Most technical founders hold two beliefs at once: that marketing is what you fall back on when the product is not good enough to sell itself, and that it needs a personality you do not have. Both are wrong, and the second is the expensive one: it stops you building something you could have built. Marketing is the work of making it easy for the right stranger to find you, understand what you do, believe it works, and act. That is a four-stage system, debuggable by exactly the sort of person who wonders where the instrumentation goes.

Four jobs, four places it breaks

Everyone who eventually pays you clears four gates, in order: nobody trusts you before understanding you, or understands you before finding you. Each gate fails differently, and the symptoms are distinguishable if you look at the right evidence.

Notice what is absent: charm, personal visibility, persuading anyone to want what they do not want. Marketing is neither self-promotion nor advertising. Both are merely tactics for the find job, and both are optional. The four jobs are not.

Three machines you are probably treating as one

A personal brand is one individual's audience, lent to whatever they attach it to. It works for some people. It is still the wrong instrument here: it takes years, demands continuous personal output, and leaves when the person does. You have said you will not do it. Say it once and stop paying rent on the guilt.

Company and product marketing is different machinery, building assets the company owns: a page that explains the product better than you do verbally, documentation a sceptical engineer can evaluate at 23:00 without talking to anyone, a case study with real numbers, a comparison against the incumbent, a pricing page, a security and GDPR page that survives procurement. They do not decay while you are on holiday, and they are the subject of this course.

Sales is the third machine: the conversations, in sequence, that turn an interested company into a signed contract. Lesson 9 covers doing that without performing a personality.

Check this yourself: look up the founders of the last ten B2B tools your company pays for. Most have no public following. Those purchases came from search results, documentation, a colleague's recommendation, a comparison page, a conference booth, or a well-timed cold email. Your constraint is the normal case.

The market is a pipeline, and it has exactly three levers

Model it as any staged system: strangers enter at the top, each stage has a transition probability, a latency and a drop-off. The funnel is that chain of stages, the conversion rate is the probability of surviving one stage, and the sales cycle is the latency between first contact and money. Output is:

revenue = volume × conversion × value

If output is too low there are only three things to change: more qualified strangers entering, higher survival per stage, or more money per closed deal. Every remaining lesson intervenes on one of those three, and knowing which lever you are pulling is most of the discipline.

Capture or creation: find out before you write a word

Demand capture means the buyer knows they have the problem, has a name for it, and is looking; your job is to be findable, credible and easy to choose against the two or three alternatives on their list. Demand creation means they do not know they have the problem, or do not believe it is worth money; your job is to change what they believe first and sell second.

Technical founders get this backwards both ways: philosophical essays about a category buyers already search for by name, or a comparison page for a product nobody knows to compare. Three tests, answerable this week:

Two capture answers and you compete on findability, proof and ease of purchase. Two creation answers and your first asset is not a landing page but an argument: writing or a calculator that makes the cost of the status quo visible in the buyer's own numbers. Most companies are partly both, and the mix drives the channel choices in Lesson 7.

Almost nobody is buying today

John Dawes, working with the LinkedIn B2B Institute and drawing on Ehrenberg-Bass Institute thinking, popularised the 95-5 rule: at any moment only a small minority of buyers in a category are in-market, and the large majority will buy later. Treat the exact split as directional, not a constant for your market; the consequence is what matters. If most of your addressable buyers cannot buy this quarter — the incumbent contract renews in fourteen months, the budget is set, the champion is not hired yet — then marketing aimed only at today's buyers fishes in a small pool, and spending on the rest pays off only if they remember you when their moment comes.

Byron Sharp and the Ehrenberg-Bass Institute name the two conditions: mental availability (being thought of, by the right person, when the need appears) and physical availability (being easy to buy once thought of). Mental availability is a cache you write to now and read from in eighteen months, and what gets cached is a short association between a situation and your name — which is why consistent wording beats cleverness.

