-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathpr-faq-framework.html
More file actions
215 lines (197 loc) · 20.4 KB
/
pr-faq-framework.html
File metadata and controls
215 lines (197 loc) · 20.4 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Amazon Working Backwards / PR-FAQ — define impact before building</title>
<link rel="stylesheet" href="framework.css">
<style>
/* Page-accent — overrides framework.css fallback */
:root{--page-accent:var(--gold);--page-accent-soft:var(--gold-soft)}
/* PR-FAQ artifact */
.doc{background:#fff;border:1px solid var(--line);border-radius:12px;padding:24px;box-shadow:var(--shadow);margin:14px 0}
.doc .pr,.doc .faq{border-left:5px solid var(--page-accent);padding:14px 18px;background:var(--page-accent-soft);border-radius:8px;margin:8px 0}
.doc .label{font-family:Georgia,serif;font-size:12px;color:var(--page-accent);font-weight:700;letter-spacing:.06em;text-transform:uppercase;margin-bottom:5px}
.doc h4{margin:0 0 6px;font-size:15px}
.doc ul{margin:4px 0;padding-left:20px;font-size:13.5px;color:var(--ink-soft)}
/* example */
.ex{background:#fff;border:1px dashed var(--page-accent);border-radius:12px;padding:18px 22px;margin:14px 0;font-size:14.5px}
.ex h4{margin:0 0 8px;color:var(--page-accent);font-family:Georgia,serif;font-size:16px}
.ex p{margin:6px 0}
</style>
</head>
<body>
<nav class="sitenav">
<details>
<summary>📑 Jump to</summary>
<div class="navmenu">
<div class="navgrp"><h4>Start here</h4>
<a href="index.html"><b>← Home (goal & map)</b></a>
<a href="impact-saas-companies.html">SaaS / B2B field study</a>
<a href="impact-consumer-companies.html">Consumer-tech field study</a>
<a href="methodologies-comparison.html"><b>All methods compared →</b></a>
<a href="experiment-trustworthiness.html">How 40k tests actually work →</a>
<a href="jargon.html">Jargon (glossary)</a>
</div>
<div class="navgrp"><h4>Scoring & Input modeling</h4>
<a href="rice-framework.html">RICE (Intercom)</a>
<a href="north-star-framework.html">North Star (Amplitude / Slack)</a>
</div>
<div class="navgrp"><h4>Goal-laddering / Define first</h4>
<a href="v2mom-framework.html">V2MOM (Salesforce)</a>
<a href="pyramid-of-clarity-framework.html">Pyramid of Clarity (Asana)</a>
<a class="cur" href="pr-faq-framework.html">PR-FAQ / Working Backwards (Amazon)</a>
<a href="heart-framework.html">HEART (Google)</a>
<a href="dibb-framework.html">DIBB (Spotify)</a>
</div>
<div class="navgrp"><h4>Experimentation (SaaS)</h4>
<a href="microsoft-exp-framework.html">Microsoft ExP / CUPED</a>
<a href="linkedin-xlnt-framework.html">LinkedIn T-REX</a>
</div>
<div class="navgrp"><h4>Experimentation (Consumer)</h4>
<a href="netflix-experimentation.html">Netflix · ABlaze</a>
<a href="booking-experimentation.html">Booking.com</a>
<a href="airbnb-erf-framework.html">Airbnb ERF</a>
<a href="uber-xp-framework.html">Uber XP</a>
<a href="doordash-switchback-framework.html">DoorDash switchback</a>
<a href="lyft-experimentation.html">Lyft</a>
<a href="pinterest-ab-framework.html">Pinterest</a>
</div>
<div class="navgrp"><h4>AI labs</h4>
<a href="anthropic-pm-on-ai-exponential.html">Anthropic · PM on AI exponential</a>
<a href="google-customer-zero-2026.html">Google · "Customer zero" 2026</a>
</div>
<div class="navgrp"><h4>Written discipline</h4>
<a href="stripe-shaping-framework.html">Stripe shaping</a>
</div>
</div>
</details>
</nav>
<div class="wrap">
<header class="masthead">
<p class="kicker">Methods · Deep-dive · Define impact before building</p>
<h1>Amazon Working Backwards / PR-FAQ — write the launch before the code <span class="srcyr">2004</span></h1>
<p class="sub">A mock <strong>press release</strong> plus an <strong>FAQ</strong> — both written before a line of code is shipped. Used at Amazon since 2004 across both consumer (Kindle, Prime, Echo) and B2B (AWS) products.</p>
<p class="sub">The discipline: if you can't write a compelling PR for the customer, you don't yet have a feature worth building.</p>
<div class="goal"><span>Goal</span><br>Decide features by data-backed expected impact — choose by outcome, not by to-do list or opinion.</div>
</header>
<div class="eli">
<div class="lbl">🎓 8th-grade version</div>
Before Amazon builds anything, somebody writes the news article announcing it — like it's already done and being sold. One page. Then a list of all the boring questions: how much does it cost, who's it for, what could go wrong, how do we know if it worked? If the news article sounds dumb or fake when people read it, the idea isn't ready — you change the idea, not the writing. Most ideas die at this step, before any code is written. The ones that survive get built, because by then they're worth building.
