Guide Writing

Guide writing (Write Useful Books method)

When to use

Use when structuring long-form educational content as a product with a promise: scoping it, ordering its contents, testing it on real readers, and planning its launch. This skill covers structure and product design of the content — voice and prose style are a separate concern, out of scope here. Only for problem-solvers (content promising an outcome); if the work is narrative, artistic, or entertainment-first, this method is explicitly the wrong advice.

The procedure

Phase 1 — Scope the promise (before writing anything)

1. Write the scope in one block: Promise + Reader profile + Who it's NOT for + What it won't cover. The promise is what's different in the reader's life afterwards, stated clearly enough that the wrong reader can decide they don't need it.
2. Sharpen by deletion, not addition. Ask: "What does my ideal reader already know and believe?" — everything they already believe gets cut. This kills introductory throat-clearing.
3. Write the recommendation loop as a literal story: one stranger, in a specific moment of pain, complains to someone — who names your thing. If no plausible story exists, adjust the promise or reader until one does. No plausible story = the growth plan is "hope."
4. Check longevity: is the promise relevant 5+ years out? Move datable tactics (tools, prices, platform mechanics) into an updatable channel (web page, appendix, changelog), not the core artifact.
5. Validate desirability by talking, not pitching: run listening conversations about the reader's life — what they tried, where they searched, what blocked them. Never ask for opinions ("asking for opinions is just fishing for compliments"). If you can't find even one interested reader now, you won't find thousands later — fix the scope today.

Phase 2 — Design the ToC for the reader

1. Every title is a takeaway, not a topic: "Good sales is about asking good questions", not "Sales 101". If a section has multiple learning outcomes, keep the category title and add takeaway subsections.
2. Test the structure by teaching it live before drafting: walk a target reader through one or two chapters' worth of value in conversation. Every improvisation, repeat, or rephrase is a ToC bug — fix the structure, then teach again.
3. Order content around the learner's goals, not the teacher's convenience. Theory is a value enabler, not value: interleave small doses of it between early wins instead of front-loading it ("let's get the theory out of the way" is the classic reader-experience killer).

Phase 3 — DEEP quality loop (draft and revise)

1. Brain-dump a messy first draft following the tested ToC. No rereading, no polishing. Write the introduction last.
2. Annotate the ToC with word counts per takeaway (250 words ~ 1 reader-minute). Scan for the three defects: slow start (how many words before the first big value?), long slogs (back-to-back sections with no a-ha), fluff (high words relative to takeaway).
3. Fix all three by rearranging, editing, and deleting — never adding. Halve the words, double the value. Keep a cutting-room-floor doc; deletions become future marketing content.
4. Front-load: cut front matter, insert real value early, start with the big reveal. Readers abandon at the start, and recommendation likelihood scales with value received before abandoning.
5. Revise for structure and clarity only — tighten the early sections hardest. Stop when the draft is coherent enough for a motivated reader to muddle through. Do NOT polish yet: polish now gets torn up, and a polished draft suppresses honest feedback.

Phase 4 — Beta reading (instrumented usability testing)

1. Recruit real target readers — never friends or paid professionals; obligation destroys the abandonment signal, which is the most valuable data.
2. Before inviting anyone: put the draft in a live-commenting tool, and add explicit instructions on what feedback you want (otherwise you get typo reports while testing core content).
3. Run 2-8 week iterations: first 1-2 weeks gather feedback, the rest revises. Invite 4x the readers you need (12-20 invited yields 3-5 engaged). Fresh readers every iteration — veterans start commenting on tone and typos instead of value.
4. Read three data streams: qualitative (what comments say — amplify what readers unexpectedly love), quantitative (where comments stop = where they got bored or quit — that's the next thing to fix), observational (do they apply it afterwards? Follow up weeks later; "sounded good in theory but didn't work" kills recommendability).
5. Fix boredom with deletion: chainsaw the surrounding sections' word counts by half or more. Never add words to cure boredom.
6. Feedback hygiene: each draft is an experiment — "the problem is the problem," not you. Never argue in comment threads. Confusion is your defect even when you're technically correct. Tangential drama-magnet sentences: clarify if you can, delete if the point isn't educationally essential.
7. Save the most influential potential readers/testimonials for a late, strong iteration — most people read a draft only once.
8. Iterate until the three finished signals fire (see quick reference), tempered by your real deadlines and energy.

Phase 5 — Seeding (launch is a year, not a day)

1. Manually acquire the first 500-1,000 genuinely-served readers. One channel is enough — pick by your assets: audience-building via writing in public (most reliable, compounding), event/bulk giveaways to exact-fit readers (fastest with contacts), a podcast/interview tour (most scalable), or paid ads (easiest, unscalable).
2. The manuscript is the marketing: pillar content becomes micro content. Share drafts, excerpts, research, process, and milestones on a schedule — consistency needs a system, because publishing early work is emotionally expensive.
3. Capture email over social follows (orders of magnitude more value per subscriber); a lead magnet or a plain "email me if you want it" both work.
4. Organic growth is earned by the product and ignited by the seed. If the seed doesn't catch, the defect is upstream (DEEP), not in marketing volume.

Rules and quick reference

• DEEP, in dominance order: Desirable → Effective → Engaging → Polished. Earlier dominates later — nobody forgives useless-but-beautiful. Effectiveness is the silent failure mode ("pseudoteaching": flawless delivery, zero behavior change).
• Scope formula: Promise + Reader profile + Who it's NOT for + What it won't cover. Sharpen with "what does my ideal reader already know and believe?"
• Recommendation loop: written as a story you can stress-test. "Nobody recommends the second-best solution — best not for everyone, but for someone."
• Value math: 250 words ~ 1 reader-minute; value = takeaways over time spent; halve the words, double the value; "deletion is your savior."
• Beta numbers: invite 4x → 3-5 engaged per round; 2-8 week iterations; at least 2 rounds (fewer if you've taught the material a lot); fresh readers each round; where comments stop = abandonment point.
• Finished signals (all three = ship): new beta readers are easy to recruit (Desirable) + most reach the end (Effective + Engaging) + some bring friends (loop is running).
• Seed bar: 500-1,000 manually-earned readers, or commit to a launch year.
• Feedback: never argue; each draft is an experiment; "the problem is the problem"; one-star reviews from wrong-fit readers mean "it's not for them."

What generalizes vs book-specific

Generalizes to newsletters, guides, courses, docs, video series, internal playbooks: the scope formula and deletion question; DEEP with dominance ordering; the recommendation loop as a design object; test-by-teaching before building; takeaway-titled outlines with word counts; value-per-page discipline and front-loading; beta reading as instrumented usability testing; manuscript-as-marketing; seed-then-organic; timelessness as an asset-design split; feedback hygiene.

Book-specific (2021 book market — treat numbers as illustrative, skip unless publishing a book): royalty splits and the self-publish-first-10k hybrid; Amazon machinery (thumbnail covers, 20-40 review tipping point, five-consecutive-days trust flag, PPC economics, $9.99/70% royalty cliff); audiobook/translation percentage boosts; $1.99 Kindle launch floors. Also keep guard up: the 800-seeded → 50,000-organic flywheel is one book's trajectory in an unusually recommendation-dense niche, and the method is quietly expensive in calendar time and social energy.

Source

Compiled from Write Useful Books — Rob Fitzpatrick (2021). The skill is the procedure; the book carries the depth (worked examples, edge cases, the author's reasoning). If this stage is where your venture lives right now, buy and read it.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: lucasheriques
• Source: lucasheriques/shipmate
• License: MIT

Install and usage instructions live in the source repository linked above.