Flower and Bees
The Coldcode Si Labs Library voice guide: how every page, product and message should feel
COURAGE GARIKAI
Version 2.0 · Second edition · November 2026
Flower and Bees · The Coldcode Si Labs Library voice guide · Version 2.0. Second edition, November 2026. First edition published November 2026. © 2026 Courage Garikai. All rights reserved. No part of this book may be reproduced without written permission, except the templates and checklists in Part Eight, which are meant to be copied. Set in a Times-based text face at 11.8 on 17.6 points. Trim 6 × 9 inches. ISBN: to be assigned. Published by Coldcode Si Labs Library, Harare.
Hype is invented feeling. Emotion is real feeling, named accurately. Almost everything in this book follows from that one distinction.
A note on what this edition is
The first edition of this book was written while the studio was called something else, and its worked examples were the six ventures that studio ran. The method did not change when the name did. The examples had to.
This edition keeps every chapter and every rule, and replaces the worked material: Part Six now walks through six books and papers from the Coldcode Si Labs Library, the toolkit’s promise Codex and venture map are rebuilt around what a press actually promises a reader, and the front matter is the press’s own. Where the first edition named a venture, this one names a title. Where it described a feature, this one describes a page of a book — which is the harder test, because a book cannot demo anything. Its proof is the writing on the page you are already reading.
One thing carries over unchanged, and it is the part worth protecting: the rule that no example is invented to win an argument. Every quoted line of reader complaint in Part Six is constructed, and it is constructed the way the book asks you to construct one — from the shape of a sentence people actually say, with no number attached that nobody counted. The books’ own figures are the figures in their manifests, and those are checkable.
A note on the examples in this book
Every example of bad copy and good copy in these pages is constructed. None of them quotes a customer. None of them describes a real person, a real business, a real price or a real result. They were written to show a failure and its repair, in the way a carpentry book shows a bad joint and a good one. This matters, because the book argues at length that you may never invent a customer, a testimonial, a number of users or a result. It would be a poor beginning to break that rule on page four in order to explain it. Where an example is drawn from the author’s own work, it is named as such and no figure is attached to it that cannot be checked. Where it is invented, it is marked. The invented ones are the majority, and the reader should assume so throughout, on every page, without having to look for a label. The book is therefore a set of instruments rather than a set of case studies. A case study tells you what happened to somebody else. An instrument lets you find out what is happening to you.
How to read this book
This is a manual, and manuals are read in the order the reader needs them. There are three reasonable orders and one unreasonable one. The first order is straight through. About five hours of reading, at the pace of a person who is thinking while reading rather than skimming. It builds: the argument first, then the five-part structure of a page, then the rules, then the surfaces where pages are actually made, then the research that makes the writing true, then the six products, then the upkeep. Read this way, the book is an education. The second order is by surface. If you have a landing page to write this week, read Part Four and Part Two and stop. The page will be better by Thursday. Come back for the rest when the deadline is not standing over you. The third order is by audit. If you suspect that your existing pages are quietly costing you something and you want to find out where, read the toolkit at the back first. It is twelve pages of checklists, templates and a lexicon of banned words. Then work backwards from whatever fails. The unreasonable order is to read it as encouragement. This book is not here to tell you that your marketing is fine. It is here to tell you, calmly, that a good deal of it is doing the opposite of what you intended, and that the repair is mostly subtraction.
The one tension a careful reader will notice
The rule at the centre of this book is that no message should be sendable, unchanged, to two people. Each one is written for a single recipient, and contains something only that recipient would recognise. This book is one message sent to many thousands of people. I am not going to pretend the contradiction is not there, because you would have found it by the second chapter and then wondered what else I had glossed over. So here is the resolution, and it is not a clever one. It is simply the truth about what you are holding. A message is an interruption. It arrives in your inbox, or on your screen, unbidden. You did not ask for it, you do not know how it found you, and it is asking for something before it has given you anything. For an interruption, the rule holds absolutely: speak to one person, or do not speak. A book is not an interruption. You picked it up. You opened it. You are reading it on purpose, because you have a problem with the words your organisation publishes and you want a method. That is closer to a workshop than to an email. So the rule changes shape rather than disappearing. A manual addressed to many readers must be written so that each reader can find themselves in it, on every page, in their own situation. Not “many businesses struggle with urgency” but “you have a countdown timer on your offer page and you know, in the quiet part of the evening, that it resets when nobody is looking.” If a page of this book could only ever have been written about somebody else’s company, that page is wrong and should have been cut. There is one more thing in the same spirit. The rules in this book govern what you publish for customers. They do not govern this book’s grammar. The book is allowed longer sentences, scenes, hesitation and doubt, because a book that never doubts itself is a book that is selling. But the book is not allowed to sound like it is selling itself. No urgency. No countdown. No promises about what will happen to your revenue if you finish it. You would be right to stop reading.
Contents
The flower and the bee
The mistake in version one
Hype is invented feeling. Emotion is real feeling, named accurately.
Where the metaphor holds, and where it breaks
The five petals
Recognition: does this person understand my day?
The cost: why does this matter now?
The after: what would it feel like?
The proof: is this real?
The small door: what do I do now, and what if it does not work?
Order, and the four ways to break it
The nine rules
Rules of truth
Rules of warmth
Rules of the door
The surfaces
The landing page
The product card
The first-touch email
Follow-up, replies, and knowing when to stop
The sales call
A page from nothing
Finding the feeling
Collecting the buyer’s words without inventing them
One feeling per page, and the four you cut
The tired-stranger test
The reader who is wrong for you
Six flowers
The Ledger and the Shopkeeper: recognition, because the failure has a name
Intelligence, Data and Judgement: earned confidence, because the material is teachable
Order Without Custody: order, because the money is never held
Grounded Cadence: calm control, because the rhythm is chosen now
The Sovereign Micro-Enterprise: sovereignty, because the rail is yours
The working papers: capability, because you can run the instrument yourself
Keeping a flower open
When three people and a system all write
Coldread, turned on your own words
The toolkit
The banned lexicon, and what to write instead
The nine rules, on one card
The 30-second check
The seven templates
The promise Codex
The collection sheet
The message audit worksheet
The venture map
The blank page
Hype is invented feeling. Emotion is real feeling, named accurately. Almost everything in this book follows from that one distinction.
The first version of the Codex got one thing wrong, and it got it wrong in the way that careful people usually get things wrong: by over-correcting. The problem it was built to solve was real. We were writing like everybody else. Every page we published had a superlative in the headline and a claim in the sub-line that nobody could check. We had a page that said a product was powerful, fast and easy to use, which meant nothing at all, and underneath it three buttons, all of them asking for the same commitment. We had an email sequence that opened with “we’re excited to announce” and closed, seven messages later, with a discount that expired in forty-eight hours and then did not expire, because the timer reset at midnight and everybody in the building knew it. So we wrote a rule set. And because the problem was hype, the rule set was assembled to remove hype. Out came the superlatives. Out came the exclamation marks. Out came the urgency, the enthusiasm, the warmth, the adjective “delighted”, the word “love”. Out came anything that could be read as a claim. What was left was accurate, compliant, checkable, and dead. I remember reading the first page we produced under the new rules and feeling a kind of pride, because it was clean. It was also the page of a company that did not want to speak to anybody. A reader arriving at it learned exactly what the product did, exactly what it cost, and exactly nothing about why they should care — and they left, quietly, in the way people leave when nothing has been done wrong to them. The lesson took a while to arrive and it arrived in the form of a sentence from a customer, who said, of the new site: “It’s honest. I just don’t think it was written by anyone.” Hype and emotion are not the same substance, and removing one does not remove the other. They sit next to each other, they sound similar in a sentence, and they are opposites in what they do to a reader. Hype is invented feeling. It is a feeling that belongs to nobody, generated in the writer’s room to move a reader who would not otherwise move. The countdown that resets. The “thousands of businesses trust us” with no number and no way to count. The testimonial that nobody said. The word “revolutionary” applied to a spreadsheet. Hype fails not because it is enthusiastic but because it is untrue, and readers have developed a reflex for it. That reflex is not admiration. It is a small closing. Emotion is real feeling, named accurately. It is the tired team on a Sunday night, the margin that disappeared between the quote and the invoice, the relief when the shop’s books balance after three months of not knowing. It is not softer than hype. It is far more specific, far harder to write, and it is the only thing on a page that a reader will believe, because they have felt it themselves and can check.
The difference is testable, which is the useful part. Ask of any line: is this feeling mine, or is it the writer’s invention? If a person in the situation described would recognise the feeling and use roughly those words for it, it is emotion, and it is allowed. If the feeling was added to make the reader do something, it is hype, and it must go. Here is the same product sold twice. Invented feeling. Take control of your logistics and watch your business soar. The reader is asked to feel empowered and adventurous. No reader has ever felt either thing about a container. Real feeling, named accurately. Your container sat two days longer than it needed to, and nobody knew until the customer asked. That is a specific Tuesday. Every person who has shipped anything recognises it before the sentence ends. It is more emotionally powerful than the version above by roughly the width of the sky, and it contains no adjective at all. This is the whole argument and it takes a paragraph. The rest of this book is the working out: how to find the real feeling, how to choose one, how to structure a page so that it arrives in the right order, how to prove it without overclaiming, and how to keep doing it when three people and a scheduling system are all publishing in your name. One more thing belongs in this first chapter, because it is the mistake that follows the first mistake. When you remove hype, you will be tempted to remove feeling. Every organisation that has ever cleaned up its language has done this, and it always feels like virtue while it is happening. The pages become accurate. The claims become defensible. The legal review gets shorter. And then nobody reads it, and the team concludes that honesty does not sell, which is the most expensive wrong conclusion in marketing. Honesty sells. Flatness does not. The two are not related. Exercise. Take the worst page your organisation currently publishes — the one you would not show a friend. Count three things: the number of superlatives, the number of claims, and the number of sentences that describe a specific moment in a customer’s day. If the third number is zero, the page has no feeling in it, and no amount of cutting will fix that. Feeling has to be added back, not preserved.
The distinction is easy to state and it only becomes useful once you can diagnose a line in the wild. So this chapter is mostly diagnosis: what hype looks like when it is wearing ordinary clothes, why readers detect it faster than writers think, and what to put in its place.
A reader who arrives at your page has read ten thousand pages like it. They have developed a skill you do not have, because you only ever read pages in your own category, and they read pages in every category, all day, while waiting for something else. That skill is a fast filter for unearned certainty. Unearned certainty has a signature. It is smooth. It contains no dates, no names, no numbers, no costs, no failures, no weather. It could have been written by anyone about anything, and it is written as though the reader’s agreement were already secured. Your page is being read by someone who has been lied to by a landing page roughly every day for fifteen years. They are not cynical. They are practised. The filter runs before they have decided to read, which is why “we’re excited to announce” is not a weak opening but a closed one.
The writer’s mistake is assuming that the filter is a defence against exaggeration. It is not. It is a defence against absence of evidence. That is why honesty works and why flatness fails: honesty gives the filter something to hold, and flatness gives it nothing, so it reads as evasion. A page that says only defensible things and never shows a moment will be filtered exactly like a page that exaggerates, because both are refusing to be specific.
Hype rarely announces itself. It dresses as practicality, as professionalism, as friendliness. These are the five costumes to look for. One: the adjective stack. Three adjectives in a row is not emphasis, it is a shrug. Fast, reliable, easy-to-use. If any of the three were removed, the sentence would mean the same thing, which means none of them were doing work. The repair is to replace all three with the one measurable consequence: the invoice exports in one click to the format your accountant already uses. Two: the borrowed plurality. Businesses everywhere are discovering… There is no list. There is no count. If there were a count, it would be in the sentence, because a real count is the most persuasive thing available. The repair is either to produce the number or to speak in the singular: one shop in Avondale discovered, in March, that… — and if that one shop does not exist, the sentence does not exist either. Three: the future tense used as evidence. This will transform how your team works. Nothing has happened. The sentence has moved from a claim about the present, which could be checked, into a promise about the future, which cannot. The repair is to describe the change in the past tense of somebody’s actual week, or to describe what the product does mechanically and let the reader infer the future themselves. Four: the urgency you control. A countdown. A banner that says the price rises on Friday. A limited allocation that is limited by a database field the marketing team can edit. The tell is agency: who decided this would end? If the answer is somebody on your side with a keyboard, it is not urgency, it is pressure, and the reader can feel the difference even if they cannot articulate it. Five: the flattering address. Dear valued customer. We know how hard you work. Praise that costs nothing and applies to everyone is read as a technique, because it is one. The repair is warmth without flattery: speak to one person, about one thing, in plain words, and skip the preamble entirely.
For any line of copy, ask: if I removed the feeling, would the fact still be true; and if I removed the fact, would the feeling still be mine? Hype fails the second half. Remove the fact from our revolutionary platform empowers teams, and there is no residue of your own experience; there is only a mood the writer manufactured. Emotion passes it. Remove the fact from your container sat two days longer than it needed to, and you are left with the specific irritation of not knowing, which is a feeling many people have had and none of them were given.
That residue is the whole difference. A page is honest when every feeling on it could survive having its claims taken away.
These are the four lines I use most often, in the order they usually find something. The two-recipient test. Could this be sent, unchanged, to a different person? If yes, it is not finished. This applies to emails most obviously, but it is ruthless on landing pages too. A hero section that could belong to any of your six competitors fails it. The substitution test. Replace your product name with a competitor’s. Does the page still make sense? If it does, the page is about a category, not about you, and the reader has no reason to choose. Categories do not create preference. The Tuesday test. Find the sentence that describes a moment. If the moment has a day of the week, an hour, a place, or a person, it is real. If it happens “in today’s fast-moving environment”, it happens nowhere. The grandmother-and-accountant test. Read the page to somebody who does not work in your industry. If they cannot say, in one sentence, what would be different about their day, the page has failed, regardless of how clean it is.
Not restraint. Specificity. This is the practical heart of the matter, and it is where most writing advice goes soft, so let us be concrete about the mechanics.
Specificity is built from four materials: — Nouns you can see. A container, a tender notice, a bank statement, a Monday morning, a shop counter, a purchase order, a renewal date. Not “solutions”, not “workflows”, not “ecosystems”. A noun you could photograph is a noun a reader can stand inside. — Numbers that can be checked. Not “significant savings” but “eleven hours a month”, and a link to where the eleven comes from. If it cannot be checked, it is a decoration, and decorations are the first thing a practised reader removes. — Time. When. Which day, which week, which season, which deadline. “Found four days before it closes” is a scene. “Discovered too late” is a summary. — Consequence in ordinary currency. Money, sleep, pride, weekends, trust. Not “efficiency”. Efficiency is what consultants call it. What the buyer calls it is that they got home at eight. The discipline is to write the sentence with all four present and then decide which three to cut, because the fourth will make the sentence too heavy. Specificity is not accumulation. It is choosing the one detail that carries the rest.
Take this line, which is exactly the kind of sentence produced by a competent marketing team on a Tuesday:
Our invoice automation solution helps small businesses streamline their billing and get paid faster.
It is not a lie. It contains no banned word. It is also worth nothing, and here is the diagnosis in four steps.