What that means for a five-person company this month

• Mental availability: pick one sentence describing the situation you solve and use identical wording on the homepage, in the email footer, in talks and in every cold email, for a year. Get your name onto the two or three directories, marketplaces or review sites your buyers actually consult. Publish where their attention already sits — a customer's blog, a partner's newsletter, an industry association — not into your own empty channel.
• Physical availability: publish a price or at least a range. Make the first meaningful step self-service at 23:00. Pre-write the security questionnaire, the DPA and the GDPR sub-processor list so procurement does not stall for three weeks. If you sell to EU public sector, register in the tender systems buyers must use; being unbuyable in the required format removes you before anyone reads the product page.

Most technical founders describe their buyer as a job title plus a company size: "CTOs at mid-sized European companies." That is a database filter, not a buyer. Buying is caused by a situation: something changed in someone's week, and the pain of staying put exceeded the discomfort of changing. Until you can name that situation and the event that starts it, your website, pricing and outbound list are guesses.

Five words, kept apart

• Market: everyone with the problem you solve, whether or not they know it. Too big to aim at.
• Segment: a slice sharing a problem, a situation and a way of buying, so one message works for all of them. Two companies needing different opening sentences are two segments.
• ICP (ideal customer profile): the type of organisation worth selling to — the segment plus your commercial reality: who closes fastest, pays fully, stays.
• Persona: the role of a human inside it. Personas have goals, fears and a boss; organisations have none of those. Confusing the two is why "we sell to CTOs" feels like strategy and functions like fog.
• Firmographics: industry, headcount, funding stage, country, stack. How you find a segment, not what defines it: the index you scan, not the predicate you care about.

People do not buy products, they hire them

Jobs-to-be-Done, developed by Clayton Christensen and Bob Moesta, says it plainly: a person hires a product to make progress in a specific situation. The unit of analysis is neither the customer nor the product; it is the situation plus the progress wanted.

Moesta's four forces explain why a good product still loses:

1. Push of the current situation: the audit failed, deploys break, the one person who understood the pipeline resigned.
2. Pull of the new solution: the better future the buyer imagines. The only force your feature list addresses.
3. Anxiety about the new thing: migration risk, "will it fit our data model", "will I look stupid for choosing this", "will these people exist in two years".
4. Habit of the present: sunk cost, existing scripts, nothing on fire today, the effort of getting five colleagues to agree.

Push and pull together must beat anxiety and habit, and features sit on one side of that inequality only. Hence the thing that irritates engineers: you can reach feature parity and still lose every deal, because deals go to whoever removed the most anxiety and made changing cheaper than staying.

The main competitor is no decision

Your most frequent loss is not a rival, it is a deal that goes quiet. Doing nothing costs no budget approval, no security review, no legal meeting, no personal risk. To beat it, make the status quo expensive in his numbers, not yours. "We cut this cost by 30%" uses your assumptions and is discounted on sight. "Your two senior engineers spent six days on this last quarter, roughly EUR 4,600 of loaded cost, and the enterprise contract you cannot sign without it is worth EUR 120,000 a year" is checkable, defensible to a CFO, and repeatable when you are absent.

Nobody buys on a random Tuesday

Buying is event-driven. Lesson 1 established that at any moment only a small minority of buyers in a category are in-market (the 95-5 rule). The operating question is what moves someone into that minority, and the answer is an event. Your addressable-now market is much smaller than your addressable market, refilled continuously by events.

The classes of trigger event — the observable change that starts a buying process — are boringly consistent:

• A new hire in a relevant role: a new head of platform, CISO or VP engineering has 90 days to show judgement and a mandate to change things.
• Funding: new money, new board expectations, a hiring plan the current systems will not survive.
• An outage or serious incident. Push force at maximum, briefly.
• A failed audit, or a security questionnaire they cannot answer.
• A regulatory deadline: NIS2, DORA, or a GDPR transfer problem after a vendor review.
• A contract renewal with notice period — the only cheap window to switch.
• Crossing a growth threshold: first enterprise or regulated customer, 50 engineers, entering DACH.
• A key person leaving with the only working knowledge of a system.

