The Press Release That Kills Products: Amazon's PR/FAQ, As Amazon Tells It

On February 2, 2005, Amazon announced a membership program: pay $79 a year, get unlimited two-day shipping. The reception, in the telling Amazon’s own veterans give, was that the idea was crazy and stupid — a company with famously thin margins volunteering to eat its shipping economics for a flat fee. Two and a half years later, on November 19, 2007, the same company launched a $399 electronic book reader into a market that had already watched several e-readers die quietly. The Kindle sold out in under six hours.

Two launches, both greeted with some version of why would anyone want this, both of which turned out to be load-bearing pillars of one of the most valuable companies on earth. And in the account written by the people who were in the room — Colin Bryar, who spent years as Jeff Bezos’s chief of staff, and Bill Carr, a longtime Amazon executive — both trace back to the same unglamorous artifact: a press release, written years before launch day, for a product that did not exist. Amazon calls the process Working Backwards, and the document at its center the PR/FAQ. I wrote about how the mechanism works elsewhere; this is the story of where the mechanism comes from — and what an honest reading of that story actually supports, which is both less and more than the legend.

The story as Amazon tells it

By Bryar and Carr’s account in Working Backwards, the process took hold around 2004, out of frustration with how new product ideas were being argued: decks, meetings, enthusiasm, and no forcing function that made anyone confront whether a customer would care. The fix Bezos and his team converged on was to make every proposed product start its life as its launch announcement. One page of press release, written in the language a customer would actually read, followed by pages of FAQ answering the questions a skeptic would actually ask. No committee could greenlight a product until the product could survive its own announcement — on paper, in draft after draft, in front of hostile readers.

The showpieces of the account are exactly the two launches above, plus AWS. Prime, in this telling, was shaped as a written narrative before it was an engineering project: the $79 price, the two-day promise, the argument for why customers would change their buying behavior — all argued in prose while the whole thing was still reversible. The Kindle likewise: years of drafts in which the team had to explain, in a customer’s voice, why anyone would pay hundreds of dollars for a worse-looking screen, and what — wireless delivery, a bookstore in your hand, sixty seconds from wanting a book to reading it — would make the announcement actually contain news. The insiders’ claim is blunt: the biggest bets of Amazon’s second decade were not products that happened to get press releases. They were press releases that eventually got products.

It’s a great story. Now hold it at arm’s length for a moment.

What the mechanism actually claims

Here’s the part of the account I find most credible, and it’s not the hits. Bryar and Carr are careful — more careful than the people who summarize them — to locate the value of the PR/FAQ in the drafts that failed. A team writes the announcement for its idea, and somewhere around draft three or four, the document quietly confesses. The customer quote reads as faint praise. The headline has no news in it. The FAQ’s pricing answer doesn’t survive its own arithmetic. And the product dies — on paper, in a conference room, for the cost of some writing time and some bruised pride.

That is the mechanism’s real claim, and notice what kind of claim it is. It isn’t “this process generates brilliant ideas.” It’s a filter claim: products that couldn’t survive their own press release died cheaply, before headcount, before roadmap slots, before sunk cost made killing them a political event. The PR/FAQ is a falsification device. Its output isn’t inspiration; it’s corpses — cheap, early, well-documented corpses — and a decision-making culture where a product’s first contact with skepticism happens when the whole bet still costs four pages to reverse.

I believe this part of the account almost independently of trusting Amazon’s self-mythology, because it doesn’t require trusting anyone. The claim that writing a concrete narrative exposes a vague idea is checkable by anybody with an afternoon: write the honest launch announcement for the thing your team is currently building, and see whether it contains news. In my own experience the document confesses with almost embarrassing speed — the vague idea that survived six roadmap reviews rarely survives one paragraph of a customer describing why they switched. Prose has no place to hide a wish. That property belongs to the format, not to Amazon.

The honest reading: insider-narrated history

And now the caveat that most retellings skip, because it makes the story less quotable: nearly everything above is Amazon’s own account of Amazon. Bryar was Bezos’s chief of staff. Carr ran businesses the process shaped. Working Backwards is an insider memoir with a thesis, written by people whose careers are braided into the thing they’re evaluating. That doesn’t make it false — insider accounts are exactly where the texture of a mechanism comes from, and the public facts that can be checked (the dates, the prices, the sellout) hold up. But the causal claim, the PR/FAQ produced Prime, Kindle, and AWS, is not an audited finding. It’s the narrators’ attribution, and every force that bends memoirs — survivorship, hindsight, the flattening of a decade of messy argument into a clean origin story — is bending this one.