Step one: find the feeling. There isn’t one. “Streamline” is not a feeling. Somewhere behind this sentence is a real one — the specific dread of opening the banking app on a Friday and finding that two customers who promised have not paid, and not being able to chase them without feeling like a beggar. The writer knew this and wrote “get paid faster” instead, because the feeling felt unprofessional. Step two: state the cost in the buyer’s currency. Not time saved. The Thursday evening spent deciding whether to send a third reminder to a customer you like. That is a real cost and it is paid in pride. Step three: find the proof. What does the product actually do, mechanically, that a reader could observe? It sends the reminder on the day the invoice becomes overdue, with the amount and the reference already filled in, so nobody has to compose an awkward paragraph. That is checkable by looking at the product, which is what proof means. Step four: rebuild in the order of the five petals — which is Part Two, and which is why the repair cannot be a rewrite of one sentence. The sentence was failing because the page had no structure underneath it. The repaired version, reading as a section heading and three lines:
The Thursday you spend deciding whether to send a third reminder.
The invoice went out on the first. It was due on the fifteenth. It is now the twenty-fourth and you have written the same polite paragraph twice. The reminder now goes on the fifteenth, and again on the twenty-second, with the amount, the reference and the bank details already in it. You read it before it sends. You do not have to compose it.
Same product. Same price. No adjectives. The reader is now in a specific Thursday, and they know whether they are in it. That is the entire craft in miniature, and it is not a trick of style. It is an act of attention. Somebody had to notice that the real problem was not speed but the awkwardness of the third reminder. That noticing is the work. Everything in the rest of this book is a way of systematising the noticing. Exercise. Take three lines from your own current copy that contain no number, no date and no specific noun. For each, write down the moment in a customer’s week that the line is avoiding. Do not write new copy yet. Just write the moment. Most of the difficulty of this work is refusal to look at the moment, and the writing after that is comparatively easy.
The flower does not chase the bee. It opens, it is honest about what it holds, and it is easy to find. That sentence is the front door of this book, so it is worth being exact about what it does and does not claim, because a metaphor that is defended too hard becomes a religion, and religions make poor manuals.
It holds on the direction of effort. Most marketing effort goes into chase: list growth, sequence length, retargeting, frequency. The flower’s effort goes into being worth arriving at — opening, colour, scent, nectar, and being findable. These are different budgets. A company that spends its energy on chase and nothing on the flower is buying traffic to a closed thing, which is the exact complaint the library’s first book was written to name: we are paying for attention and nobody reads. It holds on honesty about contents. A flower that promised nectar and held none would be visited once. A page that promises relief and delivers a feature list is visited once, too, and the second visit never happens because the bee’s memory is longer than the vendor’s spreadsheet. “Honest about what it holds” is not a moral position only. It is a durability position.
It holds on ease of finding. Bees do not search. They forage, following signals they did not have to decode. So “easy to find” is not a search-engine optimisation note. It is a note about legibility: can a person in the situation recognise, in the first two lines, that this is for them? A page that requires the reader to work out whether they are the audience has already lost the bee in the tall grass. It holds on the asymmetry of need. The flower needs the bee more than the bee needs this flower. Anybody who has run a small business will recognise that asymmetry instantly, and recognising it is what produces the calm the book keeps asking for. A vendor who believes they are doing the customer a favour writes like a vendor. A vendor who knows they are one flower among many writes like a person who is glad you came.
A flower cannot choose its bee. You can. The metaphor suggests passivity, and passivity is a real risk here: “open and honest and they will come” is only half true. The flower’s advantage is that it is one of thousands and still findable, because colour and scent are signals that work at distance. Your equivalent is not waiting. It is being unmistakably identifiable in the two seconds before the reader decides. Being findable is an active craft — how you are found, what you are called, where the notice appears, who carries the recommendation — and none of it is passive. A flower has one nectar. You have a menu. This is the largest break and it needs managing constantly. A shop selling six things cannot have one scent. What it can have is one feeling per surface, which is not the same as one feeling per company. The flower metaphor works at the level of a single page. At the level of a catalogue, the honest metaphor is not a meadow but a market stall with clearly separated trays and a person standing behind it saying “that one’s for you”. A flower’s honesty is free. Yours has a price. The metaphor makes honesty look costless. It is not. Honesty costs you the campaign you wanted to run, the comparison table you would have won, the testimonial you would have written, and sometimes the customer who would have bought the overstated version and been disappointed. That cost is real and it is paid by somebody in the business every month, which is why the rules later in this book need a governance section and not just an ethic. A season ends. Your obligation does not. Flowers are permitted to be seasonal. A product that goes quiet after launch is not a flower resting; it is a shop with the lights off. The metaphor gives no permission to stop opening.
Because it answers the question that every voice guide leaves unanswered, which is: what am I actually trying to do when I write this page? The answers people usually operate under are all wrong in the same direction. Sell harder. Explain more. Cover the objections. Rank for something. Each of those is a task, not a purpose, and a task with no purpose behind it produces the smooth, evidence-free prose of the previous chapter.
The flower’s purpose is a single, checkable outcome: that a person who needed exactly this found it, understood what it holds, and could take the next step without risk. Everything in this book is a means to that end, and every rule can be tested against it. If a sentence makes the flower easier to find, it stays. If it makes the flower easier to find while lying about the nectar, it goes. If it does neither, it was decoration and it should never have been written. Exercise. Write your own version of the epigraph, in one sentence, for one specific page you are currently working on. Not for the company. For that page. It should name what you hold, who it is for, and how they will find it. If you cannot write it in one sentence, the page has no purpose and you are about to fill it with features.
Every page and every message has all five petals. They are not a menu. If one is missing, the reader supplies it themselves, and they usually supply a reason to leave.
Every page begins with one question in the reader’s mind, and it is not “what does this do”. It is ”is this for me?” Recognition is the petal that answers it. It shows the reader their own situation, specifically enough that they believe the writer has been in it. Nothing on the page matters until this has happened, because a reader who has not recognised themselves has no reason to read the next line, and will not. The test is unforgiving and simple. Cover the rest of the page. Read only the first two lines. Does a person in the situation see their own day? Not their industry. Not their role. Their day — the specific hours, in the specific place, with the specific thing that went wrong.
Organisations put themselves first out of a reasonable instinct: the reader has arrived, so we should explain who we are. But the reader did not arrive to meet you. They arrived with a problem, at a moment, and the only currency you have in the first two seconds is whether you can describe that moment better than they can describe it themselves. There is a mechanical reason too, which is worth knowing because it explains why the order is a rule rather than a preference. Recognition is what makes the rest of the page legible. A proof point that lands on a reader who has not recognised themselves reads as a boast. The same proof point, arriving after recognition, reads as evidence. The facts do not change. Their meaning does, entirely, because of what came before them.
You cannot invent recognition, and this is the petal where invention is most tempting and most obvious. The material exists already, in four places. The words people use when they are not being polite. Support tickets. The first thirty seconds of a sales call. Churn interviews, where people are unusually honest because they have nothing to gain. The message a customer sends at eleven at night. Record these verbatim, and keep the sentences that have an odd specificity to them, because those are the ones that carry recognition. The sequence of the bad day. Not the problem in the abstract but the order in which it happened. Somebody opened the banking app, then the spreadsheet, then the phone. That sequence is a scene, and a scene is what recognition is made of. The thing they had to stop doing. Recognition is often strongest when it names an activity that ceased. We stopped going to the Monday market. We stopped answering the phone after five. We stopped quoting for small jobs because the quoting took longer than the job. Loss of a habit is more recognisable than presence of a problem.
The sentence they apologise for. People say “I know this is a stupid question” before the most useful questions. The apology marks the feeling. Write down what they apologised for and you have a recognition line with almost no work left to do.
It is short. It is in the present tense or a named past. It has a place and a time. It does not contain the word “businesses”. It does not explain why the situation is bad, because the reader already knows it is bad — that is why they are here. Here is the failure and the repair, on a page for a bookkeeping product. Failure. Running a small business is challenging, and managing your finances can be overwhelming. Two things are wrong and the second is worse. The first is that it is generic: every reader’s own description of their situation is more specific than this. The second is that it is about a category, and categories do not have Tuesdays. Nobody has ever thought “yes, that is my challenge, management of finances.” Repair. You know what the shop took this week. You do not know whether it was enough, because the wholesaler’s bill came out of a different account and you have not had time to sit down with both. That is recognition. The reader is in a shop, at the end of a week, holding two facts that will not join. There is no adjective and no claim, and the page has not yet said what it sells.
It over-reaches into invented intimacy. We know exactly how it feels when the family asks about the business. You do not know. Nobody told you. The reader can hear the difference between a described situation and a claimed relationship, and the claimed one closes the flower immediately. The rule is to describe, never to presume: “you find the notice four days before it closes” is true of a class of people and is therefore honest. “We know how much this hurts your family” is a claim about a stranger. It lands on the wrong person. This is common on products with two audiences, where the buyer and the user are different people. A bookkeeping product bought by a shop owner but used by their bookkeeper needs one recognition line for the shop floor and a different one for the back office. One line cannot serve both. The “sound familiar?” pattern from Part Four exists precisely to give each audience its own card.
Recognition is not only the first line. A page can carry a second, quieter recognition further down — usually a list of three or four situations, each one line, each one checkable by the reader against their own week. The purpose is not to cover every case. It is to let the reader find the one that is theirs and skip the other three, which is a kindness and also a filter: the reader who finds none of them should leave, and it is much better for everybody if they leave early than if they buy and are disappointed.
Exercise. Write five recognition lines for one product, each describing a different moment, none longer than twenty-five words, none containing an adjective of praise. Then show them to a person who has the problem and ask only one question: which of these is not you? Delete that one and keep the rest.
Once the reader has recognised their situation, the next thing they need is a reason to act on it today rather than in the vague future when things are quieter. That reason is the cost. The cost is what the problem takes from them, stated in the currency they actually pay in. There are five currencies and almost everything is paid in one of them: — Money. Named, sized, and traceable to something they will recognise — a month’s rent, a margin that went from nineteen points to four, a subscription they forgot. — Sleep. The two hours at three in the morning spent rebuilding a spreadsheet in your head. — Pride. The moment you have to tell a customer you made a mistake, or chase a friend for money, or ask your father for a loan. — Weekends. The specific Sunday afternoon, every week, that belongs to admin. — Trust. The version of you your team now expects — the one who said the deadline was fine.
Cost is the petal most likely to curdle. State a cost calmly and the reader feels understood. State it a second time and they feel accused. State it with a consequence appended and you have written a threat, which is hype in a dark suit.
So the discipline is: name the cost once, in their words, and do not add a consequence. You do not need to say “and if you don’t fix it, you’ll lose customers”. They know. That sentence is not information, it is pressure, and pressure is the thing this whole book is built to remove. Compare. Threat. Every day you delay, you are losing money to competitors who have already automated this. Act now before it’s too late. Cost, named once. The quote was right. Then the exchange rate moved and the shipping was quoted in a different month, and by the time the invoice went out the margin was four points instead of nineteen. The second version contains no instruction at all, and it produces more urgency than the first, because it is a specific loss the reader can price themselves. That is the trade this book keeps making: give up the pressure, gain the arithmetic. When a reader does the maths in their own head, it belongs to them. When you do it for them in capital letters, it belongs to you, and they discount it accordingly.
There is a version of this petal that is not a loss at all, and it is the strongest one available to you: urgency that already exists in the world, which you merely report. A tender closes on the fourteenth. A tax deadline is in six weeks. A renewal date is in the contract. A season ends. A licence expires. A scheme’s funding runs out on a date somebody else chose. These are real, checkable, and belong to the reader’s calendar, not to your marketing. That is the whole rule, and it is the one that separates a page a reader trusts from a page they ignore:
Urgency must be theirs, not yours. A tender deadline, a renewal date or a season is real urgency. A countdown timer you control is not.
The countdown timer is the most common lie in software marketing and the least necessary. It works once. It also teaches the reader that your deadlines are decorative, which means the next deadline you have, the real one, will be discounted too. You are trading a permanent asset — credibility about time — for a temporary bump, every time you reset the clock.
Where a number is used, three conditions apply. It must be measured, not estimated. It must be traceable to something the reader can see. And it must be stated conservatively, because the reader will halve whatever you say and be right to. You are losing thousands every month to abandoned carts fails all three. Of 1,284 carts started on your site last month, 96 completed; the rest were abandoned before payment passes all three, and it is far more frightening, because it is arithmetic rather than rhetoric. Exercise. For one product, list the cost in each of the five currencies. Then choose the single one that the buyer would bring up first if you were sitting with them. Use that one and leave the other four out of the copy. You may keep them; the page may not.
Recognition says I know your day. Cost says and here is what it takes from you. The after says: here is the same day, once this is solved. This is where most pages fail, and they fail in a specific way. They describe the absence of the problem instead of the presence of a different day. No more late payments is an absence. You open the banking app on Friday and the money is already there because the reminder went out without you is a day. Only one of those can be pictured, and people do not buy absences. They buy a picture of themselves in a better Tuesday.
The after must be the same day, not a better life. This is the discipline that keeps the petal honest and also the one that makes it believable. If recognition described a Thursday evening at the kitchen table with two accounts that will not join, the after must be that same Thursday, at the same table, with the two accounts joined. The reader must be able to see the improvement without changing the furniture. The pattern to write into, and it is deliberately small:
Before (their words). The specific moment, badly.
After (your product’s doing). The same moment, and what has changed, mechanically.
Mechanically is important. You feel calm is not an after, it is a mood claim, and mood claims are unverifiable. The two accounts are one page and the sum is at the top is an after, because the reader can see how the calm arrived. Feelings in copy should be consequences of mechanisms, never substitutes for them.
There is a temptation to write the after at the scale of a transformation: your business finally works, you get your life back, you stop worrying about money. A reader has heard this from every vendor they have ever met, and a large after is precisely what makes them suspect the product is small. A small after is more persuasive than a large one, for a reason worth stating plainly. The reader is trying to estimate risk. A promise of transformation implies the vendor has not understood how modest the real improvement is. A promise of one recovered hour on a Sunday implies the vendor has been inside the actual week. Small afters read as experience. Large afters read as sales. So the after should be the size of a single change, in a single routine, described from inside. If a page can say you will stop dreading the fifteenth and mean it, that is worth more than a hundred sentences about transformation.
The commonest failure is that the after is written by somebody who knows the product too well. It becomes: a dashboard with real-time reconciliation and multi-account sync. That is a description of the machine. The reader has no idea what their Thursday now looks like, only that there is a dashboard in it, which may well be another thing to check. The repair is a question you ask of every feature sentence: what does the reader no longer do? Features are usually the removal of an activity. Multi-account reconciliation removes the activity of copying numbers between two screens. Say that, and the feature disappears into the improvement, which is where it belongs.
The after can be one sentence or a short paragraph, and it should be the most vivid thing on the page. It comes after recognition and cost, before proof and door. One after per page section — not one per feature. If you have six features, you do not have six afters; you have one day, and six reasons it changed. Exercise. Write the after for your product as a scene with a time of day and a place. Ban the words “easy”, “simple”, “efficient”, “streamlined” and “peace of mind”. Then check: is it the same day as your recognition line, or a different, better-resourced life? Rewrite until it is the same day.
