A flight in 2004
Jeff Bezos and Colin Bryar spent a business flight reading Edward Tufte's essay The Cognitive Style of PowerPoint. Tufte's argument was that as analysis becomes more causal, comparative, and evidence-based, the more damage a bullet list does to it. His closing recommendation was blunt: make your presentation software Word rather than PowerPoint.
Amazon did exactly that. Slides were replaced with six-page narrative memos for decision meetings, read in silence for the first twenty minutes before discussion begins.
Bezos's own stated reasoning is the useful part: writing a good four-page memo is harder than producing a twenty-page deck, because the narrative structure forces better thought — better understanding of what matters more than what, and how things relate to each other.
Six pages was arrived at by trial and error. Long enough for depth, short enough to force clarity.
The mechanism: writing exposes what you do not know
A bullet list permits an idea to look complete while the relationships between its parts remain unexamined. Nothing in the format requires stating why one point follows another, and nothing exposes it when the connection does not exist.
Prose does not allow this. Sentences must connect. Causation has to be stated, and once stated it becomes checkable.
This maps onto a well-documented cognitive finding. Rozenblit and Keil described the illusion of explanatory depth in Cognitive Science in 2002: people consistently believe they understand how things work in more detail than they actually do, and the illusion collapses the moment they are asked to produce a full explanation. Asking someone to rate their understanding of a bicycle produces confidence. Asking them to draw one working produces humility.
A product idea behaves identically. It feels complete in discussion and reveals its gaps in prose.
Bezos read these memos assuming each sentence was wrong until proven otherwise. That is only possible against a document where sentences make claims.
What a PR-FAQ contains
The press release is written as though the product has shipped, aimed at the customer rather than at internal stakeholders. Roughly one page.
Headline. The product named in terms of what it does for the customer.
Sub-headline. Who it is for and the benefit.
Problem. The situation the customer is currently in, described from their side.
Solution. How it works, in plain language.
Customer quote. What a customer would say. If this quote is embarrassing to write, the value proposition is not there.
How to get started. What the customer does next.
The FAQ carries the weight, and is longer than the release. Two sections.
Customer FAQ. What a customer would ask. Pricing, availability, comparison to alternatives, what happens when it goes wrong.
Internal FAQ. The hard questions. What is the market size and how do we know. What is the hardest technical problem. What does this cost to build and serve. What has to be true for this to work. What would make us stop.
Most of the thinking happens in the internal FAQ. Teams that write a strong press release and a weak internal FAQ have written marketing copy.
AWS took more than a year of documents
Bezos and Andy Jassy identified web services as an essential technology in 2004. S3 and EC2 launched in 2006.
The gap was spent writing, revising, debating, and rewriting PR-FAQs — establishing what problem AWS solved, who the customer was, and what the right products were. Once the documents were settled, the critical decisions were made and captured, and anyone at Amazon could read them and arrive at the same understanding.
The stated trade is the point: heavy investment in planning bought speed in production. That is a specific claim about where time goes, not an argument for slower decisions.
The evidence, honestly
This is a documented practitioner method with an unusually narrow evidence base: one company, described by two of its former executives.
Three qualifications worth stating.
Amazon succeeded for many reasons. Attributing outcomes to any single internal mechanism is not supportable. The company had capital, scale, timing, and a decade of infrastructure investment alongside its writing culture.
The process did not prevent failures. The Fire Phone went through Amazon's mechanisms and was a substantial write-off. A process that produces good documents can still produce bad products, and any account suggesting otherwise is selling something.
The full ban is a simplification. Slides remain in use at Amazon for all-hands meetings and broadcast communication. What was replaced was slides in decision meetings, which is a narrower and more defensible claim.
The reason to use it anyway is the mechanism rather than the company. Prose exposes gaps that bullets conceal, and that holds independently of who first noticed it.
Six disciplines that make it work
1. Write it before resourcing, not after. A PR-FAQ produced to justify a decision already made is documentation. Written before, it is a test.
2. Read in the room, in silence. Pre-reading does not happen. Twenty minutes of silent reading at the start of the meeting guarantees everyone has actually read the same thing.
3. Give the internal FAQ more space than the press release. The hard questions are where the value is. A one-page release with four pages of FAQ is the right proportion.
