Conversion copywriting: clarity beats clever

The most expensive copy on the internet is the clever kind. The headline that made the marketing meeting laugh, the tagline that's a pun on the category, the hero that says "Reimagine everything" — all of it tests beautifully in the boardroom and converts like a wet match. Meanwhile the highest-performing headline we've shipped in years was six words long and could have been written by the customer's accountant: it named the job, named the outcome, and named the timeframe.

Conversion copywriting is not a writing skill wearing a marketing hat. It's a research skill with a typing component. The words that convert are almost always already in the customer's mouth — the craft is finding them, ordering them, and having the discipline to not decorate them. Here's the process we run, the same one behind copy for a community bank, a skincare brand with a radical honesty problem and a dozen SaaS homepages.

Mine the voice of the customer before writing a word

Every strong piece of conversion copy is eavesdropping, systematised. Four sources, in order of value:

Sales and support transcripts. The way prospects describe their problem on a call — their nouns, their verbs — is your headline vocabulary. If customers say "I'm drowning in reconciliation" and your homepage says "streamline financial workflows," you have translated the customer's pain into a language no customer speaks.
Reviews — yours and competitors'. Reviews are structured voice-of-customer data: what people valued enough to praise, what hurt enough to complain about, and the exact phrases they reach for. A hundred competitor reviews will surface the category's unmet expectations in an afternoon.
Sales-call objections, written down. Every "we went with someone else because…" is a landing-page section waiting to exist. More on this in a moment.
On-site search queries. What people type into your search box after arriving is a confession of what your navigation and headlines failed to say.

Collect phrases verbatim — not paraphrased, not summarised. A spreadsheet of fifty real customer sentences is worth more than any positioning workshop. The workshop's job is to choose from that material, not to generate alternatives to it.

Build a message hierarchy, not a headline

The mistake almost every team makes is treating "the copy" as one decision. A converting page is a sequence of decisions, each answering the question the previous one raises. We structure it as a hierarchy:

Level 1 — What is it, and who is it for? The hero's only job. If a first-time visitor can't answer both within five seconds, nothing below matters. Concrete beats elevated: "Board packs your directors actually read" outperforms "Governance, elevated" by exactly the margin you'd expect.
Level 2 — How does it work? Three steps, maximum, in the customer's verbs. This section exists because Level 1 created a scepticism debt: "sounds good, but how?" Pay it immediately.
Level 3 — Why believe you? Proof matched to the doubt: numbers for the CFO, names for the founder, a security page for the technical evaluator. Proof is not a logo strip; it's targeted evidence.
Level 4 — What happens next? One call to action with the friction stated honestly. "See the demo — 20 minutes, no credit card, bring your questions" converts better than "Get started" because it prices the click.

The test for the hierarchy: read only your headings, in order, aloud. If they tell a coherent story, the page works even for scanners — who are most of your visitors. If the headings are "Features / Benefits / Solutions / Why us," the page has no argument; it has furniture.

Objection mapping: write the page the sceptic reads

Here's the exercise that changes conversion copy more than any headline formula. List the five strongest reasons a qualified prospect wouldn't buy — not straw men, the objections your sales team hears weekly. Then check: does the page answer each one, in descending order of frequency, before the call to action?

Classic objections and their page-section answers:

"It's probably expensive" → pricing or price-framing above the fold of the consideration section, never hidden.
"Migration sounds painful" → a "switching" section with real timelines and a named human. This objection killed conversions for a bookkeeping tool we worked on until the homepage added four honest sentences: "We import your last two years. It takes a day. A person checks it. Free." Trial starts moved double digits.
"Will it work with our stack?" → integration proof in the how-it-works section, not a footer link.
"Is this for companies like mine?" → segment-specific proof points, not generic logos.
"What if it doesn't work out?" → the exit terms, stated plainly. Paradoxically, visible exits raise entries.

The principle underneath: an objection left unanswered doesn't disappear — it adjourns the decision, and adjourned decisions mostly never resume. Your page's job is to hold every objection the sceptic raises and answer it while they're still reading.