Recognition makes the reader feel known. Cost makes them feel the weight. The after makes them want it. Proof is where they decide whether to believe you — and it is the petal most often replaced with its own imitation. Proof is something true that the reader can check, without asking you, in under a minute. That definition does three things at once and each clause matters. True. Not plausible. True. If it cannot be true, it cannot be proof. Checkable. The reader can verify it themselves. This is what separates proof from a claim: a claim asks for trust, and proof does not need to. Under a minute. If checking costs an afternoon, it is not proof for a page. It is a footnote for a diligence process.
One: the free thing that does the job in miniature. A scanner that shows the reader their own result. A twenty-second audit that scores what they have. A demo that is bit-for-bit the product rather than a curated tour. This is the strongest proof because it converts the reader’s scepticism into their own data. Nobody argues with a number they generated themselves.
Two: the specific, dated, sourced observation. A real notice, with its date and its issuing body. An audit finding, with what was measured and when. A published price. A changelog. A status page. Not a statistic about the category but a fact about this reader or this market, which they can go and look at. Three: the artefact. A sample output. A template. A specimen report. An actual page of what the product produces. The reader can see the quality without being told about it, which is the point of showing a carpenter a joint rather than a brochure. Four: the verifiable statement about yourself. How many people work here. What the price is, in full. What you will not do. What happened when you failed. These are proofs precisely because they are not flattering, and a vendor who volunteers an unflattering fact is believed about the flattering ones.
— A testimonial you cannot attribute. A first name and a job title is not attribution. If the reader cannot find the person, the quote is decoration. — A logo wall with no permission and no context. Logos prove an invoice was paid once. They do not prove the product works. — A number with no denominator. “Ninety-four percent satisfaction” from how many, asked of whom, for what question. — A stock photograph of a smiling person in a hard hat. This is proof of a budget, not of an outcome. — Your own description of your own rigour. “We take quality seriously” is a claim about a claim.
Every venture in Part Six has one, and the pattern is worth stating early because it is load-bearing: every product needs something free and useful that arrives before the price, and that works even if the reader never buys. This is not a lead magnet. A lead magnet is a trick with a download on it — a checklist you must give an email address to receive, which the reader did not want. The free nectar is different in three ways. It is genuinely useful on its own. It requires no commitment, ideally no email. And it is honest about the product’s limits, which means it can tell the reader they do not need to buy. A free thing that sometimes concludes “you do not need this” is the single most credible asset a small company can have, and almost nobody has one, because it feels like giving away the sale. It is the opposite. It is what makes the sale possible at all.
Order matters here for the same reason it mattered with recognition. A price arriving before proof is read as a cost. The same price arriving after proof is read as a decision. The reader needs to have established, from evidence rather than assertion, that the thing is real, before a number can mean anything except expenditure. Which is also why the proof must be specific to this reader. A general proof — a big customer, an impressive number — is weaker than a narrow one. The reader’s own site’s audit finding, their own country’s open tender, their own shop type’s benchmark: these are proofs that were made for one person, and being made for one person is what makes them believed. Exercise. Write down every piece of proof you currently have, then mark each one against the three clauses: true, checkable, under a minute. Delete everything that fails. Count what is left. If the answer is zero, your next task is not copywriting, it is manufacturing proof — and until that is done, no amount of writing will fix the page.
The last petal is the door, and it is the one that most pages get wrong, not by being too small but by being too many. A page will typically offer: a free trial, a demo booking, a pricing link, a newsletter sign-up, a chat widget, a download, and a phone number. Each of these is a decision the reader must now make, and the sum of them is a reader who makes none. Choice is not generosity at the point of commitment. It is a queue. One section, one door, one ask. That is the rule. A long page can have several sections and therefore several doors, but each door must be inside one section, addressing one audience, asking for one thing. If two doors appear in the same section, the section has two audiences and should have been two sections.
A door is small when the risk is carried by you rather than by the reader. There are five dimensions, and a good door reduces at least three of them: — Money. Free is smaller than cheap, and cheap is much smaller than a card on file. If a card is required, say so in the button, not in the terms.
— Time. Minutes, not weeks. “Two minutes” is a door. “Book a thirty-minute discovery call” is a commitment and should be labelled as one. — Skill. No setup, no import, no configuration. If the reader must prepare something first, the door is not the door — the preparation is, and it will not happen on a Friday. — Exposure. No public commitment, no announcement to their team, no post, no review. Anything visible to their colleagues raises the price. — Reversibility. Can they undo it? A free tool leaves nothing behind. A subscription needs an obvious way out. Say what the way out is, in the place where they enter.
The pattern requires a promise line: our promise: …. It exists because the reader is estimating what happens if this is wrong, and silence on that question is not neutral — they will fill it with their worst experience of a vendor. The rule attached to it is the sharpest in this book:
Every promise is a debt. Any line beginning “Our promise:” must be something you will honour without argument.
Without argument is the operative phrase. A refund policy that requires a support conversation, a reason, and a manager’s approval is not a promise; it is a negotiation you have pre-decided to win. A promise that the reader has to fight for is worse than no promise, because it costs you the customer’s goodwill twice.
So promises must be few, written down, kept in one list, kept current, and priced into the business before they are published. Part Eight contains the list itself; it should be reviewed whenever a product changes, because a promise that has quietly become unaffordable is a debt you are still paying interest on.
The button is copy. It is not a label. Submit, Learn more, Get started are the three least informative strings on the internet. A button should say what happens next, in the reader’s words: — Score my site free — twenty seconds — See the notice for my country — Send me the sample chapter — Read the audit finding A button that names the next step and its size is itself a small act of honesty: the reader knows exactly what they are agreeing to, which is unusual enough to be memorable.
The same rule on the email side: one ask, and if there is no ask, say so. An email that says “no need to reply, here is the tender list” is a door of a different kind, and often the one that gets a reply, because it does not charge the reader anything. This is why the free nectar belongs in the message itself rather than behind a form.
A door is only small if the reader can decline it without cost. That means a visible way to say no, and it means the writer has to be at peace with being declined. Two sentences do most of the work:
If this isn’t useful, ignore it, or reply stop and I won’t write again.
The respect is not a technique for softening a reader into a yes. It is the actual offer of a no, which is what makes the yes worth having. A customer who said yes because there was no polite way to say no is a churn statistic waiting for its date. Exercise. Draw your page’s doors as a list. If there is more than one door per audience section, cut until there is one. Then write the promise line. Then check that every promise can be honoured by a member of staff without asking anybody’s permission.
The five petals are a sequence, and the sequence is the framework. Most pages that “have all five petals” and still do not work have them in the wrong order, which is like a good meal served in the wrong sequence: all the components, none of the effect.
Recognition before anything about you. Forced, because a reader who has not recognised themselves has no framework for interpreting your credentials. Introduction-first pages ask the reader to care about a stranger’s biography before establishing that the stranger has been in their kitchen. Cost before the after. Forced, because the after only has value as relief from something. Describe a solution before the problem and the reader has no weight to lift, so the solution feels like an improvement to a situation they were not worried about. This is the single most common order error in software marketing, and it produces pages that are accurate, complete and unmoving. After before proof. Forced, because proof is evidence of something, and until the reader knows what is being claimed they cannot judge evidence. Proof arriving first is trivia. Proof arriving after the after is confirmation.
Proof before price. Forced, because a price is a judgement, and judgement needs a standard. The standard is the proof. Also, practically: a reader who has seen the free tool work does not need to be talked into the price, and a reader who has not will not be. The door last, and only one. Forced, because the door is the answer to a question the reader has not yet asked until the previous four petals have been read. A door in the first paragraph is an interruption. A door after the proof is a relief.
One: the self-introduction. We are a company founded in 2016 with a mission to… The page is about the flower, at the moment the bee is looking for nectar. Move it to an about page, where the reader who wants it will go. Two: the feature-first after. The page opens with a dashboard screenshot and a list of capabilities, and the problem arrives in paragraph six, if at all. The reader’s first thought is “what is this for”, which is the question the copy was meant to answer. Three: the proof dump. Logos, awards, statistics, badges, all in the hero. This is the failure mode of a team that has been told to add credibility. It produces the opposite: a page with no claim, only evidence for a claim it never made. Four: the premature door. A form in the hero, and again in every second section, and a chat window over the third. This is the failure mode of a team measured on conversions, and it converts worse, because it asks before it has given.
Read only the first sentence of each section, top to bottom, and label each one with its petal. You should get something close to: recognition, cost, after, proof, door. If you get door, door, proof, door, recognition, the page has all five petals and none of the structure, and the repair is not editing but re-ordering. There is one permitted variation and it is worth naming so you do not invent others. On a page with several audiences, the recognition and cost may be repeated per audience — a card per persona, each with its own before, its own after and its own door. That is not a broken order; it is the order running several times in parallel, once per bee. What is never permitted is skipping a petal for some audiences because you ran out of space. If there is no room for the proof, there is no room for the door either.
Rules are what a philosophy becomes when somebody is tired. Each of the nine was learned by breaking it. Each is short enough to check a sentence against, which is the only useful length for a rule.
Four of the nine rules protect the difference between feeling and invention. They are the ones to hold hardest, because they cost the most and they are the ones a business will try to negotiate first.
You may describe a common, recognisable situation (“You find the notice four days before it closes”). You may never invent a customer, a quote, a number of users or a result.
The first half of this rule gives permission and the second half removes a shortcut, and both halves are necessary. Without the permission, writers become afraid of feeling and produce the flat pages of Part One. Without the prohibition, feeling becomes fabrication and the page becomes a liability. The line between the two is whether the sentence is a description or an assertion of fact. “You find the notice four days before it closes” is a description of a situation that recurs for a class of people. It asserts nothing about a specific individual, and no reader can check it and find it false, because it is offered as a type. “Our client Sarah found the notice four days before it closed and won the tender” asserts a specific person and a specific outcome. That assertion is either true and documented, or it is a lie with a name attached to it.
The practical test: can a reader reasonably assume the sentence refers to a real, identifiable person? If yes, that person must exist. If the sentence refers to a kind of day rather than a named individual, it is a description, and descriptions are how honest copy is built. This is also why the examples in this book are labelled. Every business-owner quoted in these pages is an invention of the author’s, constructed to demonstrate a pattern. The book says so at the front, in one place, plainly, rather than defending itself later. That is the same test applied to itself, and any voice guide worth having should be able to survive being pointed at its own pages.
“Founding price for the first five clients” is true when the promo code has stock five. “Only two left” is banned unless a system counts it.
Two conditions make a scarcity claim honest, and both are mechanical rather than moral. There is a limit that somebody chose and can name. First five clients, first fifty seats, the first cohort of twelve, the first hundred at this price. The number comes from a real constraint — hours in the founder’s week, support capacity, an introductory arrangement with a supplier — and the constraint is the reason the limit exists. A machine enforces it. If the code has stock five, the sixth person is refused by a system. If a spreadsheet is updated by hand and the count is unofficial, the limit will quietly move when the fifth client is harder to close than expected, and the reader who checked back in a week will discover that “two left” was theatre. This is one of the few places where honesty is cheaper than dishonesty, and it is worth saying because the assumption usually runs the other way. An enforced scarcity is easy: let the system refuse. A fake one requires maintenance, memory and a story for when it is questioned. The honest version is less work. Where a limit exists but is not a scarcity — a support capacity, a delivery window, a queue — say the operational truth instead of the sales version: we take four new clients a month because one person does the onboarding, and she has three slots left in November. Same number, no theatre, and it explains itself.
A tender deadline, a renewal date or a season is real urgency. A countdown timer you control is not.
The test is agency. Who decided this would end, and could they change their mind? Tender closing dates are set by the buyer and published. Renewal dates are in an existing contract. Tax deadlines are in law. Seasons are in the sky. Funding schemes close when the money runs out. None of these can be moved by a marketing team on a Thursday, which is exactly what makes them believable. A timer is a decision, not a fact. It is also self-defeating in a way that is easy to observe and hard to argue with: a reader who has seen your timer reset has learned that your time-based claims are decorative. The damage is not to that campaign, it is to the next one and to the one after. You are spending a fixed asset to buy a variable one. There is an honest version of a closing price, and it is the one that anticipates its own end. The founding price is for the first five clients; after that it goes to the standard price, and we will say when the five are gone. That sentence is a timetable, not pressure. It is also checkable — either the five are gone or they are not — which is what makes it work.
Any line beginning “Our promise:” or “If… you get your money back” must be something you will honour without argument. List them in the Codex and keep the list current.
A promise is not copy. It is a liability entered into in public, and it should be treated with the same care as a clause in a contract, because functionally it is one. Without argument is the phrase that gives the rule its teeth. Most refund policies are promises with a procedure attached, and the procedure exists to make the promise survivable rather than to make it keepable. A customer who must explain themselves, wait for a manager and accept an exchange is being argued with, and they will tell other people about it, and the telling costs more than the refund. So before you publish a promise, three questions: — Can this be honoured by any member of staff, immediately, without asking me? If not, it is not a promise, it is a policy. — Is it priced in? If five percent of customers claim it, does the business still work? If the answer is no, the promise is a loan against next year. — Is it on the list? Part Eight holds the Codex. A promise that exists only on a landing page will eventually contradict a promise on another landing page, and a reader will find both. Promises should be short, few and slightly better than the reader expects. Three good ones are worth more than twelve hedged ones: a free tool that leaves nothing behind, a sample that is the actual product, a refund with no interview attached. Each is a debt you can carry.
They are all versions of a single commitment: you do not get to borrow belief. Hype borrows belief from a feeling the reader did not have. Fake scarcity borrows it from a count that does not exist. False urgency borrows it from a calendar you control. A promise you will not honour borrows it from the customer’s future. Borrowed belief is a loan with a hidden interest rate. It is repaid in customer support conversations, in reviews you cannot answer, in a team that stops believing what it writes, and in the slow conversion of your marketing from an asset into a liability that has to be managed. The rate is not visible in any dashboard, which is why the loan is so popular. Exercise. Take one page. Find every sentence that makes a claim about time, quantity or outcome. For each, write where the number comes from and who could verify it. Sentences with no answer are not to be rewritten yet — they are to be deleted, and then the page will tell you what it actually needed to say.
Three rules regulate heat. They exist because the first attempt to remove hype removes warmth with it, and warmth is the thing that makes a page feel written by a person rather than assembled by a department.
Write to one person, as a person: “you”, “your team”, “your Monday”. Don’t write “Dear valued customer”, and don’t praise them falsely.
Warmth is created by attention, not by endearment. A sentence that shows you have noticed something about the reader’s week is warm. A sentence that tells the reader they are wonderful is flattery, and flattery reads as a technique because it is one — it costs nothing and applies to everybody, and readers know it. The difference in practice is the number of specifics. We know how hard you work is flattery: true of everyone, therefore felt by no one. You are doing the quotes at night because the days are full of the work the quotes are for is warmth: specific, slightly uncomfortable, and recognisable. The address matters as much as the sentiment. ”You”, “your team”, “your Monday” — singular, concrete, possessive. Not “our customers”, which turns the reader into a category the writer manages. Not “businesses like yours”, which is a mailing segment. A reader who is addressed as a class will behave as a class, which is to say, not at all. There are two flatteries to watch for beyond “valued customer”. The first is the compliment that is really a claim about you: we’re delighted to work with the most ambitious founders in the country. The second is the empathy performance: we understand that your time is precious. Both are the writer talking about the writer’s own sensitivity. The repair is nearly always deletion. Remove the preamble, start with the thing, and the warmth is in the attention rather than in a sentence about warmth.