Trigger awareness is the largest multiplier on outbound — contacting people who did not ask to hear from you (Lesson 8 covers the playbooks). Untargeted outbound is polling: you check everyone and mostly find nothing changed. Triggers are interrupts: same list, same copy, but you arrive in the fortnight when the problem is live.

Hence the segmentation heuristic: problem plus situation plus trigger. Industry and headcount build the list; they do not make anyone buy. None of this needs your face anywhere, only knowledge competitors did not bother to gather.

Who is actually in the room

Above roughly EUR 20,000, nobody buys alone. The buying committee is everyone who must not object.

Ten conversations, no research budget

You need ten: four recent buyers, three recent losses, three recent churns. Losses and churn carry more information than wins, which are contaminated by politeness.

Ask about one specific real decision, in the past tense. Hypothetical questions — "would you pay for this?" — produce polite lies: the person pays no cost for saying yes, and is predicting rather than reporting. Past-tense behavioural questions retrieve memory instead: dates, names, amounts, who objected. It is the difference between speculating about how a system might fail and reading the logs from a real incident.

Read these off your screen:

1. Take me back to the week this first came up. What was happening in the business?
2. What had you tried before you started looking outside?
3. Who first said out loud that this needed fixing, and what exactly did they say?
4. What did you search for, or who did you ask first? What words did you use?
5. What else was on the list, including building it yourselves and doing nothing?
6. Who else had to agree, and what did each of them ask for?
7. What nearly stopped you? What was the last thing you worried about before signing?
8. How long was it from first conversation to signature, and where did it sit still longest?
9. What did you expect to be different after 30 days, and what actually was?
10. If you had done nothing, what would be happening now?

Then stop talking; silence is where the useful sentence arrives. Record with permission and keep their exact nouns — Lesson 4 turns those into copy, and paraphrase is useless.

The usual technical view is that positioning is the sentence at the top of the homepage, written once the product works. That is a tagline, and a tagline is an output. Positioning is the input: which alternatives a buyer judges you against, and therefore which of your properties count as advantages and which are noise. Get it wrong and every downstream asset argues, competently, a case you cannot win.

It is a decision document, not a sentence

That document has a fixed list of consumers, which makes it the highest-leverage decision in this course: the homepage's first sentence, the price and packaging (Lesson 6), which comparisons you invite and which you refuse, who goes on the target list, which channels can work (Lesson 7), and which proof you must produce.

Every one is a function of the same input, so a positioning error is multiplicative, not additive, and nearly invisible. The copy renders the wrong argument faithfully, the ads buy attention from people who leave for reasons that look like price, and the salespeople answer objections produced by a comparison you never chose. Everything converts slightly worse than it should, and no single number is bad enough to investigate: the bug class where nothing throws and every endpoint is merely slower than the design said. No A/B test finds it: both variants argue the same wrong case.

Compared to what

Buyers cannot judge absolute value. Nobody knows what "good" costs for something they have never bought, so they find a reference class — the set of things they mentally file you with — and rank and price you inside it. Al Ries and Jack Trout built Positioning on this.

The filing happens in the first ten seconds, off your opening sentence. It fixes the price band that sounds sane, the criteria you are scored on, and the questions you get asked. If you do not nominate the reference class the buyer picks one from surface features — the interface looks like a log tool, therefore log tool — and imports that category's price ceiling. His default is the cheapest familiar thing in the room, so naming your alternatives is not a concession: it is choosing the denominator first.

The five components, in order

April Dunford's framework in Obviously Awesome is the most usable version of this, and the order is a dependency chain, not a form to fill in any sequence.