4. Include the stopping question. What would make us abandon this. A document without an answer is a proposal, not an evaluation.
5. Expect several drafts. The first version of anything worth building will not survive its own internal FAQ. Amazon spent a year on AWS documents.
6. Kill things with it. This is the only test that matters, and it is covered below.
The only test that matters
Has a PR-FAQ ever caused an idea to be dropped?
If the answer is no, the process is producing documentation rather than decisions. Every idea that entered the process left it approved, which means the document was written to justify rather than to evaluate.
The mechanism only works if failing is a real outcome. A team that knows nothing ever gets killed will write accordingly, and the resulting document will be fluent, complete, and worthless.
Where it goes wrong
It becomes a template to complete. Teams fill sections rather than answering questions. The format is a container; the thinking is the product.
The press release is written for internal readers. Language drifts toward strategic rationale and away from customer benefit. If the customer would not care about a sentence, it does not belong in the release.
The customer quote is invented enthusiasm. A quote nobody would actually say is the clearest available signal that the value proposition is thin.
It replaces customer research. A PR-FAQ articulates a hypothesis clearly. It does not validate it. Clear articulation of an untested assumption is still an untested assumption.
It is applied to everything. The overhead is justified for significant commitments. Applied to small decisions it becomes ceremony, and the organisation learns to treat it as such.
Diagnostic: is the process working?
Six tests.
At least one idea has been dropped as a result of a PR-FAQ in the past year.
The internal FAQ is longer than the press release.
Every document answers what would make the business stop.
Documents are written before resources are committed.
Reading happens in the meeting, not before it.
The customer quote could plausibly have been said by a real person.
Test one is the whole thing. The rest are how you get there.
What this produces
Shared understanding that survives the meeting.
The specific failure this addresses is the one where six people leave a decision meeting agreeing, and hold six different pictures of what was agreed. Slides permit that outcome because they carry the presenter's interpretation rather than the content. A document that has been read closely by everyone does not.
The second output is more valuable and less discussed: a written record of what the team believed and why, at the moment of commitment. When the thing ships and underperforms, that document is the only honest reference for what went wrong — the assumption or the execution. Most organisations cannot tell the difference afterwards, because nobody wrote the assumption down.
Frequently asked questions
What is a PR-FAQ?
A document written before building, containing a press release describing the finished product as though it has launched, plus customer-facing and internal FAQ sections. Amazon uses it to test whether an idea is worth building.
Why did Amazon ban PowerPoint?
Bezos and Colin Bryar read Edward Tufte's essay on the cognitive style of PowerPoint in 2004, which argued that bullet lists damage causal and evidence-based analysis. Amazon replaced slides with six-page narrative memos in decision meetings. Slides are still used for all-hands and broadcast communication.
Why is writing better than slides for decisions?
Prose requires stating how ideas relate to each other, which bullets allow teams to leave unexamined. It connects to the illusion of explanatory depth — people believe they understand things in more detail than they do, until asked to produce a full explanation.
How long should a PR-FAQ be?
Around one page for the press release and several more for the FAQ. Amazon settled on six pages total through trial and error — long enough for depth, short enough to force clarity.
Does the process guarantee good products?
No. The Fire Phone went through Amazon's mechanisms and failed commercially. The process improves the quality and shared understanding of a decision. It does not make the decision correct.
How do we know if we are doing it properly?
Ask whether any idea has ever been dropped because of a PR-FAQ. If nothing has ever failed the test, the documents are being written to justify decisions already made.
Does this replace customer research?
No. A PR-FAQ states a hypothesis clearly and makes its assumptions visible. Validating those assumptions is separate work, and clear articulation of an untested belief is still an untested belief.
Sources
Bryar, C. and Carr, B., Working Backwards: Insights, Stories, and Secrets from Inside Amazon (2021)
Tufte, E., The Cognitive Style of PowerPoint: Pitching Out Corrupts Within
Rozenblit, L. and Keil, F., The Misunderstood Limits of Folk Science: An Illusion of Explanatory Depth, Cognitive Science (2002)
Published accounts of the AWS PR-FAQ process, 2004 to 2006
Structure your next phase
Zerologic works with leadership teams to define what is being built and why — in a form specific enough that it can fail before resources commit to it.
Talk to us: partners@zerologic.io · zerologic.io