“The container sat two days longer” moves people more than “Optimise your logistics”. Concrete nouns, short sentences, one idea each.
Cleverness is a pleasure for the writer and a tax on the reader, and a tired reader will not pay it. Every pun in a headline is a small puzzle, and a puzzle in the first line is a reason to stop reading in the second. Specificity is not the same as detail. Detail is accumulation; specificity is selection. The container sitting two days longer is one detail, chosen because it carries a whole failure — the customer asked, nobody knew, the week was re-planned. “Two days” is doing the work of a page. That is what specific means. Three mechanics are worth practising until they are automatic.
Concrete nouns. A noun you could photograph: a container, an invoice, a till, a notice, a rota, a renewal letter, a Monday. Replace every abstract noun you can. “Solution” is not a noun you can photograph, and “optimisation” is a noun that has never been photographed. Short sentences, one idea each. Long sentences are how writers hide uncertainty: three clauses let you qualify the first two into safety. Take the sentence apart and the uncertainty has nowhere to stand. The rule is not a style preference; it is a diagnostic. If a sentence cannot be split, you probably understand it. The detail that costs you something. The most specific details are usually the least flattering: the four days, the two days, the third reminder, the Thursday. A page that only contains flattering specifics is not specific, it is selective in the other direction, and readers can feel the balance. There is a version of “specific beats clever” that applies to whole structures, not just sentences. A page with one sharp scene and six sharp lines beats a page with fourteen clever lines, because a reader remembers one thing per page and the writer’s job is to choose which.
Your buyers are tired. The emotional promise of almost every title this press publishes is calm: a calm reader, a calm evening, a book that does not raise its voice at somebody who has had a long day. Write calmly. No exclamation marks in body copy.
This is the rule most likely to be broken by somebody well-meaning, because enthusiasm feels like care. But look at who is reading. The shop owner is reading at ten at night. The founder is reading between two calls. The person who missed the tender is reading with the specific tiredness of somebody who cannot afford to miss another one. None of them needs your excitement. They need your steadiness. Calm at the sentence level has five properties, and they are all subtraction. — Low volume. No exclamation marks in body copy. An exclamation mark is the writer asking the sentence to be more important than it is. — No hurry. No “now”, no “today”, no “don’t miss”. A calm page states a fact and leaves the reader to decide when. Counter-intuitively, this produces faster decisions, because there is nothing to resist. — Even temperature. No alternating between crisis and delight. A page that swings makes the reader brace, and a braced reader does not buy. — Declarative mood. Statements rather than questions and rhetorical nudges. “The notice closes on the fourteenth” is calmer than “Ready to stop missing deadlines?” — Room around the sentences. Short paragraphs, white space, one idea per line. Calm is partly typographic. A dense paragraph is a raised voice. There is a failure mode here worth naming before it happens to you, because it is the one that turns calm into something worse than hype. Calm is not coldness, and restraint is not indifference. Written badly, calm becomes the voice of a company that has decided not to care: accurate, distant, faintly administrative. The distinction is that a calm page is still warm
— it has recognised a specific person, named their cost in their own currency and given them a small door. Calm is low temperature, not low attention. A funeral is calm. So is a good doctor. Write like the doctor. The test for whether you have written calm or merely flat: does the page contain at least one sentence that could only have been written by somebody who was paying attention? If not, you have removed the feeling along with the noise, and you are back in Part One. Exercise. Take a page and remove every exclamation mark, every question in a heading, every “now” and “today”, and every sentence longer than twenty-five words. Read it aloud. Then add back only the ones whose absence made the page colder rather than calmer. Usually that will be none of the exclamation marks and one of the long sentences.
The last two rules govern the moment of asking, and the practice that makes all the others possible.
Every outreach message makes “no” easy and honourable (“A no is completely fine”). Every message has an unsubscribe or a “reply stop”. That respect is itself the most attractive thing in the message.
A no that is difficult to say is not a yes. It is a complaint in waiting, and it arrives later, in public, in a form you cannot answer. Respecting the no has three parts. It is visible. Not buried in a footer in eight-point grey. In the body, in the same size as everything else, in the sender’s own voice. If this isn’t useful, ignore it, or reply stop and I won’t write again. One sentence. The reader stops bracing, and a reader who has stopped bracing can actually read what you wrote. It is honourable. Not “if you’d rather not hear from us, click here to manage your preferences”. That sentence makes the reader feel like a problem in a system. A no is completely fine costs three words and treats the reader as an adult with a busy week.
It is kept. Reply stop and the writing stops. Not “stops for six weeks”, not “moves to a monthly digest”. If the machinery cannot honour the no reliably, the message cannot be sent, because a promise about stopping that is not honoured is the exact category of debt the fourth rule forbids. There is a commercial argument for this and it is worth stating, since the ethical argument is usually enough for good people and this one is for the ones under pressure. The reader who is given an easy no is more likely to read the next message, and much more likely to remember the sender when the problem arrives. A list of four hundred people who could say no and did not is worth more than four thousand who could not find the link.
Each message is written for one recipient and contains something only that recipient would recognise: their site’s audit finding, a tender in their country, their shop type, their course. Systems drip them out slowly. A message that could be sent unchanged to two people is not finished.
This is the rule that costs the most and defines the company that can keep it. It rules out the entire architecture of modern outbound: the sequence, the personalisation token, the “Hi {first_name}” that produces the illusion of attention while delivering none. The test is brutal and easy to apply. Show the message to a second recipient. Does any sentence now read as wrong, or merely as true of them too? If it is merely true of them too, it is not finished. What makes a message finished is one observed thing. Their site’s scan result. A tender published in their country on Tuesday. The fact that their checkout requires an account before payment. The course they are halfway through. Something you found by looking, that they did not tell you, that is checkable and correct. That one observation is the entire difference between a message and a mailing. This is why the systems must drip slowly. A person cannot observe a thousand things a day. Ten is a good day. Ten a day is three thousand a year, and three thousand observed messages will outperform a hundred thousand templated ones, not by a small margin but by the margin that decides whether outbound works at all — and if the volume is wrong, the honest fix is to reduce the volume, not to weaken the rule. The same rule scales downward inside an organisation. A support reply that could be sent to any customer is not a reply. A release note that could describe any release is not a note. The rule is not about email; it is about the difference between addressing a person and addressing a list.
The rules so far assume a normal week. Businesses do not only have normal weeks, and a voice guide that has no emergency clause will be abandoned on the first bad day — which is a problem, because the first bad day is exactly when the way you write matters most. So here is the clause, and it is deliberately narrow.
When something has gone wrong — an outage, a security incident, a data loss, a billing error, a missed deadline you promised — the shape of the message changes but none of the rules are suspended. You still name real feelings. You still do not invent numbers. You still do not manufacture urgency, because the urgency is now real and needs no help. You still have one door. You still respect the no, though in an incident the “no” is usually “stop emailing me about this incident”, which must also be honoured. What changes is the order and the temperature, and the pattern is worth memorising because you will not have time to design it while it is happening. Say what happened. One sentence, plain, no hedging. Not “some users may have experienced intermittent degradation”. The dashboard was unavailable from 09:14 to 11:02 on Tuesday. Say who was affected and how. Specifically. Counts if you have them, honestly bounded if you do not. All accounts on the EU region. No data was lost; writes made between those times were not saved. Say what you have done. Past tense, mechanical. We restored the service from backup at 11:02 and added the check that should have caught it. Say what happens next, with a time. Affected accounts will be credited automatically on the next invoice. I will post an update here on Friday whether or not there is more to say.
Say what the reader should do, and make it small. Usually nothing, and saying so plainly is a form of respect. Do not apologise more than once, and do not perform distress. A page of apology is the writer asking the reader to comfort them. One accurate apology, then the facts, is what a person in a bad week can actually use. Do not promise it will not happen again. You cannot honour it, and the fourth rule applies. Promise the specific thing you can do: the check, the alert, the credit, the update. There is a temptation in an incident to use plainness as an excuse for coldness. Resist it in one place only: acknowledge that this cost the reader something. That is an hour of your Tuesday that we took, and it was our fault. One sentence, at the point where the facts stop and the repair begins. Then stop. The composure is the apology. Exercise. Write the incident page for a plausible failure of your own product — pick the one you would least like to happen. Write it in ten minutes, in the five parts above, before you have time to become a committee. Keep it. On the day, you will be glad of a draft that was written by somebody calm.
A framework is tested only when somebody has to publish something by Thursday. These are the seven surfaces where it is tested, with the pattern for each and the failures that recur.
The landing page is where the five petals are visible in sequence, and it is the surface where a single order error costs the most. The pattern for a hero has four elements and there is room for nothing else.
Headline: the feeling or the moment of loss, in the buyer’s words. The tender you never saw is the one somebody else won.
>
Sub-line: what the product does, for whom, at what entry cost. Free or low-risk entry goes here.
>
Two buttons: the small door (free) and the proof (sample or free tool).
>
Nothing else. No logo wall. No carousel. No chat bubble in the hero. No “trusted by”.
The headline is recognition. It is not a slogan, and it is not a description of the category. It should be a sentence the buyer could have said out loud on the day it happened, which is why the example above works: it is said by a person after a tender result, not by a company. The sub-line does the work that the headline is forbidden to do, because the headline is carrying feeling and cannot also carry explanation. The sub-line is plain, factual, and contains three facts: what it does, who it is for, what it costs to start. Entry cost belongs here because a reader scanning the sub-line is deciding whether to keep reading, and “free” is the most persuasive word available to you — if it is true. Two buttons, and their purposes are different. The first is the small door: the free thing, minutes, no card. The second is the proof: the sample, the demo that is the actual product, the free tool. Note what is missing. There is no button for pricing, no button for “contact sales”, no button for the newsletter. A hero with four buttons has asked four questions and received no answer.
Below the hero, one card per audience. Each card has three lines and one door.
The bad moment — red before line, specific and short. The same moment, solved — green with [product] line. The door for that audience.
The colour convention is not decorative. It lets a reader find their own card in a second without reading the other three, which is the practical version of the recognition rule at page scale. Someone who has a shop and not a clinic should be able to locate the shop card by shape and colour, then read only that. Three lines is the discipline. The moment, the same moment solved, the door for this one. What usually creeps in is a fourth line — a feature list per audience — and it should be cut, because the card’s job is to sort the audience, not to sell to it. The selling happens on the page the door leads to.
Both are constructed examples, written to show the sequence. Version A — the failure.
Reimagine how your team manages procurement. Our intelligent platform brings clarity and efficiency to every decision, with powerful tools designed for modern organisations. [Request a demo] [Speak to an expert] [View pricing] [Watch the video]
Diagnosis, in the order the failures occur. The headline is a slogan with no moment in it and contains a command, which is the least recognisable form of speech available. The sub-line is four adjectives in a sentence about the machine. There is no cost, no after, no proof. Four buttons, all of them asking the reader to commit time. And the reader — a procurement person who missed a notice on Tuesday — has no way to know that this page understands their year.
Version B — the repair.
The notice you found on the day it closed. Tenders for your country, in one list every Monday, with the deadline, the buyer and the sum already checked. Free to start. [See this week’s list for my country — free] [See a sample audit]
>
If this was your week. You heard about the tender from a competitor at a conference, and by then the closing date had passed. With Threshold. Every open tender in your country arrives on Monday morning, sorted by closing date, with the ones you are eligible for marked. [Show me this week’s list]
>
If this was your month. You priced the job correctly and still lost money, because the exchange rate and the shipping were quoted in different months. With Threshold. The quote carries both rates, and the margin is recalculated on the day it goes out, with the difference shown. [Open the margin calculator]
Look at what Version B does not contain. No adjectives. No logo wall. No “trusted by”. No exclamation marks. Two hero buttons, one free and one a sample. Each card has three lines and one door, and the doors differ because the audiences differ. Total claim count: zero unverifiable ones. Total feeling: high, because every line is a moment somebody has lived.
The carousel. The reason it persists is that it flatters everybody in the meeting: each team puts its message in a slide, and the hero becomes a compromise in rotation. It converts badly because the reader cannot see all of it and will not wait. One headline. The convertible hero. Ready to take control of your tenders? A question mark invites a reply the reader has no way to give. Questions in headlines are the writer asking permission to make a claim. The feature grid as hero. Nine boxes with nine icons, immediately below the fold-flap. A reader cannot derive a feeling from a grid. The grid belongs far below, as reference for a reader who has already decided. The proof before the claim. Awards, certifications and logos directly under the headline, before any statement of what the page is about. This is what happens when a team is told to “build trust” without being told what to build it in. Evidence for an unspecified claim is noise.
The fold is not where you think it is. On a phone it is roughly one headline, one sub-line and one button. Whatever you were planning to establish “above the fold” on a desktop has to fit in that. Write the phone hero first, on a narrow column, and let the desktop version be the relaxed one. And a landing page may be long. Length is not the problem, repetition is. A long page that runs the five petals once, then again for a second audience, then again for a third, is three pages stitched, and stitched pages read fine. A short page that says the same thing four times is long. Exercise. Take your current hero. Delete every element that is not one of the four. Read what is left to a person outside your company and ask what the product does and what it costs to start. If they cannot answer both, the hero is not finished, regardless of how it looks.
The product card is the shortest surface in the book and the one where restraint shows most clearly. It has six lines and a button, and the pattern is fixed:
[Name] · [price] [Feeling line in italics: the after, in one sentence] For: [who, plainly] You get: [3–5 concrete items] When: [time you commit to] Our promise: [the risk reversal you will honour] [One button]
Nine out of ten product cards fail on the second line, so start there.
It is the after, in one sentence, in italics, and it is the only place on the card where feeling is permitted because it is the only place with room for it. It must be a day, not an attribute, and it must be small. Failure. Powerful, flexible, built for scale. Three attributes, no day, no reader. Failure. Transform the way you work. A large after, which reads as sales. Repair. You know what the shop took this week, and the wholesaler’s bill is already in the same number. That sentence is the product. Everything above it in the card is administration — name, price, who, what, when, promise — and the feeling line is the reason anybody reads the rest.
Name and price. In full, together, no asterisk. A price that requires a page to understand is not a price, it is an enquiry. If the price genuinely varies by size, print the smallest and say so: from £18 a month, by number of shops. For. Who, plainly. Not “ambitious teams” but a shop with one counter and one person doing the books. The purpose of this line is to let the wrong reader leave, and it should be written with that intention. You get. Three to five concrete items. Concrete means a noun the reader could point at: the till exports to your accountant’s format, the reminder sends on the fifteenth, the notice list arrives on Monday. Five is a maximum; three is usually better, because the reader will remember three. When. The time you commit to. Onboarding, response time, delivery, how long the free tool takes. This line is where most of the trust is built, because it is the line that most companies refuse to write, since a committed time is a debt. Write it anyway. A competitor who will not commit to a time has handed you the choice. Our promise. One, from the Codex, that you will honour without argument. Not two. Not a paragraph.
