story: The Kindle concept started as a press release describing a book reader that could hold a library, download any book in 60 seconds, and last a week on battery. The press release was compelling. The product was built. It launched in 2007 and created an entirely new product category.
β Full model: Working Backwards
Principle 3: Day 1 Mindsetβ
This is Bezos's most recurring theme β referenced in nearly every shareholder letter across 25 years. Amazon's original headquarters building in Seattle is named "Day 1."
Day 1 vs. Day 2β
Day 1: Constant invention, customer obsession, high-velocity decision-making, comfort with being misunderstood for long periods, willingness to cannibalise existing business.
Day 2: Process worship, metric proxies for customer outcomes, slow decisions by consensus, defensive strategy, avoidance of cannibalisation.
Bezos's definition: "Day 2 is stasis. Followed by irrelevance. Followed by excruciating, painful decline. Followed by death. And that is why it is always Day 1."
The Four Day 2 Trapsβ
Bezos identified four specific patterns that signal a company has entered Day 2:
Trap 1: Process Worship When process becomes the goal rather than the output, you're in Day 2. The signal: when people can't remember why a process exists, only that it does. Processes are tools for achieving outcomes; when they survive the outcomes they were designed for, they've become a disease.
Related: β Goodhart's Law β metrics become targets and stop being good metrics
Trap 2: Proxy Metrics Using metrics as substitutes for the actual customer outcome. "Our NPS score went up" can coexist with customers getting worse service if NPS is being gamed. The signal: when improving the metric becomes possible without improving the underlying reality.
Related: β Goodhart's Law
Trap 3: Consensus Decisions Waiting for everyone to agree produces decisions that everyone can tolerate and no one believes in. Bezos's response is "Disagree and Commit" β explicitly articulated disagreement followed by genuine commitment. The worst outcome is uncommitted agreement: people who voted yes and act as if they voted no, waiting for the decision to fail.
Related: β Precommitment
Trap 4: Trend-Following Adopting new methodologies (AI, agile, blockchain at various historical moments) because competitors are adopting them rather than because they serve customers. The signal: the language of the methodology substitutes for customer outcome thinking.
Principle 4: Long-Term Thinkingβ
Bezos's 1997 shareholder letter β which he attached to every subsequent letter as a reminder β established the core framework: optimise for long-term free cash flow per share, not quarterly earnings.
What This Required in Practiceβ
This wasn't just rhetoric. Amazon accepted:
- 8 years of losses in its core retail business while building the infrastructure for scale
- Analyst and media criticism for prioritising investment over profit through the early 2000s
- Cannibalisation risk β Kindle risked the physical book business; Amazon Web Services risked the retail hosting business
- Long payback periods β AWS launched in 2006; it became Amazon's largest profit centre in 2015
The long-term framing allowed these investments. A company optimising for quarterly earnings would have made none of them.
The Regret Minimisation Test: Bezos has described the decision to start Amazon using this framework β imagining himself at 80, looking back, and asking which regret would be larger: not trying and missing the opportunity, or trying and failing. Applied to long-term decisions, it consistently points toward action and investment rather than short-term optimisation.
β Full model: Regret Minimization Framework
Principle 5: The Flywheel Effectβ
In 2001, Bezos sketched a diagram on a napkin that described Amazon's growth model as a virtuous cycle:
Lower prices β more customers β more sellers β wider selection β better customer experience β lower costs (scale) β lower prices
Each element reinforces every other element. Investing in any element of the loop accelerates all elements. This is not a sequential strategy β it's a reinforcing system. The correct strategic goal is not to win at any single element but to spin the flywheel.
The flywheel insight most companies miss: Prime membership was not primarily a revenue stream. It was a flywheel accelerant β a mechanism to increase the frequency and depth of customer relationships with Amazon, which accelerated every other element of the loop. By 2023, Prime had over 200 million members globally.
β Full model: Flywheel Effect
Principle 6: High Standards as a Learnable Skillβ
In his 2017 shareholder letter, Bezos argued that high standards are teachable and contagious β not fixed character traits.
The Two Requirements for High Standardsβ
Requirement 1: Recognition β You must be able to identify high-quality output in this domain. You can't enforce standards you can't recognise.
Requirement 2: Realistic Expectation β You must know how difficult it is to achieve high quality, so you don't abandon the effort when it proves harder than expected.
Most high-standard failures happen because people can recognise quality but underestimate the effort required to achieve it. They set high standards, encounter difficulty, and conclude the difficulty means something is wrong β rather than that producing excellent work is simply hard.
The application: When setting standards in an organisation, Bezos recommends being explicit about both what the standard looks like and how hard it is to reach. "We want world-class customer service" is insufficient. "We want world-class customer service, which means X, Y, and Z specifically, and which typically takes 18 months of sustained effort to build from scratch" is actionable.