1. Competitive alternatives. What the buyer would do if you did not exist — behaviours, not a vendor list. A spreadsheet one analyst maintains. An internal tool a senior engineer keeps alive. And doing nothing, usually the largest single line. Source them from lost and stalled deals, not a market map.
2. Unique attributes. Capabilities no alternative on that list has, stated as capabilities rather than adjectives. The test: could a competitor's salesperson say your sentence with a straight face? "Fast, reliable, easy to use" — yes, so delete it. "Reconstructs which user read which tenant's records, per request, retained seven years" — no.
3. Value. What those attributes let the buyer do or avoid, in his units: hours, euros, audit dates, contracts unblocked. Force the two-step — attribute, then "which means you can" or "which means you avoid". If the second half will not finish, the attribute is a hobby.
4. Who cares a lot. The characteristics of buyers for whom that value is urgent and expensive. Characteristics, not demographics: "has a contract clause requiring per-tenant access records" is a characteristic; "500 to 2,000 employees" is only a way to build a list.
5. Market category. The frame of reference that makes the value obvious to that buyer: not what you are internally, but what you are best understood as. Chosen last, because it depends on the four above.

Trends are seasoning. One belongs only if it makes your value more urgent for your segment now — a compliance deadline like NIS2. Never make it the subject: the season's fashionable technology files you beside hundreds of identical claims, and dates the document within eighteen months.

Choosing the category is choosing the price

An existing category buys instant comprehension: the buyer knows what it is, what it costs, who signs, and which criteria to score, and you inherit a budget line. The bill is that you are scored on the category's whole criteria list, including the parts you are weak at, and its cheapest credible member sets your price ceiling.

Inventing a category replaces the comparison with an education bill: every buyer must be taught the name before he can want the thing, paid for in years of content, talks and sales calls that start from zero. Most companies under fifty people cannot fund it.

Failure modes, and how to test it in a week

• Positioning for investors. Market-size language, "the X for Y", platform ambition. It feels right because you rehearse that sentence for investors more than for buyers. Investors buy a future market; buyers buy relief from a present problem.
• Positioning for everyone. Narrowing feels like discarding revenue, so the language hedges until it moves nobody. A small home market sharpens the instinct; the arithmetic contradicts it, since one narrow segment across the Nordics and DACH is bigger and cheaper to reach than everyone in Estonia.
• A feature list in a positioning costume. Features are the true, checkable, hard-won part, so listing them feels honest. It hands the synthesis to the buyer, and he will not do it.
• Leading with technology instead of the job. The stack, the model, the architecture. It feels right because that is how you would pitch it to yourself. Only the technical evaluator cares, and he does not hold the budget.
• Copying the category leader. Their broad, confident line works because a hundred thousand companies already know who they are; from an unknown company it is empty. Their positioning is a consequence of credibility you have not accumulated.

Say the positioning aloud to five people in the target segment — not friends, not investors — then stop talking. You are not collecting opinions, which are polite and worthless, but measuring which bucket they file you into. Two signals:

• "So it is basically like X?", where X is not on your list of alternatives. Wrong reference class: you led with a mechanism that resembles a cheaper product, and he filed you there.
• "Who is this for?", or "would this work for us?" No segment: a capability described without the characteristics of the person who needs it badly.

It is working when they finish your sentence with an example of their own.

The belief is that once positioning is settled the words are a formatting exercise. That is why the homepage gets rewritten four times a year and performs identically each time. Positioning is one decision; messaging is the set of claims derived from it, served differently depending on what the reader already knows. Most "the site does not convert" problems are a mismatch between what the reader believes and what the page assumes he believes.

One decision, many renderings

The tree has four levels, kept in one file:

• The one-liner. One sentence, whose job is not elegance but surviving repetition by someone who does not work for you.
• The value proposition: two to four sentences on who it is for, what changes for them, and why it works.
• Three pillars. A pillar is one claim about one kind of value. Three, because a champion relays you to his committee from memory, and past three he picks the survivors instead of you.
• Proof under each pillar, one to three checkable items. A pillar with no proof is a slogan.

