A stranger, on a phone, with four seconds

The one-job rule

A landing page has exactly one job. Not one primary job among several. One.

The test is a sentence: "This page exists to get a stranger to ___." Book a call. Join the waitlist. Buy the $49 tier. Download the template. If the blank needs an "and" or an "or" to fill in, the page is going to underperform on every option in the list, because a visitor asked to choose between two actions frequently chooses neither.

The most common violation is not two buttons. It is a page trying to serve two audiences — a product page that also wants to impress recruiters, a service page that also wants to sell a course. Both audiences read copy addressed to someone else and conclude, correctly, that it is not for them. Split the page. Pages are free.

What a cold visitor actually is

The distinction that matters is not "traffic" versus "no traffic" — it is temperature. A warm visitor arrives already knowing who you are and what you sell; the page only has to close them. A cold visitor arrives from a link, on a phone, mid-scroll, with no context and no reason to extend patience.

• They do not know what category the product is in, so they cannot fill in gaps from experience.
• They will decide whether this is relevant to them in a few seconds, and that decision is mostly made from the headline and the first image.
• They are on a phone, so "above the fold" is roughly one headline, one subhead, and one button.
• They assume, by default, that this is like everything else they have already dismissed today.
• They will not scroll to find out what it is. They scroll only after deciding it might be relevant.

Two cheap tests catch nearly all of this. The five-second test: show the page to someone unfamiliar for five seconds, take it away, and ask what it does and who it is for. The squint test: blur the page until text is unreadable and check whether the visual hierarchy still points at one headline and one button. Both take minutes and both find problems that months of self-review will not.

Above the fold: five elements, in order

1. A specific headline naming what it does, who it is for, and the outcome. This carries most of the page's weight.
2. One subhead sentence explaining the mechanism, plus whatever qualifier keeps the wrong people out.
3. One primary call to action, an action verb plus an object, visible without scrolling on a phone.
4. One risk-reducer or proof element directly beneath the button — a price, a guarantee, "no card required", or a genuine number.
5. One visual of the actual thing. A screenshot of the real interface, the real garment, the real document. Not a stock photo of a team pointing at a laptop.

Everything else on the page — the manifesto, the founder story, the feature grid — belongs below that. Not because those things do not matter, but because they are answers to questions the visitor has not asked yet, and answering unasked questions is how a page loses someone in its first screen.

Specificity beats adjectives, every time

The single highest-leverage edit available on most pages is deleting adjectives and replacing them with nouns, numbers, and verbs. Adjectives are claims anyone can make. Specifics are claims only you can make.

There is a rule that catches almost all of it: if a competitor could put your headline on their page without lying, it is not a headline. "Fast, reliable, and easy to use" passes that test for every product ever made, which is why it converts nobody.

The same principle governs the qualifier. "For small freight brokerages running under 500 loads a month" loses you every visitor who was never going to buy and gains credibility with every visitor who fits. Naming your audience narrowly feels like turning away business. It is the opposite: it is the fastest way to make the right person feel spoken to.

Social proof when you have none

The advice to "add testimonials" is useless for a product that has not launched. The temptation is to manufacture them, and that is worth naming plainly as a bad idea: fabricated reviews, invented testimonials, and logos of companies that are not customers are a legal exposure under advertising rules in multiple jurisdictions, and separately they are the thing most likely to be found out by exactly the person you wanted to impress.

The workable substitutes, roughly in order of strength:

1. Show the work. A live demo, an interactive sample, a public repository, a real screenshot with real data in it. Demonstration outranks description and outranks testimony.
2. Be specific about something checkable. A number you can defend, a method you can describe, a document someone can open. Specificity itself reads as evidence.
3. Be transparent about the stage. "Founding-client pilot, three slots, $999 setup plus $249 a month" is credible precisely because it admits earliness. "Trusted by thousands" from a brand nobody has heard of is not.
4. Reverse the risk. A refund policy, a no-card trial, cancel-anytime, or a pilot with a defined exit. Every one of these converts better than a testimonial from a stranger.
5. Borrow credibility that is genuinely yours. Published work, a real employer, the standards you follow, the stack you built on. Real and checkable, never implied.
6. Put a real name and a real face on it. A human being who can be looked up is meaningful proof that someone is accountable for the thing.
7. Describe the process. "Week one: we connect the inbox and map your invoice formats" tells the visitor you have thought this through further than they have.

And then, from the first real user, ask for one sentence. One specific sentence from a named person beats five generic ones, and beats a logo wall entirely.

Objections as page sections

The most reliable way to structure everything below the fold is to stop thinking about it as content and start thinking about it as a queue of objections, answered in the order a real person raises them.

The "who it is not for" block is the one people resist and the one that most reliably raises conversion. Ruling people out is what makes ruling people in believable. A page that says "not right for you if you run fewer than 20 loads a month" has just told every visitor above that line that the claims are calibrated rather than aspirational.

Section order should follow the objection order, not a template. And the page should be exactly long enough to answer the objections that exist for that price point — a $19 download needs far less page than a $999 engagement, because the amount of reassurance a visitor requires scales with what they are risking.