Related: β Pygmalion Effect β high expectations consistently held tend to produce high outcomes
Principle 7: The Narrative Memo (No PowerPoint)β
Amazon banned PowerPoint in senior meetings in 2004. Every significant proposal begins with a 6-page narrative memo, read silently by all attendees in the first 15β20 minutes of the meeting.
Why Narrative Over Slidesβ
Bezos's reasoning: "PowerPoint-style presentations somehow give permission to gloss over ideas, flatten out any sense of relative importance, and ignore the interconnectedness of ideas." Bullets make it easy to hide weak reasoning in confident-sounding fragments.
Narrative prose requires the writer to have actually thought through the argument. It forces a logical structure β sentences connect causally, not just sequentially. It reveals gaps in reasoning that bullets allow you to skip.
The silent reading rule: Everyone reads the memo before discussion begins. This eliminates the "first speaker" effect β where whoever presents first anchors the entire discussion β and ensures all attendees have processed the same complete argument before any responses are given.
Related: β Brainwriting β parallel silent processing eliminates dominant-voice dynamics
Principle 8: Customer Obsession Over Competitor Obsessionβ
Bezos consistently distinguished Amazon's culture from competitor-obsessed cultures:
"We don't try to figure out what Amazon should be doing. We try to figure out what customers need and then work backward. We're customer obsessed, not competitor focused."
Why Competitor Focus Is Dangerousβ
Companies that focus on competitors fight over existing territory. They improve on what already exists. They respond to moves rather than creating new ground.
Customer focus asks: what do customers actually want that doesn't exist yet? What would they want if they knew it was possible? This produces expansion into new categories and the creation of new markets β what Bezos called "wandering" β rather than incremental competition for existing share.
The practical test: In product reviews, count questions that reference customers vs. questions that reference competitors. In companies that have drifted toward Day 2, competitor-referencing questions dominate.
Related: β Jobs to Be Done β understanding what customers are actually trying to accomplish, not just what they're currently buying
The Amazon Decision-Making Framework Appliedβ
| Principle | The Question It Answers | Model |
|---|---|---|
| Two-Way Door | How much process does this decision warrant? | Two-Way Door |
| Working Backwards | What are we actually trying to build and why? | Working Backwards |
| Day 1 | Is this decision coming from invention or inertia? | Goodhart's Law |
| Long-Term Thinking | Am I optimising for the right time horizon? | Regret Minimization |
| Flywheel | Does this investment accelerate the system? | Flywheel Effect |
| High Standards | Do we know what good looks like and how hard it is? | Pygmalion Effect |
| Narrative Memo | Have we actually thought this through rigorously? | Brainwriting |
| Customer Obsession | Are we solving real customer problems or playing competitive games? | Jobs to Be Done |
Frequently Asked Questionsβ
Q: Which Amazon principle is most applicable to early-stage startups?
Two-Way Door decisions and Working Backwards. Early-stage companies typically err in two directions: either overthinking reversible decisions (slowing everything down) or underthinking irreversible ones (making permanent architectural choices casually). The Two-Way Door framework addresses the first error; Working Backwards addresses the second by forcing customer clarity before commitment.
Q: How do you actually run a "Working Backwards" process?
Start by writing the press release headline and first paragraph only. If you can't write a headline that someone would want to click, the concept needs more development. Work backward through the document β the FAQ is often the most revealing: if the answers to the hard questions are weak or unclear, the product isn't ready. Bezos's original memo requirement was that the document should be crisp and compelling β if it takes 8 pages to explain why a customer should care, the value proposition isn't clear enough.
Q: Is the "Day 1" concept applicable to older, larger companies?
Yes, but it requires deliberate structural protection. Large companies naturally drift toward Day 2 because the incentive structure of a large organisation (stability, risk-aversion, consensus) differs from an early-stage company's (speed, experimentation, tolerance for failure). Bezos's internal solution was to create distinct business units (AWS, Alexa, Kindle) that operated as separate companies with Day 1 cultures, insulated from the Day 2 gravitational pull of the core retail organisation.
Further Readingβ
- Bezos, J. (1997β2021). Amazon Shareholder Letters β the primary source; all 24 letters available free at ir.aboutamazon.com
- Bezos, J. & Stone, B. (2021). Invent and Wander β curated essays and shareholder letters
- Stone, B. (2013). The Everything Store β narrative account of how these principles operated in practice
Apply This Collection with AIβ
π Apply Bezos's decision principles to your situation in MindMax β
Part of the MindMax Mental Models Knowledge Base. See also: The Founder's Toolkit Β· Elon Musk's Frameworks