Everything external is a projection of that tree: the homepage is the whole tree, one screen per pillar; a cold email is one pillar plus one proof point; a case study proves one pillar end to end. A claim living in an asset but not in the tree means one of the two is wrong.

The reader's state decides the message

Eugene Schwartz, in Breakthrough Advertising (1966), set out five stages of awareness.

The mechanism behind the dead homepage: you live at stage four, so you write at stage four. Traffic arrives at stage two — a search for a problem phrase, a colleague's remark, a line in a tender. He meets a page comparing you with firms he has never heard of, on criteria he has not decided matter, and leaves. That is a type error, not a tone problem, and persuasion does not repair it. Diagnose it from your search queries: problem phrases landing on a competitor grid is the mismatch. Fix it by stepping the page down the ladder — problem, approach, product, proof, price.

The "so what" ladder

Every claim sits on one of four rungs: the feature (what it is), the function (what it does), the consequence (what that causes at the company), and Monday (what changes in one person's week).

Technical founders stop at rung one, because rung one is true, checkable and hard-won, while three and four feel like presumption about someone else's business. Marketers with no substance start at rung four — "ship faster, sleep better" — unfalsifiable, identical to every rival's headline, discounted to zero by the person you need. A credible message carries all four in order, because the lower rungs license the higher ones: alone, "nobody assembles the pack by hand" is a stranger's boast, but behind the store and the detection it is an argument he can attack, and repeat.

Claim, mechanism, proof

A claim is free to make and therefore carries no information. A technical buyer hunts for the because clause: the mechanism that would have to be true for the claim to hold. Give it and he can check it and see its limits, and a stated limit is the strongest trust signal a company nobody has heard of can offer. Then proof, in three classes that are not interchangeable.

For the evaluator, credibility runs demonstration, evidence, social proof; for the person who signs, roughly the reverse. Hence one of each rather than three of the cheapest.

Their words, and how to know it worked

Thirty minutes of language mining

His phrasing beats yours mechanically: a phrase he already uses needs no translation, and translation is where attention leaks.

1. Ten minutes. The last twenty inbound emails and contact-form messages, plus any recorded first calls. Paste every sentence describing the problem verbatim — no paraphrase, no tidied grammar.
2. Ten minutes. Support tickets; two recent tenders in your segment, whose requirement wording is the buyer's own specification language; then two- and three-star reviews of rivals, which describe the job better than five-star ones.
3. Five minutes. Search one community your buyers use for the problem phrase, taking the threads where someone is complaining.
4. Five minutes. Sort into problem, outcome and objection words and count repeats. A phrase used by four different companies goes on the page unedited.

Spend them: the headline takes his problem noun, headings his outcome verbs, the FAQ his objection wording, the email subject line the phrase from the tender. Words mined from people who never buy attract more of them.

Adjectives out, numbers and nouns in

The test is falsifiability: a sentence that could not conceivably be false carries no information, and your competitor published it already. Any surviving adjective must attach to a number or a noun.

Four tests that work at 200 visitors a month

You cannot A/B test a page at that volume; detecting a small difference needs thousands of samples per variant. Test comprehension instead, which needs eight people.

• Five-second test. Show the page to someone in your segment for five seconds, hide it, then ask what it does, who it is for, and what he would do next. That measures comprehension, not preference.
• Explain-it-back. End a first call with "describe this to your CTO in two sentences" and record the answer verbatim. Your message must survive retelling by someone who does not work for you; five different answers from five people means you have none.
• Cold email as the A/B channel. Email is the one place you control the denominator: a hundred addresses on variant A, a hundred on B, same week, same ask, changing only the first sentence. Two replies becoming eight is real; four becoming five is noise. GDPR and national ePrivacy rules apply to EU business addresses, Germany strictest.
• Ask the newest customer. Within 48 hours of signature: "before we spoke, what did you think we did, and what made you take the call?" That is the gap between your message and the belief it produced.

