TL;DR: Write the press release before you build the product. Amazon’s PR/FAQ process forces teams to articulate the customer value proposition in plain language before writing a single line of code. Most PR/FAQs never ship — and that is the entire point. The kill rate is the feature.
What it means
Working Backwards is Amazon’s mechanism for deciding what to build. Before any project gets funded, a team writes a mock press release announcing the finished product, followed by an FAQ addressing both customer and internal questions. The document is written in the future tense, as if the product already exists, with the launch date as if the work were done (working-backwards).
The press release format is deliberately constraining and that’s the point. You can’t hide behind technical jargon, market sizing, or competitive analysis. You have to answer the only questions that actually matter: What is it? Who is the customer? Why should they care? What makes this meaningfully better than what already exists? If you can’t write a compelling one-page PR, the product isn’t ready to build. Often the product isn’t ready to exist.
The argument
The kill rate is a feature. Most PR/FAQs die in review. Teams discover that they cannot articulate a clear customer benefit, or that the FAQ exposes fatal assumptions about willingness to pay or competitive response. This is the entire intended outcome. The process is designed to kill bad ideas cheaply, before engineering resources are committed. The cost of a failed PR/FAQ is a few weeks of writing and arguing. The cost of a failed product is months or years of building, plus the opportunity cost of every other thing you could have done with those engineers (working-backwards).
Kindle as the proof. Amazon had no hardware capabilities when they decided to build the Kindle. The PR/FAQ process forced them to confront the question: if the customer wants to read any book, anywhere, instantly — what do we need to build? The answer required hardware, which Amazon had never done. But because the customer value proposition was so clear, the organization committed to acquiring the capability rather than constraining the vision to existing competence (working-backwards).
This is the method’s deepest lesson and the part most teams fail at: start from what the customer needs, not from what you can currently do. Capabilities can be acquired. Customer insight that shapes the right product is much harder to develop, and easier to fake.
Single-threaded leadership. Amazon pairs Working Backwards with a structural principle: one leader owns one initiative, full-time. No splitting attention across projects. The single-threaded leader is empowered to make decisions without navigating committee approval. This is how a massive organization maintains startup-like focus on individual products. The trade-off is real — you need more leaders — and Amazon decided that was a cost worth paying. Most companies decide otherwise and then wonder why their products feel watered down (working-backwards).
Two-pizza teams. Teams small enough that two pizzas can feed them. Small teams reduce communication overhead, increase ownership, and make it harder to hide. Combined with single-threaded leadership, this creates autonomous units inside a large organization that can move with the speed of an ooda-loop-optimized startup. This is how Amazon stays fast at scale — by refusing to centralize.
Mechanisms over intentions. Amazon’s broader philosophy: don’t rely on people to do the right thing. Build processes and systems that make the right thing the default. Working Backwards is itself a mechanism — it mechanizes the discipline of starting from the customer. Other Amazon mechanisms include the six-page narrative memo (replacing slide decks, because slides hide bad thinking), the “disagree and commit” framework, and the bar raiser process for hiring (working-backwards).
The synthesis
Working Backwards is the anti-ooda-loop in one sense: it’s slow, deliberate, and front-loaded with analysis. But in Boyd’s framework, it’s actually optimizing the Orient step — building the sharpest possible mental model of the customer before acting. A company that spends four weeks on a PR/FAQ and then builds with conviction will outperform one that ships fast, discovers the positioning is wrong, and pivots three times. Working Backwards isn’t slow strategy. It’s front-loaded strategy that pays off in execution speed later.
The method also connects to simplicity-as-strategy: the press release format forces simplicity. You cannot announce a complex, multi-feature product in a press release that a customer would actually read. The constraint of the format pushes toward the simplest version of the idea that delivers real value. If your PR has more than two features in it, your product is probably already too complicated.
Loose threads
- Working Backwards assumes you can articulate the customer before building. For genuinely novel categories — where the customer doesn’t yet know they want the thing — the method may over-constrain exploration. ChatGPT is the obvious counterexample: no PR/FAQ in 2022 would have predicted “people will use this to think out loud about their relationships.”
- The method works at Amazon’s scale because they can afford to kill most PR/FAQs. Startups may not have enough shots on goal to tolerate that kill rate. The startup version is closer to “kill the idea fast in conversation” than “write a six-page document about it.”