Read the six lines together and check the sequence: the reader learns what it is, how it will feel, whether it is for them, what they get, when, and what happens if it is wrong — and then one door. That is a complete decision in under a hundred words, and it is the most efficient piece of writing most companies never do. The commonest failure is not any single line but the loss of the feeling line under review. Somebody in a meeting says it is “a bit soft”, it gets replaced with a benefit statement, and the card becomes accurate and unmoving. If you only defend one line on this card for the rest of your career, defend the second. Exercise. Write the feeling line for each product you sell, one sentence each, no adjectives, containing a day. If you cannot, you do not yet know what the product changes, and no other part of the card will help.
The first-touch email is where the no-bulk rule gets tested by reality, and it is the surface most likely to be quietly abandoned in favour of a sequence. So this chapter is as much about the discipline of producing them as about the writing. The pattern:
Subject: [their name/firm/site] + the specific thing you noticed
>
Hi [Name or “there”],
>
[Recognition: how you came across them, truthfully, in one line.] [Proof: the one specific, checkable thing — an audit finding, an open tender, with the link.] [The after, in one sentence: what that could mean for their week.] [The small door: a free thing they can use without replying.] [The easy no.]
>
Courage Garikai · [venture] · [postal address] · [unsubscribe]
It has two components and both are factual: who they are, and what you noticed. Nairobi Dental — your booking page asks for a card before the appointment. It is long by the standards of subject-line advice and it is opened more often, because it is obviously about them and obviously written by a person who looked. What is banned here is the same thing that is banned everywhere: curiosity gaps, fake replies (“Re: your site”), lowercase intimacy designed to look like a friend, and any subject that would work equally well on a thousand recipients. Quick question is a subject line addressed to nobody.
One line, and it says how you came across them, truthfully. Truthfully is the whole rule, because the standard opening of outbound — I was browsing your website and loved your work — is the most common lie in business correspondence. Honest versions: I found you on the county supplier list. Your shop came up when I was looking at who stocks in Avondale. You appeared in Tuesday’s tender results. Each is a fact about how a person spent an afternoon. It is unglamorous and it is believed.
One specific, checkable thing, with a link. Not three. The single most effective first-touch email is one observation the recipient did not know about themselves, especially if it is mildly unflattering:
Your booking page asks for a card before the appointment is confirmed. On a phone, that step takes four screens and asks for the card before the price. I ran the check twice to be sure.
That is proof, it is checkable (they can go and look), and it is the only reason the email is worth opening. The link goes to where the thing can be seen: their own page, the published notice, the audit output.
One sentence for the after, and it should be small. If that step went, the booking would take two screens and you would see the drop-off. One sentence, then the free thing they can use without replying. The door is deliberately weak here, and this is where most writers ruin a good email. The reader does not know you. Anything that requires a reply is too expensive on a first touch. So the door is a thing that arrives whether or not they respond: the report, the list, the sample, the tool they can run themselves. The full check is here, no need to reply — it takes about a minute. That sentence has produced more replies than any call to action, because it charges nothing.
In the body, in the same voice, in the same size.
If this isn’t useful, ignore it, or reply stop and I won’t write again.
Then the signature: name, venture, postal address, unsubscribe. These are not only legal furniture. In a message this personal, they are part of the honesty: a person with an address, who can be found, who will stop if asked.
The operational question is how to produce these without a machine and without self-deception. Three practices hold it together. One at a time, researched first. The observation comes before the email exists. You look at a business, find a thing, and then write. If you find nothing, there is no email — and skipping that business is not a lost lead, it is the rule working. Ten a day, no more. A person can observe ten things carefully. At ten a day, five days a week, that is two thousand six hundred observed messages a year, which is a large outbound operation and a tiny one compared to what a sequencer promises. It will outperform the sequencer, and it will not exhaust the sender into writing “quick question” by week three. A record of what was noticed. Keep the observation, the date and the source for each message. This is how you follow up honestly (Chapter 16 needs something new to say) and how you avoid writing to the same firm twice with the same finding.
If a system writes the first line — a scanner that produces a finding, a database that produces a tender — the message can still be honest, but only if a person has read the output and the output is specific to one recipient. The test does not change because a machine did the looking: could this be sent, unchanged, to a second recipient? A machine that produces one true, specific observation per recipient is doing the same job a person does, faster. A machine that produces “{first_name}, I noticed your website could be improved” is producing a mailing with a variable in it. Exercise. Find one business you could help, and write down three things you observed about it, with the date and where you saw them. Then write the email using the pattern, keeping to the word count of a person who is busy. Send it to nobody. Read it back tomorrow and ask whether it is the sort of message you would be pleased to receive.
Only three follow-ups exist in the pattern: the first message, one at day three, one at day seven, and then silence. Not day fourteen, not the monthly newsletter they did not ask for, not the “last one from me” that is not the last one. The condition attached to each follow-up is the same, and it is the line that separates follow-up from pestering:
One or two lines with something new: a second finding, a new notice, a free tool. “Just checking in” with nothing new is banned.
A follow-up carrying something new is a small gift and the reader is glad of it. A follow-up carrying nothing is a request for attention with no content, and the second one of those closes the flower permanently. The two-line rule matters as much as the “something new” rule: if the follow-up is longer than the thing it is following up on, it is a second email wearing the first one’s clothes. Here is what day three actually looks like.
Subject: the second thing on the booking page
>
The same page also asks for a date of birth after the card. Two steps would become one. Still no need to reply — the full check is here.
Two lines, one new fact, one free thing, no pressure. If there is no second finding, the correct action is not to send a shorter version of nothing. It is to wait until there is something, or to stop.
When somebody replies, everything about the register changes and most of the rules become easier, because you are now in a conversation rather than an approach. The pattern is four moves: Thank them, briefly and without gush. One clause. Thanks for coming back to me. Answer exactly what they asked. Not the adjacent question you would prefer, not the objection you have prepared a paragraph for. If they asked about the price, answer about the price, in the first line, in full. If the answer is no, say no. Offer one next step with a time. I can send the full audit on Wednesday, or if you would rather talk, I have twenty minutes on Thursday at ten. One step, and a named time — a time is the smallest possible commitment and costs you nothing to offer. Then stop. No second question. No “and by the way”. No attachment they did not ask for (the no-attachments rule is for first touches, but the spirit applies: send what was requested).
The most common error in a reply is answering a question with an invitation to a meeting. It reads as evasion, and it is, because a question answered is a small door and a meeting invitation is a large one.
There are four stopping points and each should be honoured without interpretation. No reply after three messages. Stop. No fourth. The reader has answered, in the only way available to them. A no. Stop, and say so once, warmly, so that they know it is real: Understood, and thank you for replying — I won’t write again. Do not add “if anything changes”. That sentence re-opens the door you have just closed, and it turns a clean no into a standing offer they must now remember. A wrong person. If they say “I’m not the right person”, do not ask them for the right person’s name unless they offer it. Asking turns them into an unpaid lead generator and it is the most common way a good first touch is turned into an imposition. Say thank you and stop. An unsubscribe. Stop, permanently, mechanically, and no other list. This is the one place where a system failing is not an inconvenience but a broken promise, and the fourth rule applies to promises made by machines just as much as by people.
All of this is the same rule applied to time. Attention is spent, not claimed. Every message takes something from the reader, and the honest question before sending anything is not “is this message good” but “is this message worth what it costs the person receiving it”. A follow-up with a new finding is worth it. A follow-up with nothing is theft of two minutes by a stranger who has already been given a way to say no. The reason this matters beyond politeness is that it is the only way outbound scales honestly at all. A company that stops when told, follows up only with substance, and never sends the fourth message earns something that cannot be bought: the reader who did not reply last year but who remembers that you were fine about it. Exercise. Open the last outbound message you sent to somebody who did not reply. Count how many of the words were new information. If the answer is under ten, you know why they did not reply, and the repair is not a better subject line.
The sales call is a surface like any other, and it has the same structure. It is included because a page and a call fail in the same way — by talking about the flower before the bee has been recognised. The pattern is four moves and one silence:
Open with their story (“What made you take this call?”). Listen for the feeling words they use, then use their words when you describe the after. State the price once, then go quiet.
”What made you take this call?” — and then nothing. This is the only opening worth having, because it is the only one that produces the material the rest of the call needs. It works because it is a question about them, asked by somebody who is genuinely willing to hear the answer. Then listen for two things specifically: The feeling words. Not the requirements. The words they use when they are not being professional: it’s embarrassing, I dread it, we look amateur, I can’t keep doing this, it’s the not knowing. Write them down while they talk. You will need them in twenty minutes.
The sequence of the bad day. Ask walk me through the last time it happened and let them take as long as they want. The order of events in that story is your recognition line, and it is worth more than any discovery framework.
When you describe the after, use their words rather than yours. If they said I dread the fifteenth, the sentence is so the fifteenth stops being a day you dread. Not so you experience improved cash-flow visibility. This is not mirroring as a technique; it is the minimum courtesy of a conversation, and it is also the single most effective thing available to you on a call, because a person recognises their own phrase instantly and hears it as understanding. The discipline is to resist replacing their vocabulary with the product’s. The temptation is strong because you know the product’s words and they do not, and it feels like clarity. It is not clarity. It is the beginning of the reader no longer recognising themselves.
One number, in full, including anything that will be added. Then stop talking. The silence after a price is the hardest four seconds in sales and the reason people fill it. Filling it with a discount, a justification, a breakdown or a story about value converts a stated price into a negotiation the customer did not open. Say the number, and let them think.
If they ask a question, answer it. If they say nothing, wait. The number has to be allowed to sit in the room for a moment, because a person cannot decide about a price that is still being discussed.
Every call should contain one thing that is mildly against your own interest: a limitation, a case where the product is the wrong choice, a competitor who is better at something. Not as a technique, though it works as one, but because a person who is deciding whether to trust you needs at least one piece of evidence that you are not simply optimising the conversation. The ticketing is not as good as the dedicated tools. If your main problem is the queue and not the tenders, I’d go elsewhere and I’ll say where. That sentence, on the right call, closes more business than any close.
End with one small step, agreed, with a time — the sales equivalent of the door. Not “I’ll send some information”. I’ll send the audit on Wednesday by eleven, and if it looks right we can set a start date. If it doesn’t, I’ll tell you so. Then do exactly that, on Wednesday, by eleven. Exercise. On your next call, say nothing for four seconds after the price. Count them. Then write down the customer’s own feeling words, in their words, in the notes you keep. Those two sentences are worth more than anything else you will hear.
Everything so far has been shown in parts. This chapter builds one page whole, from an empty sheet, for a business that is not the library’s, so that the method can be seen operating on something unfamiliar. The business is invented, and its details are invented with it. The name is Ashgrove Certificates. Two people, a van, and a business doing gas safety certificates for landlords of small rented flats. A certificate is legally required every twelve months; without one, a landlord cannot lawfully let the property, and if something goes wrong the fine is not small. The work is unglamorous, the deadline is a date on a piece of paper, and the customer is usually a person with one or two flats who has forgotten the date. No claim in what follows is real. The method is.
The temptation at this point is to start with a headline. Instead, four questions on an empty page, in this order, and no writing until they are answered.
Whose day is this page about? What happened, in order, on the day it went wrong?
What did it cost them, in which currency? What can they check, in under a minute, without asking us?
The answers do not come from imagination. They come from the collection described in Chapter 19, and for this business there were three sources: the phone calls in the fortnight before a certificate expires, the messages after a booking was missed, and the way landlords talk about the whole subject when they are not being charged for it. The notes, verbatim, from those sources — invented here, but of the shape you would actually hear:
I didn’t know it had run out. It’s the date, I can’t keep the date in my head. The agent asked for the certificate and I had to say I’d look for it. Last time the man came out and the boiler had a fault, so it was two visits and two fees, and nobody told me that could happen. I’ve got two flats and they expire in different months and I only find out when something goes wrong. It’s not the money, it’s being made to feel stupid about it.
That last sentence is the whole page, and it took three sources to find. Note that it is not about certificates at all. It is about being made to feel stupid.
Four candidate feelings are visible in the notes. Anxiety about the date. Irritation about the two-visit fee. Confusion about which flat expires when. Shame, from the last sentence. Applying the four questions from Chapter 20: — Which feeling did the buyer actually have? Shame is the one they said out loud, unprompted, and it is the one the phone calls are full of — the apologetic opening, the “I know I should know this”. — Which feeling does the service actually produce? Not pride, and not expertise. It produces the plain relief of being told, in advance, that everything is in order. That is the removal of shame, which is close enough to be honest and specific enough to be provable. — Which can we prove? The renewal date, held for them, and a named date for the visit. Both checkable in a minute. — Which does this page need? The shame one, because the page is meeting a landlord who has just been asked for a certificate they cannot find. One feeling: the relief of never being the one who forgot. Everything that does not serve it is cut — the parts-and-labour price list, the engineer’s qualifications, the service history, the company’s founding year, the testimonials from letting agents. Each is a real asset. None of them serves this feeling on this page.
The first draft exists to be diagnosed, so it is written fast, in the voice the company actually uses when nobody is watching.
Ashgrove Certificates — Gas Safety Certificates for Landlords We are a family-run business with over fifteen years of experience providing high-quality gas safety certification across the region. Our fully qualified engineers use the latest equipment to ensure your property is compliant and your tenants are safe. Why choose us? — Competitive pricing — Fast turnaround — Friendly, professional service — Fully insured [Get a quote] [Call us] [Contact form]
Diagnosis, petal by petal. Recognition: absent. The first line names the service and the audience and says nothing about the reader’s day. A landlord reading it learns that gas certificates exist, which they knew, because that is why they came. Cost: absent. There is no mention of the fine, the date, or the feeling of being asked for a document you cannot produce. The page’s most powerful material — a legal deadline with a real penalty, and a specific embarrassment — has been left in the notes. After: absent. “Compliant” is a state, not a day. The reader cannot picture Tuesday. Proof: zero. Fifteen years, latest equipment, fully qualified, fully insured: four claims, none checkable, none specific. There is no date, no name, no document, no sample. Door: three. A quote, a call, and a form — and all three ask the reader to start a sales conversation before anything has been given to them. The smallest door on the page is “call a stranger”. Voice: four exclamation-adjacent superlatives in the first paragraph, no exclamation marks (the writer restrained themselves) and no warmth either. The page is written about the company to a category. It is, in short, the page that ninety percent of small service businesses publish, written by somebody competent who did not have a method. Nothing in it is dishonest. It simply does not do the job.
Recognition. From the notes: you didn’t know it had run out, and the agent asked and you had to say you’d look for it. Narrowed, because this page is for the landlord with one or two flats:
The agent asks for the certificate, and you are not sure whether it expired in March or whether that was the boiler service.
Two specifics — a month and a confusion between two documents — and no adjectives. A landlord with two flats recognises the exact flavour of the uncertainty in the first clause. Cost. The currency here is not money, and this is the decision that most writers would get wrong. The fine is real but it is remote; the embarrassment is immediate. So the cost is stated in pride, once, without a threat attached:
The worst part is not the fee. It is being asked a simple question about your own property and not being able to answer it.
One sentence, no consequence appended, no “don’t let this happen to you”. After. Same day, mechanically changed. The mechanism has to be visible, because “peace of mind” is exactly the mood claim this book bans:
We hold the date. Six weeks before it falls, you get a message with the date, the flat and the price, and a visit that you can move with one reply. The certificate arrives as a PDF the same evening, and a paper copy goes to the agent if you tell us to.