The technical view of copy is that it is decoration: the product does the work, the words describe it, and people who are good at words do that part. That misreads what the reader is doing. A stranger on your page is not being persuaded, he is trying to disqualify you cheaply, because a wrong vendor costs him a quarter and closing a tab costs him nothing. Copy is not persuasion. It is the removal of specific doubts, in the order a skeptic raises them, with evidence attached to each.

Pages: seven blocks, one job each

A landing page is any page a stranger arrives on cold; the homepage is the one that must work for the widest mix of arrivals. Treat it as a pipeline in which each block hands the reader to the next. Throughput is the product of the pass rates, so a missing block does not cost you a little, it multiplies by something close to zero.

A call to action is the one thing you ask the reader to do: one per page, in the hero and again at the close. Everything else — pricing, docs, sample export — is a text link. Two co-equal buttons is not choice; it is a decision handed to someone with no basis for making it, and the usual result is neither.

The FAQ is an objection-handling device

An objection is the reason a buyer gives himself for stopping. On calls you answer roughly eight of them, repeatedly. The FAQ answers them once, in public, in his phrasing, available at 22:00 to the engineer told to evaluate you: "Can we self-host?", "What happens to our data if you go under?", "How long is the integration, honestly?", "Why is this not just a Postgres table?". A question nobody has asked — "Do you offer support?" — signals you do not know who is reading. Answer with a number or a limit. "Two weeks for a typical integration, four if your logs are unstructured" beats "fast and easy" because it can be wrong.

What hiding your price costs

Hiding price feels like protecting negotiating room. It mostly moves cost onto deals you would have won. Three mechanisms. Buyers cannot judge absolute value, so with no number they take their reference point from competitors' published prices, usually the cheapest familiar option, which is the wrong band for you. Disqualification is a service you are refusing to perform: at a loaded EUR 80 an hour, twelve unqualified demo calls a month is about EUR 12,000 a year of senior time spent telling people they cannot afford you. And in EU public procurement, or any process with a budget owner, a supplier who indicates no price often never reaches the shortlist, because the buyer must forecast a figure before he can start. Hiding is correct in a narrow band: bespoke work where scope swings the price more than roughly 5x between clients, tender business where a published number becomes a ceiling forever, and products too young to know their worth. Even then publish a floor and a shape — "projects start at EUR 25,000, typically EUR 40,000–90,000 over three to five months". The floor disqualifies; the range preserves the negotiation. Lesson 6 covers what to charge for.

Cold email: relevance is the whole product

Cold email is a message to someone who did not ask to hear from you. Copy quality has a ceiling and list quality sets it, the way the index rather than the wording of the query decides whether you get a scan or a seek. Spend your time on the trigger. The shape: one observed trigger, one specific hypothesis about their situation, one small ask, under 120 words. First, the version most people send, illustratively bad:

Subject: Quick question

Hi Anna, hope this email finds you well! I was really impressed by the great work the team at Nordkraft is doing in fintech. At Auditrail we are a leading provider of innovative, enterprise-grade compliance automation that helps companies like yours streamline audit readiness and unlock efficiency across the security stack, with real-time monitoring, powerful dashboards and seamless integrations. Would you be open to a quick 30-minute call next week to explore synergies?

"Quick question" is not a question. "Hope this finds you well" marks it as bulk before the first claim. "Impressed by the great work" is fake personalisation: it contains nothing that could only be said to her, which tells her the rest is a template. The middle sentence makes four unfalsifiable claims and gives no reason any is true. Thirty minutes is not a small ask, it is a meeting she must justify. And nothing explains why this arrives in August 2026 rather than at any point in the last three years. The same company, done properly:

Subject: per-tenant access logs before your ISO 27001 audit

Anna — you have two SRE roles open with ISO 27001 in the requirements, so I am guessing Stage 1 lands this autumn.

The access-control clause is where teams your size lose time. The usual answer is exporting logs into spreadsheets, which cost the last four companies we worked with about ten engineer-days each.