</div>
<nav class="toc">
<a href="#headline">Honest headline</a>
<a href="#anatomy">The artifact</a>
<a href="#mechanism">How it picks work</a>
<a href="#example">Worked example</a>
<a href="#apply">Apply to a sheet</a>
<a href="#limits">Where it breaks</a>
<a href="methodologies-comparison.html" style="color:var(--gold);font-weight:700">Comparison table →</a>
</nav>
<div class="finding" id="headline">
<h2>The honest headline: the strongest pre-build forcing function on the list</h2>
<p>Most "estimation" methods score features after the team has decided what to build. PR-FAQ inverts the order: <b>you write the customer-facing launch first</b>. If the press release reads as boring, irrelevant, or implausible, you re-shape the problem — long before you've sunk engineering effort. Ex-Amazon execs Colin Bryar & Bill Carr describe it (verbatim) as <em>"a systematic way to vet ideas and create new products. Its key tenet is to start by defining the customer experience, then iteratively work backwards from that point until the team achieves clarity of thought around what to build."</em></p>
<p>It's the only method on this site that produces a <b>readable artifact a non-PM can react to</b>. That's why it travels well — but also why it's hard: writing one good PR-FAQ takes days, not minutes.</p>
</div>
<!-- ANATOMY -->
<h2 class="sec" id="anatomy">The artifact — what a PR-FAQ actually contains</h2>
<p class="secsub">Two parts, both short. The strict format is the discipline: limited length forces clarity.</p>
<div class="doc">
<div class="pr">
<div class="label">PR — Press release (≤1 page)</div>
<h4>Reads exactly like a real announcement</h4>
<ul>
<li><b>Headline</b> — the product, in the language the customer would use.</li>
<li><b>Subheading</b> — who it's for and the key benefit.</li>
<li><b>Summary</b> — one paragraph the customer would understand.</li>
<li><b>Problem</b> — the specific pain being solved (in customer's words).</li>
<li><b>Solution</b> — how the product solves it.</li>
<li><b>Quote from a leader</b> — captures the strategic intent.</li>
<li><b>How to get started</b> — what the customer does next.</li>
<li><b>Customer quote</b> — a hypothetical real reaction.</li>
</ul>
</div>
<div class="faq" style="margin-top:14px">
<div class="label">FAQ — Internal & external (≤5 pages)</div>
<h4>The honest part — where edge cases get surfaced</h4>
<ul>
<li><b>External FAQ</b> — questions a customer or journalist would ask: <em>price, how it works, support, where to buy</em>.</li>
<li><b>Internal FAQ</b> — questions senior leaders and stakeholders would ask. The source lists: <em>competitors, problems solved, TAM (<a class="j" href="jargon.html#tam">Total Addressable Market</a>), challenges, capabilities needed, dependencies, regulatory issues, unit economics, investment required, payback timeline, success assumptions, failure risks</em>.</li>
<li><b>Success metrics live in the Internal FAQ</b> — explicit numeric outcomes ("we expect X% adoption in 6 months because…"). This is the line that makes PR-FAQ a pre-build impact estimator.</li>
</ul>
</div>
</div>
<div class="src">Source: <a class="cite" href="https://www.aboutamazon.com/news/workplace/an-insider-look-at-amazons-culture-and-processes">About Amazon — culture & processes</a> · <a class="cite" href="https://workingbackwards.com/concepts/working-backwards-pr-faq-process/">Working Backwards (Bryar & Carr) — PR-FAQ process</a>.</div>
<!-- MECHANISM -->
<h2 class="sec" id="mechanism">How a PR-FAQ actually picks the work</h2>
<p class="secsub">The mechanism is iterative and brutal: most first drafts get killed. Bezos's well-known rule: the PR-FAQ must be revised until the room can't poke a useful hole in it.</p>
<div class="step"><div class="num">1</div><div><h3>Pick the customer and the bad-day moment</h3><p>Start from a real customer in a specific situation. The PR will be written <em>for them</em>, not for the org chart.</p></div></div>