Notice what the after is doing: it is a schedule, not a sentiment. The relief is produced by a mechanism the reader can picture — a message arriving six weeks early. Proof. What can a landlord check in under a minute, without asking? Three things exist, and the page should carry all three:
This month’s list of certificates we have renewed, with the street and the date. (A dated, checkable list — not testimonials, a record.) A sample certificate, exactly as it arrives, with the real one’s fields. (The artefact.) Our price, printed in full: £78 for one appliance, £96 for two, and the second visit is free if the boiler fails. (The price, and the honest handling of the two-visit fear from the notes.)
That third item is the strongest proof on the page, and it costs money to publish. It answers the irritation in the notes — two visits and two fees, and nobody told me that could happen — before the reader has to ask, and it converts a fear into a promise the company can afford. Door. One, small, and free: the renewal date check. Not a quote. Not a call.
[Tell me my renewal date — free, one minute]
The reader supplies an address, gets a date, and has received something before spending anything. The quote and the phone number move to the bottom of the page, where the reader who has decided will look for them.
Promise. From the Codex, one line, honoured without argument:
Our promise: if we miss your renewal date, the next certificate is free.
That promise is only publishable if the system holds dates reliably and somebody checks the diary every Monday. Which is the point made in Part Seven: a voice is a set of operational commitments wearing a sentence.
You are not sure whether it expired in March. Gas safety certificates for landlords with one or two flats. We hold the date, tell you six weeks early, and print the price in full. [Tell me my renewal date — free, one minute] [See a sample certificate]
>
The worst part is not the fee. It is being asked a simple question about your own property and not being able to answer it.
>
What we do. We keep the date. Six weeks before it falls you get a message with the flat, the date and the price. You can move the visit with one reply. The certificate arrives as a PDF the same evening, and a copy goes to your agent if you ask.
>
What it costs.£78 for one appliance. £96 for two. If the boiler fails on the first visit, the second visit is free — that is the part most people do not get told.
>
What you can check before you book. This month’s renewals, with the street and the date. A sample certificate, exactly as it arrives.
>
Our promise. If we miss your renewal date, the next certificate is free. [Tell me my renewal date — free]
Nineteen lines. No adjective of praise. Every claim checkable. One feeling. One door, and it is free.
The first draft was 96 words of copy and 4 unverifiable claims. The second is 214 words, 3 claims (a price, a list, a sample) and every one of them checkable by the reader in under a minute. The page is longer and takes less time to read, because there is nothing in it that the reader has to work out for themselves. It is also written entirely from six sentences collected from three sources. The writer invented no feeling at all. That is the whole method, and it can be repeated on any page in any business: collect, choose one, order, prove, open one door. Exercise. Take the page you are dreading. Fill in the four questions from step one without writing a word of copy. Then write the first draft badly on purpose, and diagnose it petal by petal. Do not skip the bad draft. It is the only way to see what the page was going to do on its own.
The rules in Part Three are easy to agree with and hard to obey, because obeying them requires material. This part is about where the material comes from, and how to choose one true thing out of the twenty you will find.
Every rule in this book depends on a supply of real sentences. If you do not have them, you will write plausible ones, and plausible ones are the beginning of the whole failure this book is about. So the collection is not a preliminary to the writing. It is the writing, done in advance, in a different room. There are six sources, roughly in order of value.
Before the pitch, before the requirements, people explain why they took the call. In those ninety seconds they use their own words, because they have not yet learned yours. This is the richest material available and it evaporates within a week, which is why every call should end with one sentence written down while the call is still warm: the phrase they used for the problem. Not a summary. The phrase. It’s the not knowing that gets me. We look like amateurs. I’m doing the job twice. Keep the words, not the meaning.
A support queue is a collection of people describing their worst hour in the vocabulary available to them. Read a hundred tickets and you will find that the same three or four phrasings recur — those are the recognition lines, and they are exact. People say it just disappears, it won’t let me, it’s showing the wrong amount and I don’t know which one to trust. Nobody in marketing would write that. Everybody who has used the product has said it. The method: once a month, take twenty tickets and copy out only the sentences that are not describing a bug but describing a feeling. You will get three or four sentences, and they are worth more than a survey.
The person who has just left is the most honest source in business, because they have nothing to gain and no reason to be kind. Five questions produce almost everything: — What were you trying to do on the day you decided to stop? — What did you say to your colleague about it? — What would have had to be true for you to stay? — What did you buy instead, and what did it say it would do for you? — If you met somebody with the same problem, what would you tell them to do? The fourth question is the one people forget. Somebody who left you for a competitor has already told you what was missing, in the competitor’s language, which is a finished piece of market research obtained at the cost of a phone call.
People preface the most useful question they will ever ask with an apology. Sorry, this is probably a stupid question, but... — and then the question is: what actually happens if the invoice is wrong? The apology marks the feeling. Collect the apologies. Each one is a page heading that writes itself, because a person who felt stupid has told you exactly where the product failed to hold them.
Not your inbox. Their WhatsApp group, their trade association meeting, their forum, the comments under a supplier’s post, the local business group where somebody asks the question at eleven at night and gets six replies by morning. Read for the question rather than the answer. The question, as asked by a person who does not know the vocabulary yet, is a recognition line written by the audience. Read, do not participate. And never quote anybody from these places without asking.
Not invention, recollection. What did you dread this week, and what words did you use inside your own head? Some of the best lines in this book came from the author’s own irritation rather than from a customer, and they work because the irritation was real and specific. The test is the one from Part One: is this feeling mine, and did it happen on a named day?
Whatever the source, the material has to be kept, and the form matters. Not a document of insight. A list of verbatim sentences with the date and where it came from. One line each. No paraphrase, no tidy-up, no writing them into the brand voice. The whole value is in the roughness: the phrase that is slightly wrong is the one a reader recognises, and the moment you smooth it you have written a sentence about a customer rather than a sentence from one. Then use them. A row of sentences you never consult is a library nobody reads. The practice is to open the list before writing any page, find the one that belongs to that page’s reader, and build the recognition line from it, changing as little as possible.
— Anything you would have to pretend not to know. If a sentence was said in confidence, it cannot appear on a page with a name on it, and it cannot appear without one either if the speaker could be identified. — Numbers that have no denominator, however often they are repeated at conferences. — Testimonials written by the vendor and approved by the customer. A quote the customer edited out of politeness is no longer their words, and its evidential value drops to zero the moment anybody notices the register does not match the person. — Anything you intend to improve. If it needs improving it was not their words, and you are back to invention with extra steps.
Exercise. Open your support queue, or your last five calls, and copy out five verbatim sentences that describe a feeling rather than a bug. Put the date and the source beside each. You now have five recognition lines, and they are true.
The collection produces abundance, and abundance is its own problem. A page built from everything you know becomes a page about everything, which is a page about nothing. The rule is: one feeling per page or message, not five. This chapter is about how to choose the one, and it is mostly a chapter about cutting.
Because a reader retains one thing. Not because they are inattentive but because the page is competing with their whole week, and one clear thing survives a week of interference while five things do not survive the afternoon. And because feelings interfere with one another. Relief and urgency do not sit comfortably in the same paragraph: one says this is handled, the other says this is on fire. A page that tries to produce both produces a reader who cannot decide whether to relax or to hurry, and a reader in that state does nothing, which is the most expensive possible outcome for a page that was trying to be persuasive on five fronts.
Four questions, asked in order. The one feeling that survives all four is the one to write. One: which feeling did the buyer actually have? Not which one would be most useful to you. What did they feel, on a named day, at the moment the problem got bad enough to act? The collection from Chapter 19 answers this if it was kept honestly. Two: which feeling does the product actually produce? This is the discipline that prevents over-promise. A bookkeeping product produces relief and order. It does not produce pride in the business or confidence in front of investors, even though those are nice, and claiming them is how cards become untrue. Three: which feeling can you prove? If the feeling is relief, the proof is the free tool that demonstrates the change. If the feeling is certainty, the proof is an audit finding. If the feeling is capability, the proof is a person doing the thing. A feeling with no proof attached is a claim with a mood over it. Four: which feeling does this specific page need? A homepage, a product card, a churn-saving email and an incident notice all serve the same product and need different feelings, because they meet the reader in different states.
The title map in Part Six gives each book its feeling: recognition for The Ledger and the Shopkeeper, earned confidence for Intelligence, Data and Judgement, order for Order Without Custody, calm control for Grounded Cadence, sovereignty for The Sovereign Micro-Enterprise, and capability for the working-paper series. Those are the catalogue-level answers. At page level, each one has to be narrowed further, and the narrowing is where the work is. Take Grounded Cadence, whose feeling is calm control: the rhythm is chosen and can be paused. Now apply the four questions to a page aimed at a small firm that has automated its follow-up and stopped trusting it. Which feeling did the buyer have? Not the abstract one. The one from the call: the shame of hearing about the tender from a competitor. That is not calm-control-absence, it is embarrassment, and embarrassment is a sharper feeling than untidiness. Which does the product produce? Relief from embarrassment, specifically: the Monday list, so that the answer to “did you see the Ngong Road one” is always yes. Which can you prove? The list itself, this week, for their country. Free, checkable, dated. Which does this page need? The embarrassment one, because the page is aimed at firms who have already missed something and are looking for a way not to miss the next one. So the page’s one feeling is the relief of never being the last to know, and everything that does not serve it goes: the margin calculator (a different feeling, control of sums, which is a different page), the invoice features, the company story, the team photographs. Four feelings cut, and the page gets shorter and better in the same edit.
They are not lost. Each cut feeling is a page waiting to exist, and the discipline is to write down the cut list rather than discard it. — Cut: the margin worry. Becomes a page about the calculator. — Cut: the pride of a tidy office. Becomes an about page section, or a different product’s card. — Cut: the relief of a fast reply. Becomes the first-touch email’s after, not the landing page’s. — Cut: the delight of a good interface. Becomes nothing. Some feelings are the writer’s, not the buyer’s, and this one usually is. That last line is worth sitting with. A good proportion of what writers want a page to feel is the writer’s own pleasure in the product, and it has no counterpart in the reader’s week. Interface delight is the commonest example. It is real, it matters to the people who built it, and it is not why anybody buys.
There is one legitimate place for a second feeling: the page of the reader who has already decided. Pricing pages, documentation, comparison pages for people in a trial, and post-purchase emails can carry a second, quieter note — confidence, reassurance, welcome — because that reader is no longer deciding whether to care and is now deciding whether they were right. Even there, one feeling leads and the second is subordinate. What is never permitted is two feelings of equal weight on a page where the reader is deciding. That is not richness. It is a page with two doors.
Exercise. Write down the five feelings your last landing page tried to produce. Then choose the one the buyer actually had. Then, beside each of the other four, write the page it belongs to instead. You have just planned your content for a quarter.
Everything in this book converges on one moment: a person who does not care reads the page for four seconds and decides. That person is the only audience that matters and the hardest to arrange, because the people who will read your draft for you are colleagues who care, or friends who are kind. So the test has to be designed deliberately.
Find somebody who is not in your industry, has no stake in your success, and is tired. The tiredness is not a joke. A rested stranger will read anything and find it interesting. A tired one will reproduce, fairly exactly, the experience of your actual reader on a Thursday at nine in the evening. If you can find somebody who has the problem your product solves, better still, and better again if they have never heard of you.
Four questions, in this order, with no explanation and no defensiveness while they answer. One: what does this company sell? If the answer takes more than one sentence, the page has failed at its first job.
Two: is it for you? This produces the most useful refusal available. If a stranger says “I don’t think this is for me”, and they were right, the page works — it has filtered correctly. If they say it and they were the target reader, the recognition failed. Three: what would be different about your day? The after test. If they can describe the change in a day, the after worked. If they say “it would be easier” or “more efficient”, the after is an attribute and not a scene. Four: what would you have to do next, and what would it cost? The door test. If they cannot say what happens when they press the button, the button is a label rather than a door. A fifth question is optional and often the most revealing: what does this company remind you of? If the answer is “every other software company”, the page is written at category level and recognition has not happened.
The most useful results are not the answers but the hesitations, and there are four to listen for. The pause before “what does it sell”. Means the sub-line described a category rather than a product. The phrase “I think it’s for…”. The hedge means the recognition line narrowed to a class rather than a person. You want “this is for me” or “this isn’t for me”, both of which are correct outcomes, and nothing in between.
The word “easier”. The after has failed. Easier is what people say when they cannot picture anything, and it is a polite way of reporting nothing. The question “what does it actually cost?” The pricing or entry cost is not where the reader looks, or is not there at all. Entry cost belongs in the hero sub-line for exactly this reason.
Every organisation has a group of people whose job, on some days, is to approve the page. They are the worst possible test subjects and the hardest to avoid, because they will read carefully and tell you what the page means rather than what it communicates. Two practices help. Test before the review, not after. Bring the stranger’s four answers to the review meeting as evidence rather than opinion; a meeting arguing about a headline is a meeting about taste, and a meeting looking at a tired person’s confusion is a meeting about a problem. Do not attend the test in a helpful mood. The temptation is to explain, and every explanation you give is evidence of something the page failed to say.
One more thing worth doing, cheaply, once. Take a page you published a year ago and read it as though you had never seen it. Then ask the four questions. Pages age badly in a specific way: the recognition line, which was sharp when the product was young, becomes a description of a category once the market fills with similar products. The tests in this book should be run on your own archive at least once a year, with the same coldness you would apply to a competitor. Exercise. Arrange one tired stranger this week. Four questions, ten minutes, no explaining. Write down their exact answers, including the hesitations. Then change one thing on the page. One, and see whether the next stranger’s answer changes.
There is a kind of honesty that costs nothing, and there is the kind this chapter is about: telling somebody, in public, that your product is not for them. Every rule in this book has been about making the right reader feel recognised. This one is about the opposite, and it is the rule most companies cannot bring themselves to keep, because the arithmetic looks wrong. It is not wrong. It is the arithmetic that makes the rest work.
A page that is for everybody converts at the rate of a page for nobody, because no reader can find themselves in a description of anyone. The narrow statement — for landlords with one or two flats — is more attractive to landlords with one or two flats than the broad one, and it is genuinely useful to landlords with forty, who now know to look elsewhere and save themselves an afternoon. Three things follow, and all three are measurable. The wrong readers leave earlier, which costs less. A wrong reader who leaves at the hero costs you nothing. A wrong reader who buys, fails to get value, cancels, and writes about it costs you a great deal. The page’s job includes preventing the second thing.
The right readers believe you more. A page that disqualifies somebody has demonstrated, in the only way that can be demonstrated, that it is not simply optimising for the maximum number of clicks. Credibility is produced by visible costs, and this is one. Support gets quieter. A large fraction of unhappy customers are people who were never going to be happy and were persuaded anyway. Each one you decline politely is a complaint that does not happen, a refund that does not occur and a member of staff who is not worn down on a Tuesday.