We keep per-tenant access events in a hash-chained store and export the evidence pack directly. A Nordic payments customer passed its surveillance audit in March on that export.

Would the sample export be useful? I will send it, no call needed.

— Martin, Auditrail, Riga. Reply "no" and I will not write again.

Order of magnitude, not a promise: from a hand-built list of 200 companies where you verified the trigger yourself, plan for replies in the low single-digit percentages and a handful of calls; from a bought list of 5,000, plan for approximately nothing plus a damaged sending domain. Send 40 to 60 a week from a real mailbox and change one variable at a time. Your own first 200 are the only numbers that mean anything.

Case studies: the constraint carries the load

A case study is a short account of one customer's problem and result, written so a similar buyer recognises himself. Six parts: situation and constraint; what they tried before; what you did, including the mechanism; the measured outcome with a number; a quote in the customer's voice; and what it cost or how long it took. Most published cases carry parts three and four only, which is why they read as advertising.

The constraint section is what makes the rest believable. A skeptic's default explanation for any success story is that the conditions were unavailable to him: greenfield system, real budget, a cooperative team. Naming the constraints puts his situation inside the same class, and it is the one section a vendor has no incentive to fabricate. A stated limitation is cheap for you and expensive to fake, which is precisely why it reads as true.

When the customer will not disclose revenue, and most will not, use proxies: percentages instead of absolutes ("39% fewer billing tickets"), time in engineer-days or working days, counts of things avoided (incidents, audit findings, manual reconciliations), and cycle times such as close from 11 days to 4. Much of this you can measure from your own delivery records without asking. A number you measured and can explain the method for beats a customer-supplied figure you cannot defend.

Sign-off is a process, not a favour. Write the draft yourself, send it as an editable document, name the approvers you expect (the champion, his manager, someone in marketing or legal), set a date two weeks out, and offer the anonymised version in the same email so refusal is not the only alternative to yes. Written approval by email is normally enough. When they cannot be named, publish the shape: "a 200-person Nordic payments provider, named on a reference call under NDA". That converts better than most people expect, because the constraint and the mechanism are what the reader is checking, not the logo.

The proposal is a marketing asset

For services and larger deals the proposal is the last piece of copy in the deal and often the only one three of the deciders read. Five parts: their problem restated in their words, taken verbatim from the call; two or three options with prices; explicit scope boundaries, including a short "not included" list; a timeline with real dates; and one next step carrying a date — "confirm by 26 August and we start on 8 September".

A proposal that restates the problem better than the client did beats a cheaper competitor, for a reason worth understanding. His exposure is not the EUR 8,000 price difference; it is the project failing, which costs him a quarter and some credibility. He cannot inspect your delivery quality before buying, so he substitutes the only available signal: whether you appear to understand the job. A precise restatement, including the constraint mentioned in passing that everyone else ignored, is the cheapest evidence of comprehension there is. The "not included" list looks like surrendered scope and buys trust instead, because a supplier who has thought about boundaries has evidently done this before. Three priced options give him something to decide other than yes or no, and reveal his real budget without your asking.

Editing rules that work without talent

• Cut 30% of the words, not the ideas. Set the number before you start or you will trim two commas and stop.
• One idea per paragraph. If it needs "and also", it is two paragraphs.
• Front-load every sentence: point first, qualification after. A skeptic reads four words and moves on.
• Delete the adverbs. "Significantly faster" survives only as "8 seconds instead of 40 minutes".
• Read it aloud. Anything you would not say to a customer's face comes out. This catches more than any other rule.
• The opposite test. Would a competent competitor ever write the opposite? Nobody claims poor quality and slow support, so your sentence claiming the reverse carries no information: delete it, or replace it with the number that makes it falsifiable.

This lesson is still being written.

This lesson is still being written.

This lesson is still being written.

This lesson is still being written.

This lesson is still being written.

This lesson is still being written.

This lesson is still being written.

This lesson is still being written.