One CTA, repeated

The same action, in the same words, in three to five places down the page: in the hero, after the how-it-works section, after the objections are handled, after pricing, and in the footer. Not because visitors miss it, but because different visitors become ready at different points, and the CTA needs to be within reach at the moment readiness arrives.

• Use identical wording every time. "Start the pilot" in the hero and "Get in touch" at the bottom read as two different offers, and the visitor has to work out whether they are the same. That work costs conversions.
• Verb plus object. "Join the waitlist", "Book the call", "Get the template". Not "Submit", not "Learn more", which is the button people press instead of the one you wanted.
• One secondary action at most, visually subordinate and lower-commitment — a sample, a spec sheet, a demo. It exists to catch people who are not ready, not to compete for the ones who are.
• Reduce the form. Every field costs conversions. If the follow-up is by email, do not ask for a phone number. Name and email is usually the entire form.
• Say what happens next, right under the button. "I reply within a day" removes more hesitation than another paragraph of copy.

Speed and phones are conversion features

A landing page is mostly text and a couple of images. There is no good reason for it to be slow, and slow pages lose visitors before they read anything — the visitor who leaves during load is invisible in every metric you have. Build it as a static page, inline the critical styles, compress the images, and be extremely suspicious of anything that delays the headline from rendering, particularly a hero video or a font that blocks paint.

Then check it on an actual phone, on cellular, not on a desktop browser resized to look like a phone. That is where the traffic is, and it is where the four-second decision gets made. The technical side of building pages this way is covered in the client-side tools guide, and how the same page gets found in the first place is in the AI search optimisation guide.

Measuring it without a tracking stack

You do not need a tag manager, a session recorder, and a heatmap tool. At the traffic levels a new page actually gets, those tools mostly produce confident-looking noise. What you need is one number and enough structure to attribute it.

the only metric that matters early: conversions per 100 sessions not pageviews, not time on page, not bounce rate

Instrumenting it for free

• Use a distinct destination per CTA placement. A different anchor, a different form URL, or a different mailto subject line for the hero button versus the footer button. Where the click came from is then inferable from the destination, with no JavaScript at all.
• Use a distinct link per channel. One link in a post, another in a newsletter, another in a bio. Which channel produced conversions matters much more than which button colour did.
• Put a hidden source field in the form, or a source parameter in the booking link. One field, filled automatically, answers most attribution questions.
• Count sessions with something lightweight — server logs or a privacy-friendly counter. You need the denominator, not a behavioural profile.
• Ask every single inbound lead what made them reach out. Twelve qualitative answers are worth more than a dashboard with no conversions in it, and they will tell you which objection the page is failing to answer.

The sample-size honesty section

This is where most solo landing-page optimisation goes wrong. At a 2% conversion rate, the numbers you see at small volumes are dominated by chance. Here is the range you would expect purely from randomness, with a true 2% rate:

Read the first row carefully: with a genuinely good page and 100 visitors, getting zero conversions is entirely normal. Redesigning after that is not optimisation, it is reacting to noise, and it destroys more pages than bad copy does. The discipline is to hold the page still until there is enough traffic for the number to mean something.

Which leads to the rule that follows from the table: only test big changes. A different headline and a different offer might be detectable at a thousand sessions. A different button colour is not detectable at any volume you will reach this year. Spend the effort on the headline, the offer, and the objection sections, in that order.

A pre-publish checklist

1. The one-job sentence is written down and contains no "and".
2. The headline is a claim a competitor could not truthfully make.
3. There are no adjectives in the headline or subhead that survive a "prove it" question.
4. The primary CTA is visible without scrolling on a phone.
5. Price, or the reason there is no price yet, appears on the page as a number.
6. There is a "who this is not for" statement.
7. Every proof element is true and checkable, and nothing is invented.
8. The same CTA wording appears at least three times.
9. The form asks for the minimum, and the button says what happens next.
10. Someone unfamiliar passed the five-second test.

Pages built on these rules across this site include FreightDesk, which has exactly one job and states its pilot terms as a number, KXNG SEF, which is honest about being a pre-launch waitlist rather than pretending to be a shop, and the services page. The pricing arithmetic that sits behind whatever number goes on the page is in the digital product pricing guide, and free starting points are on the templates page.

The uncomfortable summary

Most landing pages do not fail on design. They fail because they were written for someone who already understands the product, they ask for two things instead of one, they make claims a competitor could copy word for word, and they get redesigned after a hundred visitors on the strength of two conversions that were always going to be noise.

One job. One audience. Specifics instead of adjectives. Objections answered in the order they arrive. The same button, repeated. And then enough patience to let the number mean something before touching it again.

Tools referenced in this guide

• FreightDesk — a single-job page with the pilot terms stated as a number.
• KXNG SEF — a pre-launch waitlist page that says so rather than pretending to be a shop.
• Templates — free starting points, including page and copy structures.
• Digital product pricing — the arithmetic behind whatever number ends up on the page.