The line has to be specific, calm and free of contempt. There are three forms and they do different jobs. The scope line. Stated as a fact about the product: For landlords with one to three properties. Above that, the paperwork needs a system with multiple users and we do not have it yet. The word “yet” matters. It is honest about the limit and honest about the direction. The wrong-problem line. Stated as a fact about the reader’s situation: If your tenants pay by standing order and have never missed one, this will not earn its price. Most of what we do is chase people who are late. The alternative line. Stated as a fact about the market: If you need certification for commercial premises, go to a firm that does commercial; the checks are different and we would be learning on your time. This is the most generous form and the most effective, because it hands the reader a real recommendation rather than a nudge. What all three avoid is the insincere version: We might not be the right fit for everyone! That sentence disqualifies nobody while appearing humble, and readers hear the emptiness in it. A disqualification that costs you nothing is not a disqualification, it is a courtesy noise.
On a long page, the honest form is a short two-column block, placed after the proof and before the door, where a reader who is still deciding can check themselves against it.
This is for you if you own one to three rented flats, the certificate is in your name, and the date has surprised you before.
>
This is not for you if a letting agent handles your compliance, or you need commercial premises, or you want somebody to chase the tenants as well. We do none of those, and we will say so on the phone rather than after you have paid.
Two sentences, and the second one is the whole credibility of the page. It also does something the rest of the copy cannot: it proves the writer is willing to lose a sale, which is the one claim about honesty a company cannot make about itself.
The objection is always the same: this will cost us sales.
It will cost you some. That is the point, and the honest answer is not “it will pay for itself in no time”, which is exactly the kind of unverifiable claim this book bans. The truthful answer is narrower. Disqualification costs you the sales that would have churned anyway, and it buys you three things you cannot otherwise get: a lower cost of support, a higher proportion of customers who stay, and a page that a person can believe. Whether that trade is worth it depends on your churn and your support load, and you can find out — which is a different thing from being told by a book that it always works. What can be said without qualification is this: a business that cannot name the customer it does not serve has not chosen a customer at all, and is therefore not in a position to write a page about one.
A related discipline, harder to hold. Some readers will recognise themselves in a page and then be wrong — they have the problem, but they lack the thing that makes the solution work. The shop with no bookkeeping habit. The firm with no one to send the applications. The landlord whose agent does, in fact, handle it. The product cannot fix that, and the page should not pretend it can. The honest version names the precondition plainly: this works if the till is used for every sale. If some sales go through the phone and get written on a pad, the number at the top of the page will be wrong, and you will stop trusting it in a fortnight.
That sentence is unattractive, and it is the difference between a customer who succeeds and a customer who leaves. Every page has a precondition; most pages hide it; hiding it is how support queues are built. Exercise. Write three sentences: who this is for, who it is not for, and what has to be true about the reader’s week for the product to work at all. Put the second and third on the page. If you cannot write the second, you have not yet decided who the page is addressing, and no amount of headline work will supply the decision.
Each title exists because a particular sentence was said by somebody tired. These six chapters are short on purpose: a book’s whole voice fits in a page if the feeling has been chosen honestly, and if it needs more than one, the feeling probably has not been chosen. A press has an advantage here that a software vendor does not. Nobody arrives at a book expecting a demonstration. They arrive expecting to be told the truth by somebody who did the work, and they will forgive a slow paragraph far more readily than they will forgive a claim.
The bee’s pain, in their words. We built the shop, the customers came, and the money still did not add up.
The feeling we serve. Recognition. Not insight, which flatters the writer, and not hope, which the reader has been sold already. Recognition is the specific relief of reading a description of your own week in somebody else’s words, and finding that it has a name and a chapter. The book’s argument is that a small business fails in five recognisable classes, and that almost nobody can see which class they are in from inside it.
The free nectar. The first chapter and the failure taxonomy, free, on the library’s own page, with no form in front of it. A reader can learn whether this book is about their problem before deciding to buy it. That is the whole of the free nectar, and it is enough, because the book’s value is not the diagnosis — it is the working out that follows.
A shopkeeper reading the contents page and finding the sentence that describes their Tuesday. The line below is constructed, as every quoted line in this book is:
Three months in, the shop is busy and the account is empty, and I cannot tell you which of the two numbers is lying.
The cost, in their currency. Not the money lost, which is countable, but the private suspicion of being bad at the thing you are visibly doing every day. That suspicion is expensive and it is never invoiced.
A name for the failure, and with it the dignity of a problem that has a shape: this is a pricing class, or a reconciliation class, or a follow-up class, and each one has a fix that does not require you to become a different person.
Where the voice goes wrong: It starts reviewing the book instead of doing the work — “a must-read for entrepreneurs”, “essential”. A blurb is the writer asking the reader to trust the writer’s judgement instead of their own. The page should contain one scene from the book and the price.
The bee’s pain, in their words. I have read about this for a year and I still cannot do any of it.
The feeling we serve. Earned confidence. Not excitement about a subject, and not the vague reassurance of “anyone can learn”. Earned confidence is the state of a person who has finished a chapter, done the exercise, and found that they can now do a thing they could not do this morning. It is a small feeling and it is the only honest one a textbook can produce.
The free nectar. The learning outcomes and one complete worked chapter, free. A reader can judge a textbook in ten minutes by reading how it teaches a single idea, and withholding that is how a publisher tells a buyer they do not trust their own book.
A reader who has watched the videos, followed along, and then opened a blank file and found nothing there. The quoted complaint is constructed:
I understand it while I am watching. The moment the screen is blank, I have nothing.
The cost, in their currency. The accumulation of evidence against yourself. Every half-finished course is a small document about your own character, and the stack of them is heavier than the money they cost.
One chapter finished, one exercise done badly, and the discovery that the second attempt is easier. The page’s job is to promise exactly that and not a career.
Where the voice goes wrong: It promises transformation, or an income, or a career change. A textbook that sells a future is selling the thing this book spends Part Two removing. Teach the sample, state the page count, state the price, and stop.
The bee’s pain, in their words. We cannot tell which payment belongs to which order, and we are afraid to touch the money.
The feeling we serve. Order. This is the one title in the set whose feeling is not primarily relief, and the difference matters. The reader is not only worried; they are somebody who runs a business on rails they do not control, and what they want is not comfort but a system whose arithmetic can be recomputed by a stranger. The paper’s design rule — the operator never holds the money — is a promise the reader can check, which is why it belongs on the page in the first sentence rather than in a footnote.
The free nectar. The design rule and the reconciliation arithmetic, free, in the sample chapter. Both are useful on their own, whether or not the reader ever buys the paper, and one of them will change how they build.
A specific afternoon: two customers paid the same amount and nobody can say which order either payment settled. The quoted line is constructed:
Two payments came in for the same amount and I still do not know which one was the groceries and which one was the gas.
The cost, in their currency. Not the sum involved, which is usually small. The cost is the dread of being asked to account for money that moved through hands that were not supposed to hold it.
Every order carries an amount unique to the cent, so the payment identifies itself, and the reconciliation is a subtraction a stranger can repeat. The after is dull and that is the point: it is the dullness of a thing that no longer needs attention.
Where the voice goes wrong: It drifts into the language of fintech — “seamless”, “frictionless”, “bank-grade”. Those words are the trade’s, not the reader’s, and they describe a feeling the vendor wants rather than a fact the reader can verify.
The bee’s pain, in their words. The systems run all night and I no longer know what they are doing.
The feeling we serve. Calm control. The reader has automated something and discovered that automation has a tempo of its own, and that the tempo was never chosen — it was whatever the default interval happened to be. Calm control is the state of a person who knows what runs, how often, and what stops it.
The free nectar. The cadence worksheet and the pause rule, free. A reader can set one interval correctly this afternoon with what is given away, which is the test the book applies to every promise it makes.
The morning after an unattended night, opening a log to find that something ran four hundred times. The quoted line is constructed:
It did more than I asked it to, and I found out from a customer.
The cost, in their currency. Not the compute. The cost is the loss of the ability to say, truthfully, what your own business did yesterday.
A stated cadence, a pause that a person can reach, and a log short enough to read. The reader gets their own account of the night back, which is what they actually lost.
Where the voice goes wrong: It sells autonomy. The word is the vendor’s fantasy and it is the opposite of what the paper argues, which is that a rhythm nobody chose is not freedom but an unattended machine.
The bee’s pain, in their words. The platform changed its terms and my business changed with it.
The feeling we serve. Sovereignty. Not independence, which is a slogan, and not scale, which is somebody else’s ambition. Sovereignty is the ordinary condition of owning the thing you sell through: your domain, your list, your files, your ability to move. It is a quiet feeling and it is only visible when something goes wrong elsewhere.
The free nectar. The architecture chapter and the failure taxonomy, free. The taxonomy alone is worth the visit, because it lets a reader name which dependency they are standing on before it moves.
The week a platform changed a policy and a business spent it rewriting things that had worked. The quoted line is constructed:
Nothing about my business changed, and everything about it changed.
The cost, in their currency. Not revenue. The cost is the low background fear that the ground is not yours, which makes every plan provisional and every hire a gamble.
A named dependency list, a rail you can point at, and an export that already works because it is the same rail. The after is a sentence the reader can say to a bank.
Where the voice goes wrong: It becomes a manifesto about platforms and freedom. The reader does not want a politics; they want an architecture drawing and the arithmetic of what it costs to own it.
The bee’s pain, in their words. I read the research and I still cannot apply it to my own case.
The feeling we serve. Capability. The four working papers — Order Without Custody, Grounded Cadence, The Sovereign Micro-Enterprise and Constraint-Native Business Model Design — are sold as a series because they share one property: each one hands the reader the instrument it used. A paper that only reports is a story about somebody else. These are meant to be run.
The free nectar. Every paper’s method section is free to read before purchase, and the method is the paper. What is sold is the typeset edition, the worked case and the corrections; what is given away is the ability to check the argument while standing up.
A reader who agrees with a paper in principle and cannot find the first step in it. The quoted line is constructed:
I understand the argument. I do not know what to do on Monday.
The cost, in their currency. Not the price of a paper. The cost is the quiet conviction that research is for other people’s businesses, which is the reason good papers go unread and bad consultancies prosper.
One instrument, run on the reader’s own numbers, producing an answer they can disagree with. Disagreement is the proof it worked.
Where the voice goes wrong: It sells the series as a set to collect. A paper is not a bundle item; each one has to stand alone at the price printed on it, and the page has to survive a reader who buys exactly one.
Six titles, six feelings, one method. In each case the voice was found by asking what the reader actually had, on a named day, in their own words, and then choosing the smallest honest feeling that fitted it.
A voice survives a document and dies of an ordinary Wednesday. This part is about the ordinary Wednesdays: the writer who has just joined, the system that sends at two in the morning, the page that failed its own test.
The rules in this book were written for one person writing one page. Organisations do not work that way, and the failure mode is predictable: the founder writes beautifully for two years, hires a marketer, hires a freelancer, buys a scheduling tool, and within eighteen months there are four voices and a machine, all technically compliant, none recognisable. Nothing about that decline is a failure of intelligence or goodwill. It is what happens when a method lives in a document rather than in a process, so this chapter is the process.
You do not need a review committee and you do not need a style council. You need one gate that any member of staff can apply to their own work in about thirty seconds, and it has two questions. Question one: whose day is in the first two lines? If the answer is “nobody’s” or “anybody’s”, the piece is not ready. It does not matter how good the rest of it is; the first two lines are the whole test of recognition. Question two: could this go, unchanged, to somebody else? If yes, it is not finished. For a page, this means: could this sentence appear on a competitor’s site unchanged? For a message, this means: could it be sent to a second recipient? Two questions, no jargon, applyable by a person on their first day. Everything else in this book is explanation for people who want to know why the gate exists. The gate itself fits on a card, and should be on one.
The hardest governance question is not what the rules are but who is allowed to overrule them, because in most organisations the answer is “whoever is most senior on the day”, and that produces drift with a clear conscience. Three roles, and the authority is deliberately narrow. The writer owns the draft. They must have both questions answered before it goes to anyone else. Reviewing an unfinished draft wastes a meeting and teaches everybody that the gate is optional. The reviewer owns the two questions only. They may not change the feeling, the price, the promise or the strategy. They check the gate and the banned lexicon, and they are allowed to say no on those grounds alone. A reviewer who starts rewriting the argument has become a second writer, and two writers produce the fourth voice. The founder owns the Codex. Promises, entry prices and the list of what the company will and will not claim. Nobody else may add a promise to the Codex, and nobody may make one up in a draft and hope it survives the approval.
A voice guide cannot be read. It can be exercised. So the onboarding is three exercises, and the writer’s first paid task should be the first of them, on real work, with real stakes. Exercise one: the repair. Give the new writer your three worst pages and ask for one repaired section each, chosen by them. The repairs do not ship. The purpose is to find out, in a day, whether their instinct is subtraction or decoration — and to have a conversation about the two questions while there is still time. Exercise two: the collection. Send them to the support queue and the last five sales calls to collect five verbatim sentences. Then ask them to build one recognition line from one of them, changing as little as possible. This one exercise predicts most of what you need to know, because it reveals whether they trust the buyer’s words or their own. Exercise three: the stranger. Have them run the tired-stranger test on somebody they recruit themselves, and report the four answers rather than their opinion of the page. If they come back with taste rather than evidence, you have time to say so. The system, and what it is allowed to do Systems now write: subject lines, product descriptions, support replies, follow-ups, and whole pages. There are three rules and they are simple to enforce, which is useful, because enforcement is where such rules usually die. Rule one: a system may look; a person must read. A tool may scan a site and produce a finding, query a tender database, or flag an overdue invoice. Every output that goes to a stranger is read by a person first, in full. Not skimmed for errors — read, because the question is not whether it is correct but whether it is recognisable. Rule two: no system generates a feeling it did not observe. A model can write a good sentence about a situation that has been described to it. It will write a hollow sentence about a situation it has been told to assume. Every generation needs an input that came from the collection: a ticket, a call, a scan, a notice. No input, no sending. Rule three: machinery failures are promise failures. If the unsubscribe breaks, the fourth rule is broken, and it is broken invisibly, which is worse. So the mechanical promises get tested on a schedule, like a fire alarm. Once a quarter, unsubscribe from your own list. Once a quarter, reply stop. Once a quarter, buy something as a customer and watch the promise arrive.
You can hold a voice with one number, and it is not the number of pages published. It is: how many pieces went out last month that a second recipient would have made wrong? The target is zero, and the number is only obtainable by keeping the drafts and pulling three at random at the end of the month. That is the audit, and it is deliberately small. A company that can say honestly that nothing went out last month that could have been sent to somebody else is a company whose voice is not a document anymore.
They will. The rule is broken roughly every week, and the response to a broken rule decides whether the rules survive. The principle: judge the pattern, not the sentence. A single banned word in a busy week is nothing. A writer who has used banned words in six consecutive drafts has a different understanding of what the company is doing, and that is a conversation worth having in private and in time. Conversely, a public reprimand about a phrase teaches the whole team that rules are for the benefit of the person enforcing them, which is the fastest possible way to make them decorative. And one practice is worth more than all the correction: when something goes out that fails the gate, name it yourself. The founder or the marketing lead says, in front of everybody, that one was mine and it was wrong, here is why. A company that only ever examines other people’s failures will not have its own examined for long, and the whole methodology depends on the examination being ordinary.