Headlines: frameworks without the formula smell

Frameworks are scaffolding, not paint. Use them to generate candidates, then choose by ear and evidence. The ones that keep earning their place:

Customer-quote headlines. Lift the best sentence from research, lightly trimmed. It sounds like a person because one said it.
Outcome-plus-timeframe. "Close your month-end in two days, not nine." Specificity is credibility; vagueness is poetry.
The enemy headline. Name the thing the customer is fed up with: "Timesheets are a tax on honest people." Enemy headlines work because they signal we see your actual life — but only if the product absolutely delivers on the promise.
Plain statement. Sometimes the winning move is to simply say what it is, unusually clearly. In a category of superlatives, plainness is the loudest voice in the room.

What to avoid: wordplay that needs decoding (the reader owes you no effort), category adjectives with no nouns ("powerful, intuitive, seamless"), and the "It's like X for Y" construction unless Y's customers already love X.

Testing copy without a twelve-variant circus

Small teams can't run landing-page factories, and don't need to. Copy testing works at modest scale if you respect three rules:

Test messages, not words. "Free trial" vs "Try it free" is a coin flip dressed as science. Test arguments — speed versus accuracy, price versus risk-reversal, one proof type versus another — because Argument A vs Argument B teaches you something portable about your market, whatever the result.

Sequence the tests by hierarchy. Hero first (it gate-keeps everything below), then the objection sections, then the CTA framing. Testing the footer while the hero is broken is reorganising deck chairs with a spreadsheet.

Use cheap research for directional answers. Five-second tests tell you whether Level 1 lands. Message-market-fit interviews tell you which argument resonates before you spend traffic. Save A/B testing for the finalists — it's the expensive, precise instrument, not the idea generator. This is the same discipline we apply to experiment design generally and to landing-page programs: velocity of learning, not of variant-count.

And a note on where copy fits the broader growth system: copy changes are the cheapest lever in the funnel — no rebuild, no new traffic, often a same-week ship. Before any client spends a dollar on more acquisition, we audit the words the existing traffic reads. It's unglamorous, and it's usually the highest-ROI week of the engagement. If you'd like that audit run on your pages, the brief takes ten minutes.

Key takeaways

Mine verbatim customer language from transcripts, reviews and search queries before writing. Your headline vocabulary is already in their mouths.
Structure pages as a message hierarchy — what/who, how, why-believe, what's-next — and test it by reading headings aloud as a story.
Map the five strongest objections your sales team hears and answer each on the page before the call to action. Unanswered objections adjourn decisions.
Use headline frameworks as scaffolding, then choose by ear. Clarity is the loudest voice in a category of superlatives.
Test arguments, not words; sequence tests by hierarchy; and use five-second tests and interviews for direction before spending traffic on finalists.

FAQ

Should the founder write the copy, or a copywriter? The founder should be interviewed, at length, by whoever writes. Founders carry the sharpest version of the pitch — honed in a hundred sales conversations — but often can't see it anymore, or bury it under the investor-deck version. The copywriter's job is to excavate the founder's best material and the customers' language, then arrange it. Pure founder-written copy is usually too inside; copy written without the founder is usually too outside.

How do we write conversion copy for a product with multiple audiences? One page, one primary audience — full stop. A homepage that speaks to everyone converts no one; the hierarchy collapses because each level tries to answer three audiences' questions at once. The scalable pattern is a sharp primary page plus dedicated paths (segment pages, industry landing pages) for secondary audiences, each with its own objection map. It's more pages and less compromise, which is the correct trade.

Does longer or shorter copy convert better? Wrong axis. The right question is: does every sentence answer a live question or objection, in order? Complex, expensive, high-commitment decisions need more of those answers than a $9 subscription. Cut everything that isn't doing a job and keep everything that is — the length that survives that edit is the right length.

Can AI write our conversion copy? It can produce competent category-normal copy at speed — which is exactly the problem: category-normal is the baseline your competitors also just generated. AI is genuinely useful in the research phase (clustering review themes, drafting interview questions, producing variants for cheap message tests) and as a sparring partner for edits. The judgment layer — which customer phrase is the headline, which objection ranks first — still belongs to someone who has listened to the calls.