<div class="step"><div class="num">2</div><div><h3>Draft the PR</h3><p>Write the launch as if it ships tomorrow. If the headline is weak, the product is weak. The team often discovers in this step that the feature they wanted to build isn't compelling to the customer.</p></div></div>
<div class="step"><div class="num">3</div><div><h3>Write the FAQ — both halves</h3><p>External FAQ surfaces customer objections; Internal FAQ surfaces team disagreements and forces <b>numeric success metrics</b>. This is where "expected impact" becomes a written number.</p></div></div>
<div class="step"><div class="num">4</div><div><h3>Review in a silent-read meeting</h3><p>Amazon's distinctive ritual (verbatim from the source): <em>"The document is circulated at the start of the meeting, followed by 15–20 minutes of silent reading and note-taking."</em> No slides, no selling — the doc has to stand on its own. Most bets die here.</p></div></div>
<div class="step"><div class="num">5</div><div><h3>Iterate or kill</h3><p>If the doc survives critique, work begins from the PR backwards — what would have to be built to make the PR true? Many candidates iterate the PR for weeks before any code is written.</p></div></div>
<!-- EXAMPLE -->
<h2 class="sec" id="example">Worked example — the original Kindle PR-FAQ (the famous story)</h2>
<div class="ex">
<h4>1-page PR, written before any Kindle hardware existed</h4>
<p><b>Headline:</b> "Introducing Kindle — every book ever printed, in any language, all available in under 60 seconds."</p>
<p><b>Subheading:</b> Read books anywhere, instantly, without losing your library.</p>
<p><b>Key Internal-FAQ pressure tests it had to answer:</b> Why would a reader pay for a device? How does it work without Wi-Fi (→ became free 3G)? What if publishers won't sign? What does success look like in year 1?</p>
<p style="color:var(--ink-soft);font-size:13px">The Kindle PR-FAQ went through many iterations before the device was built. The team didn't just <em>design</em> the Kindle from the PR — they kept revising the PR until the customer story was undeniable.</p>
</div>
<p style="font-size:13px;color:var(--ink-soft)">Source for the Kindle PR-FAQ narrative: Bryar & Carr, <em>Working Backwards</em>, Chapter 5. The "every book ever printed... in under 60 seconds" line is Bezos's well-known Kindle vision — used here as a stand-in for the headline. The actual PR-FAQ text was never published by Amazon; the structure and discipline are the verifiable parts. AWS is a more documented example: Bryar & Carr describe the AWS team spending "more than a year writing, revising, debating, and discussing PR/FAQs" between identifying web services in 2004 and launching S3/EC2 in 2006.</p>
<!-- APPLY TO A SHEET -->
<h2 class="sec" id="apply">Apply to a feature sheet</h2>
<p class="secsub">PR-FAQ doesn't produce a numeric score either — it produces an <strong>artifact gate</strong>. Your sheet doesn't rank features; it tracks <em>readiness</em>. A feature does not enter build until every cell is green and the silent-read review passes. The columns are a checklist, not a calculation.</p>
<div class="note" style="background:var(--teal-soft);border-left-color:var(--teal)"><b>Try it Monday morning (30 minutes).</b> Pick your next "big bet" feature. Open a doc. Write only the <em>headline</em> and the <em>subheading</em> of its press release — as if it shipped tomorrow. Don't write the body. Just those two lines. Show them to one person who doesn't know the feature. If they can't tell you who it's for and why they'd care, the bet isn't shaped yet. That's a 30-minute version of the PR-FAQ pain — small enough to do this week, painful enough to surface the truth.</div>
<div class="extable">
<table class="ex">
<thead><tr><th>Column to add</th><th>What it captures</th><th>How you fill it</th></tr></thead>