Consider what we can’t see from outside. We don’t know how many strong ideas the process killed alongside the weak ones; a filter’s false negatives are, by construction, invisible. We don’t know how much of Prime’s success belongs to the document versus to Amazon’s logistics network, capital patience, and a founder with unusual appetite for margin-eating bets — advantages the press release format does not confer on anyone who adopts it. And there’s a quieter absence worth sitting with: for a process this famous, adopted (reportedly, enthusiastically, everywhere) for over a decade, the well-documented non-Amazon PR/FAQ success story is strikingly hard to find. Plenty of companies say they use it. I can’t point you to a rigorously told case, with names and numbers, of the process producing a Prime-sized outcome anywhere else — and when a mechanism’s evidence base is one company narrating itself, intellectual honesty requires saying so.

None of that is a debunk. It’s a re-scoping. The strong claim — adopt this document and you’ll find your Kindle — is unsupported and probably unsupportable. The narrow claim survives every dose of skepticism I can administer: writing the narrative before the code is a cheap, repeatable way to make bad bets fail early, and the discipline is reproducible even if the outcomes aren’t.

What transfers, and what doesn’t

So what do you actually get to take from this story? Not the outcomes — those came bundled with Amazon’s balance sheet and Bezos’s risk appetite. What transfers is the epistemic move: before committing to a build, force the idea to make a falsifiable public promise, in customer language, and then attack the promise in writing. That move is a risk-reduction instrument in exactly the sense I mean in the decision-making post — it’s the cheapest available test of the most expensive assumption, which is almost always “someone will care.”

What doesn’t transfer is the ceremony without the hostility. The failure mode I’ve watched in teams that import the format is press releases written to be approved rather than to be attacked — beautifully worded, FAQ questions chosen for answerability, nobody in the review empowered to say this contains no news. That’s the pageantry of Working Backwards with the falsification removed, and it’s worse than nothing, because the surviving document now functions as evidence for a bet that was never actually tested. The Amazon story, even taken fully at face value, never claims the writing is the mechanism. The dying is the mechanism. If nothing you write is allowed to die on paper, you’re not working backwards; you’re decorating a decision that was already made.

That’s the lens I’d offer for the whole legend. Prime at $79 sounded crazy and stupid; the Kindle at $399 sold out in under six hours; and the people who were there say a document made the difference. Believe the dates, respect the discipline, hold the attribution loosely — and steal the one part that never depended on being Amazon: make your ideas survive their own announcement before they’re allowed to cost real money.

Put it to work

1. Run one idea through an honest death-check this month. Take the initiative your team is most attached to and write its launch announcement — one page, customer language, a quote a real customer could plausibly say. Then apply the single test the whole Amazon account rests on: does the announcement contain news? If the honest version doesn’t, you’ve just gotten the mechanism’s entire promised value — a cheap corpse — for the price of an afternoon.
2. Count your paper deaths. Look back over the last year and ask how many initiatives your team killed in a document review versus in a retro after shipping. If the answer is zero-in-review, you don’t have a working-backwards problem, you have a falsification problem: nothing in your process is allowed to fail while failing is still cheap, and no imported format will fix that until someone in the room is empowered to say “there’s no news here.”
3. Label your evidence the way this story should be labeled. Next time you cite a company’s process as the reason for its success — Amazon’s PR/FAQ, anyone’s anything — say out loud whose account it is. “Per the insiders who ran it” is a different epistemic grade than “independently documented,” and teams that keep the distinction make better decisions about which borrowed practices deserve real investment.

Further reading

• Colin Bryar & Bill Carr, Working Backwards: Insights, Stories, and Secrets from Inside Amazon — the primary account of the process, Prime, Kindle, and AWS; read it as an insider memoir with a thesis, which is both its value and its caveat.
• Colin Bryar, “Working Backwards: How to Write an Amazon PR/FAQ” (coda.io) — Bryar’s own public walkthrough of the document’s structure, the closest thing to a first-party template.
• Jeff Bezos’s shareholder letters — the narrative-over-slides culture in its original voice, and useful context for why prose carried so much decision weight at Amazon.