This book has argued that a page can be beautiful, accurate, feature-complete and still be dishonest — not because it lies, but because it makes the reader do work that only the truth can do for them. There is one final instrument in the set, and it is the one to keep for last because it is the least flattering. Coldread asks a single question of whatever is in front of it: this page reads beautifully — is what is behind it real? That question, asked in the context of a product, is a diligence question. Users of an audit will recognise the shape of it: the demo works, the numbers are impressive, and then somebody asks what the mechanism is and the room goes quiet. What we find, when we audit software, is almost never a lie. It is usually a demonstration that contains more than the product does — a curated dataset, a deleted error state, a flow that exists only in the demo build. Turned on your own marketing, the question becomes: this page reads beautifully — is what is behind it real?
Audit one: every claim has a source. Take a page and number every factual assertion: each number, date, quantity, comparison, result and named customer. Beside each one, write where it came from and when it was last checked. Claims with no source are removed, and removing them is the audit passing, not the audit failing. A page that loses thirty percent of its claims and gains a hundred percent of its credibility is the normal outcome. Audit two: every promise has an owner. List every promise on the page, in emails, in the product itself — including the small ones nobody notices, like “we reply within a day”. For each, name the person who will honour it and the system that makes it happen. A promise with no owner is a lie that has not happened yet. This is the Codex, checked rather than written, and Part Eight contains it. Audit three: every page has a bee. For each published page, name the one person it was written for. Not a segment — a person, described in a sentence, who has a specific week. If a page’s bee cannot be named, the page was written for a committee, and it will do what committee pages do, which is be ignored by everybody except the committee.
Six findings, in the order they most often turn up in real work. One: the pages written for a category. Most companies have two or three of these, usually the homepage, and everybody knows they are bad and nobody knows what to replace them with. The audit gives the answer: the bee was never named, so write the bee first and the page second. Two: the claims that were true eighteen months ago. A result from an earlier version of the product, still on the site, no longer checkable. These are the most dangerous claims in circulation, because nobody is lying and nobody is responsible. Three: the promises with no mechanism. We reply within a day — with no rota, no alert, and a support inbox that one person reads when the mood takes them. The audit does not ask you to remove the promise. It asks you to build the rota, because the promise is the asset and the rota is what it is made of. Four: the free things that are not useful. Lead magnets nobody would download if they did not have to give an email. The audit’s finding is not “make it better” but “this is not proof of anything, and it makes your proof look like bait”. Five: the scarcity that nobody counts. A countdown, a “last few seats”, a quote that “expires Friday” and does not. Each of these is a small, permanent withdrawal from the same account, and the audit reports the balance. Six: the one nobody wants to find. There is always a page somewhere that the team knows is overstated, and it survives because it converts. The audit asks the only question worth asking about it: what would have to become true for this page to be honest, and how much would it cost to make it true? Sometimes the answer is a small engineering change, and the page can be kept. Sometimes the answer is that the honest version of this page converts worse, and the choice is now explicit rather than unspoken. That is the most valuable output the audit produces, because it replaces a silence with a number.
Everything here is a method for making claims checkable. It cannot make them true. If the product does not do the thing, no arrangement of the five petals will help, and a well-written page will simply produce disappointed customers more efficiently. This is worth stating plainly because a voice guide, read enthusiastically, can create the impression that the writing is the work. The writing is the smaller half. The larger half is the company deciding, in a meeting, that it will make the free tool actually free, that the promise will actually be honoured without an argument, that the audit finding will actually be published even when it is unflattering. Those are decisions with budgets attached, and no sentence written beautifully substitutes for them.
The flower does not chase the bee. It opens, it is honest about what it holds, and it is easy to find. You are the flower. Your customers are the bees. Every word you publish either opens the flower or closes it. That is the whole of it, and it was whole at the beginning. What this book has added is only the working out: how to find out what you actually hold, how to describe it without inventing anything, how to prove it before you ask for money, and how to keep doing that when the week is difficult and the meeting wants a countdown. There is no trick at the end, which is the point. The method is available to anybody who is willing to be specific about their own product in public, and it is the only method that gets easier the longer you use it, because the collection of real sentences grows and the temptation to invent shrinks. Open, be honest about what you hold, and be easy to find. The bees will come.
Everything in this part is meant to be copied, printed and marked with a pen. Nothing here is new; it is the rest of the book with the explanation removed.
The list below is not a list of forbidden words in the sense that a word will ruin a sentence. It is a list of words that are almost always doing the work that a fact should be doing. When you reach for one, the reliable move is to stop and find the fact.
BANNED LEXICON Word or phrase Why it fails What to write instead revolutionary asserts scale nobody measured the specific change, with a date game-changing a claim about the future what changed in somebody’s week unlock metaphor hiding a mechanism what the reader does now that they could not before leverage jargon for “use” use seamless asserts an absence of problems what does not happen anymore, specifically cutting-edge, best-in-class comparisons with no comparison a checkable fact, or nothing 10x, skyrocket numbers with no denominator the measured figure and its source guaranteed results a promise you cannot honour the promise you can honour limited time urgency you control the real deadline and who set it trusted by thousands a plurality with no list the count, the names, or the singular world-class nobody can check it the artefact the reader can look at solutions a noun nobody can photograph the thing itself
Word or phrase Why it fails What to write instead empower feeling claimed, not shown the day, after effortless denies the reader’s effort how long it actually takes state-of-the-art an assertion of quality the spec, or the sample “we’re excited to announce” the writer’s feeling, not the reader’s the news, in the first word “valued customer” flattery that costs nothing you, your team, your Monday “just checking in” a message with no new thing the second finding “Don’t miss out” pressure you created the date it closes and who set it exclamation marks in body copy asks a sentence to be louder than it is a full stop
How to use the list. It is a search-and-replace aid, not a doctrine. On a long page, run a find for each term, and for every hit write the fact that the word was standing in for. If there is no fact, the sentence was decoration and the honest repair is deletion. The list should also be run over emails and product copy before sending, because the lexicon decays fastest in systems, where nobody is reading with care.
THE NINE RULES OF THE HONEST FLOWER 1 · Name real feelings only. Describe a recognisable situation; never invent a customer, a quote, a number of users or a result. 2 · Scarcity must be true. A limit somebody chose, enforced by a system that refuses the next person. 3 · Urgency must be theirs. A tender deadline, a renewal date, a season. Never a clock you control. 4 · Every promise is a debt. Honour it without argument, price it in, and keep it on the Codex. 5 · Warmth without flattery. One person, as a person: you, your team, your Monday. No “valued customer”. 6 · Specific beats clever. Concrete nouns, short sentences, one idea each. The container sat two days longer. 7 · Calm is the luxury. No exclamation marks in body copy. Room around the sentences. Low temperature, high attention. 8 · Respect the no. An easy, honourable no in every message, in the sender’s own voice, honoured permanently. 9 · No bulk, ever. One recipient, one observed thing. If it could go unchanged to two people, it is not finished.
Run this before publishing anything. Thirty seconds is the honest estimate if the answer to the first question is no, because at that point the rest is a formality.
BEFORE YOU PUBLISH 1 · Would the buyer recognise their own day in the first two lines? 2 · Is there exactly one feeling, and is it real? 3 · Is every number, notice, finding and claim checkable right now? 4 · Is there something free and useful before the price? 5 · Is every promise one I will honour without argument? 6 · Is there one door, and is “no” easy? 7 · For a message: could it be sent unchanged to anyone else? If yes, it isn’t finished. 8 · No banned words, no exclamation marks in body copy. THE TWO-QUESTION GATE (FOR REVIEWERS) 1 · Whose day is in the first two lines? 2 · Could this go, unchanged, to somebody else?
LANDING PAGE HERO Headline: the feeling or the moment of loss, in the buyer’s words. Sub-line: what the product does, for whom, at what entry cost. Free or low-risk entry goes here. Two buttons: the small door (free) and the proof (sample or free tool). Nothing else in the hero. “SOUND FAMILIAR?” CARD (ONE PER AUDIENCE) Before (red): the bad moment, specific and short. With [product] (green): the same moment, solved. The door for that audience. PRODUCT CARD [Name] · [price] [Feeling line: the after, in one sentence] For: [who, plainly] You get: [3–5 concrete items] When: [time you commit to] Our promise: [the risk reversal you will honour] [One button]
FIRST-TOUCH EMAIL Subject: [their name/firm/site] + the specific thing you noticed Hi [Name or “there”], [Recognition: how you came across them, truthfully, in one line.] [Proof: the one specific, checkable thing, with the link.] [The after, in one sentence: what that could mean for their week.] [The small door: a free thing they can use without replying.] [The easy no: “If this isn’t useful, ignore it, or reply stop and I won’t write again.”] Courage Garikai · [venture] · [postal address] · [unsubscribe] FOLLOW-UP (DAY 3, THEN DAY 7, THEN STOP) One or two lines with something new: a second finding, a new notice, a free tool. “Just checking in” with nothing new is banned. REPLY TO AN INTERESTED BEE Thank them, briefly. Answer exactly what they asked, in the first line, in full. Offer one next step with a named time. Then stop.
SALES CALL Open with their story: “What made you take this call?” Listen for their feeling words; use their words when you describe the after. State the price once, then go quiet. End with one small step, agreed, with a time. INCIDENT NOTICE What happened, with times. Who was affected, and how. What you have done, in the past tense. What happens next, with a date. What the reader should do, usually nothing. One sentence acknowledging what it cost them. Then stop.
The rule: a promise is a debt. This is the list, and it must be current. Every entry needs an owner, a mechanism and a cost. An entry with no mechanism is not a promise, it is an intention, and it belongs nowhere near a page.
| Promise | Where it is made | Who honours it | Mechanism | Priced in? |
|---|---|---|---|---|
| The sample chapter is the actual book, set the same way | Every product page | Press | Sample file generated from the same source as the edition | Yes |
| Every price is the price, with the currency named | Product pages, receipts | Whoever sets prices | One price table, one source, checked before publish | Yes |
| Openly licensed titles say so, and point at the free source | CC BY 4.0 listings | Press | Licence line printed inside the file itself | Yes — in the copy, not the margin |
| A sold copy carries the buyer’s name in its footer | Delivery | Delivery service | Stamped at download; the download is logged | Yes — in the file |
| A correction reaches everybody who bought the edition | Buyer shelf | Press | Correction note on the shelf, no marketing attached | Yes — one message, no list |
| A refund on request, before or after download | Product pages | Whoever answers the mailbox | Full amount, no interview; written into the policy | Budgeted at 5% |
| Author-review editions are labelled as such | Every listing | Press | Edition line on the cover and in the file | Yes — it costs sales and stays |
Reviewing the Codex. Once a quarter, three questions per line: is it still advertised where the table says, is the mechanism still working, and can we still afford it? A promise that has drifted out of affordability is a debt accruing interest, and the honest move is to stop making it aloud rather than to honour it badly.
For Chapter 19. Keep one file. One line per sentence. The roughness is the value, so do not tidy.
VERBATIM COLLECTION # The sentence, verbatim Source Date Where it can be used call, first 90 seconds / ticket / churn / apology / forum / own week page, email, card
For any message about to be sent, and for three messages pulled at random at the end of every month.
MESSAGE AUDIT Recipient, named: ______ The one thing I observed about them: ______ Where I saw it, and when: ______ Whose words are in the first line? ______ Could this go unchanged to a second recipient? Yes / No If yes — do not send. The free thing, usable without replying: ______ The no, in my own words: ______ Unsubscribe present and working: ______ Banned words checked: ______ Exclamation marks in body copy: ______
The whole company’s feelings on one page. This is the table to keep on the wall, because it is the answer to “which feeling is this page for” before anybody starts writing.
| Title | The reader’s pain, their words | The feeling served | The free nectar |
|---|---|---|---|
| The Ledger and the Shopkeeper | Three months in, the shop is busy and the account is empty. | Recognition: the failure has a name and a chapter. | Chapter one and the failure taxonomy, free. |
| Intelligence, Data and Judgement | I understand it while I am watching. With a blank file, I have nothing. | Earned confidence: the material is teachable. | The learning outcomes and one full worked chapter. |
| Order Without Custody | Two payments for the same amount and I cannot say which is which. | Order: the money is never held, and the arithmetic can be re-run. | The design rule and the reconciliation worksheet. |
| Grounded Cadence | It did more than I asked, and I found out from a customer. | Calm control: the rhythm is chosen and can be paused. | The cadence worksheet and the pause rule. |
| The Sovereign Micro-Enterprise | Nothing about my business changed, and everything about it changed. | Sovereignty: the rail is yours to point at. | The architecture chapter and the failure taxonomy. |
| The working papers | I understand the argument. I do not know what to do on Monday. | Capability: you can run the instrument yourself. | Every method section, free, before purchase. |
| Coldread | It reads beautifully. I am not sure what is behind it. | Certainty before the reader finds out. | The fifteen-minute call and the published sample finding. |
For the day when none of the above helps and a page has to be written.
STARTING FROM NOTHING Whose day is this page about? One person, a sentence. What did they do, in order, on the day it went wrong? What did it take from them — money, sleep, pride, weekends, trust? Choose one. What is the same day like afterwards? Name the change, mechanically. What can they check, in under a minute, without asking me? What is the one door, and what does it cost them to walk through it? What happens if it is wrong? Write the first two lines. Then stop and read them.
Colophon
Set in Georgia for the text and headings, with a single monospaced face for the templates, in the belief that a book about plain writing should not be set in a typeface that is showing off. Trim size 6 × 9 inches. Text measure 4.16 inches, which is a long line for a book and a short one for a screen, and it was chosen deliberately: it is the measure at which a paragraph has to be about one thing, because a paragraph containing two things becomes visible at this width. Chapters begin on a new page. Part titles stand alone. Exercise paragraphs are set as body text rather than boxed, because an exercise that looks like a form gets filled in reluctantly, and one that looks like a paragraph gets read. The book is set to be read at a pace of about thirty pages an hour, which is slower than a business book and faster than a manual. That was the target because that is the speed at which somebody can hold an argument in their head and check it against their own week.
A note on what this edition is
This is version one point zero of a guide that will be wrong about something. The first version of the Codex was wrong about emotion, which is why the first chapter of this book admits it, and the honest expectation is that a later edition will have to admit something else. The parts most likely to need revising are the two that depend on the world rather than on the method. The venture voices in Part Six are tied to products that will change. The legal furniture around consent and unsubscribe will change as the rules around it change, and it should be checked by somebody qualified before this book is used as a compliance document, which it is not. The method itself — collect the real words, choose one feeling, order the five petals, make the proof checkable, open one door and respect the no — is unlikely to need revising, because it was not invented here. It is what a good shopkeeper does across a counter, and the only thing this book has added is a way of doing it at a distance, in writing, at scale, without letting the words drift into the hands of whoever is in the meeting.
Where the promises live
The Codex in Part Eight is the part of the book that must be kept alive. Every promise printed anywhere in this company’s pages should have a row there, an owner, a mechanism and a cost. When a promise on a page and a promise in the Codex disagree, the Codex governs and the page is amended that day. A promise that has drifted out of affordability should be retired in public, in the same calm voice in which it was made. That is unpleasant and it is the only version of the transaction that leaves the company worth trusting afterwards.
The last word
Everything in this book is a way of paying attention, written down so that attention can be paid by more than one person. The rules are not the point. The point is that somebody, somewhere in your week, is tired and specific, and the whole of this craft is to describe their Tuesday accurately enough that they believe you about Wednesday. Open, be honest about what you hold, and be easy to find.