<tbody>
<tr><td>Bet</td><td>Candidate feature / product</td><td>Backlog item — usually a "big bet," not routine work</td></tr>
<tr><td>PR drafted?</td><td>One-page mock press release exists</td><td>Y / N</td></tr>
<tr><td>Headline strong?</td><td>The launch reads compelling in the customer's own words</td><td>Y / re-write — subjective, decided by the silent-read room</td></tr>
<tr><td>Customer & bad-day moment</td><td>The specific user and situation the PR addresses</td><td>One sentence — "X user, when Y happens, today they Z"</td></tr>
<tr><td>Customer quote drafted?</td><td>Hypothetical real reaction to the launch</td><td>Y / N</td></tr>
<tr><td>Numeric success metric</td><td>"We expect X% Y in 6 months, because Z"</td><td>Internal-FAQ field — the line that makes PR-FAQ a pre-build estimator</td></tr>
<tr><td>Kill criteria</td><td>What would convince us to stop?</td><td>Internal-FAQ field — explicit thresholds</td></tr>
<tr><td>External-FAQ count</td><td>Customer/journalist objections answered (pricing, availability, limits, comparisons)</td><td>Integer (target ≥ 5–8 honest answers)</td></tr>
<tr><td>Silent-read passed?</td><td>Did the 20-min silent-read review survive critique?</td><td>Y / N / re-draft</td></tr>
<tr><td>Ready for build?</td><td>The gate</td><td>Ready (all green + silent-read passed) / Re-draft / Kill</td></tr>
</tbody>
</table>
</div>
<h3 style="font-family:Georgia,serif;font-size:18px;margin:24px 0 8px">Worked example — a PR-FAQ queue snapshot</h3>
<p style="font-size:13.5px;color:var(--ink-soft);margin:0 0 12px">Eight bets at various readiness stages, shaped after the Bryar & Carr template. The point is the <em>gate behaviour</em> — most first drafts re-draft or kill at the silent-read review; very few sail through to "Ready." That ratio is the discipline working as intended.</p>
<div class="extable">
<table class="ex">
<thead><tr><th>Bet</th><th>Headline</th><th>Numeric metric</th><th>Kill criteria</th><th>FAQ count</th><th>Silent-read</th><th>Verdict</th></tr></thead>
<tbody>
<tr class="top"><td>Standalone e-reader device</td><td>"Every book ever printed, in 60 seconds"</td><td>Units y1, % trade-book share by y3</td><td>Yes — publisher sign-on threshold</td><td>12</td><td>Passed</td><td class="score">Ready</td></tr>
<tr class="top"><td>In-store pickup lockers</td><td>"Skip the line — your order, in your locker, in your lobby"</td><td>Pickup% of orders, time-to-collect</td><td>Yes — utilisation floor 35%</td><td>8</td><td>Passed</td><td class="score">Ready</td></tr>
<tr class="top"><td>Free-shipping tier for sub-segment</td><td>"Two-day shipping on everything you buy this year"</td><td>Subscriber <a class="j" href="jargon.html#ltv">LTV</a> vs control, retention</td><td>Yes — subscriber margin floor</td><td>10</td><td>Passed</td><td class="score">Ready</td></tr>
<tr><td>Smart speaker for kids</td><td>Drafted but weak — "Voice that listens like a parent"</td><td>Partial — adoption stated, no retention floor</td><td>Missing</td><td>4</td><td>—</td><td class="score" style="color:var(--gold)">Re-draft</td></tr>
<tr><td>B2B subscribe-and-save</td><td>"Restock once. Never run out."</td><td>Yes — repeat-order frequency</td><td>Missing</td><td>6</td><td>Failed — legal flagged contract impact</td><td class="score" style="color:var(--gold)">Re-draft</td></tr>
<tr><td>Voice-ordering for delivery</td><td>"Order dinner by talking" — customer too generic</td><td>Voice-order share of all orders</td><td>Partial</td><td>5</td><td>—</td><td class="score" style="color:var(--gold)">Re-draft</td></tr>
<tr><td>AR fitting-room overlay</td><td>"Try things on virtually" — no customer-pain story</td><td>—</td><td>—</td><td>2</td><td>—</td><td class="score" style="color:var(--accent)">Kill</td></tr>
<tr><td>Crypto checkout</td><td>"Pay in crypto" — no customer pain</td><td>—</td><td>—</td><td>3</td><td>—</td><td class="score" style="color:var(--accent)">Kill</td></tr>
</tbody>
</table>
</div>
<div class="note" style="background:var(--accent-soft);border-left-color:var(--accent)"><b>The most important reading skill on this page.</b> Look at the four "Kill" and "Re-draft" rows. None of them died because the engineering was wrong — they died because the <em>customer story</em> was wrong. AR fitting-room overlay and Crypto checkout died on "no customer-pain story" — they never named who hurts and why. PR-FAQ's whole leverage is moving the death-of-bad-ideas from the post-launch retrospective (expensive) to the pre-build <a class="j" href="jargon.html#silent-read">silent-read meeting</a> (cheap). The ratio of Kill+Re-draft to Ready in this table is the framework working as designed.</div>
<div class="note"><b>Decision rule.</b> <b>Ready for build</b> only when <em>every</em> column is green <em>and</em> the silent-read room can't poke a useful hole in the doc. <b>Re-draft</b> if any field is missing or weak — go back to step 2, no engineering effort yet. <b>Kill</b> when the headline can't be made compelling after iteration; that's the PR-FAQ doing its job — most first drafts die at this gate, which is the point. PR-FAQ filters at the bet level; once a bet is built, individual changes still go through <a class="cite" href="microsoft-exp-framework.html">experiment-style A/B</a> readouts.</div>
<!-- LIMITS -->
<h2 class="sec" id="limits">Where PR-FAQ breaks</h2>
<div class="warn">
<b>It's expensive.</b> Days to weeks per artifact. Reserve PR-FAQs for <em>big bets</em>, not every small change. For routine work use RICE-style ranking.<br><br>
<b>It can ossify the product.</b> If the PR was wrong, the team is anchored to it. Mitigation: explicit "kill criteria" in the Internal FAQ — what would convince us to stop?<br><br>
<b>It is <em>not</em> a substitute for experimentation.</b> The PR-FAQ estimates expected impact; <a class="cite" href="microsoft-exp-framework.html">controlled experiments</a> measure actual. Amazon does both — PR-FAQ at the bet level, A/B at the change level.
</div>
<div class="note"><b>Why PR-FAQ is on the list.</b> It's the only Consumer-side framework that produces an <em>up-front written estimate of customer value</em> rather than relying on post-launch measurement. The closest cousin in the SaaS set is <a class="cite" href="stripe-shaping-framework.html">Stripe's deep-dive memo</a>, but PR-FAQ adds the strict format that makes it teachable.</div>
<footer>
Companion to <a href="impact-consumer-companies.html#define">← Consumer case studies · Define impact first</a> · <a href="methodologies-comparison.html">All methods compared</a> · contrast: <a href="stripe-shaping-framework.html">Stripe shaping</a> (the SaaS-side cousin)<br>
<b>Grounded in</b> Bryar & Carr's <em>Working Backwards: Insights, Stories, and Secrets from Inside Amazon</em> (HarperBusiness, 2021) and the official <a href="https://workingbackwards.com/concepts/working-backwards-pr-faq-process/">workingbackwards.com</a> process page. Practice originated at Amazon in 2004. <b>Verbatim from the source:</b> the Working Backwards definition ("a systematic way to vet ideas..."), the silent-read meeting framing ("15–20 minutes of silent reading and note-taking"), and the Internal-FAQ topic list (competitors, problems solved, TAM, dependencies, regulatory issues, unit economics, payback, success assumptions, failure risks). <b>Paraphrased from public retellings:</b> the Kindle headline ("every book ever printed... in under 60 seconds" is Bezos's well-known Kindle vision used as a stand-in — Amazon has not published the actual Kindle PR-FAQ). <b>Added by us, not in the source:</b> the readiness-checklist sheet columns, the eight-row bet-queue worked example, the kill/re-draft/ready verdict logic.
</footer>
</div>
</body>
</html>