Practical guides on websites, online stores, and getting found on Google — written for business owners, not developers.
Tell us about your business today — get a free quote within 24 hours. No pressure, no confusing tech talk.
Tell us about your business today — get a free quote within 24 hours. No pressure, no confusing tech talk.
Most landing page advice assumes you bought the traffic. Where the button goes, what it says, and how much to explain first, when nobody clicked an ad.
Every guide to landing pages tells you to put the button where people can see it. Above the fold, before the scroll, no hunting required.
In 2012 a conversion consultant tested that on a long B2C page. He moved the call to action to the bottom, underneath all the explanation, and conversions went up 304%.
That number has been travelling ever since. It now appears on marketing sites as evidence that CTAs above the fold outperform ones below by 304% — the same figure, pointed the opposite way, source long gone.
You may not agree with this one. Where the button goes isn't settled by the fold. It's settled by how much the decision costs the person making it, and by how much they need to know before they can make it. Someone downloading a free checklist and someone commissioning a $4,000 website are doing different arithmetic, and the same layout can't serve both.
So this article is about the page: where the action sits, what the button says, how much you explain before you ask, and what to take off the page entirely.
It assumes people are arriving. If they're not, that's a traffic problem and no amount of layout work fixes it. Seven things, roughly in the order worth doing them.
There's a reason so much landing page advice feels slightly wrong when you try to apply it, and it isn't you.
Take the two most repeated rules. Strip the navigation off the page. Match your headline to your ad. Both are sound, both come out of real testing, and both assume the same thing: that your visitor arrived by clicking an ad you paid for, onto a page built for that one campaign.
Unbounce's Oli Gardner formalised the first as attention ratio: the number of things a visitor can click, against the number of things you want them to do. The target is one to one. The tests behind it are real. Removing links from a page took one Unbounce test from a six-to-one ratio to one-to-one and lifted conversions over 40%; another went from ten-to-one to one-to-one and gained 31% on ebook downloads. VWO stripped the navigation from one of its own pages and doubled conversions.
Now look at where your visitors actually come from. A search result. A link in your own menu. An email you sent. Somebody's recommendation. They didn't click an ad, and the page they landed on is one of eight service pages that all need to sit inside the same site.
Strip the navigation off that page and you haven't built a landing page. You've broken your website.
The underlying principle survives the move and the tactics don't. One page, one job, and nothing on it competing for the click — that holds anywhere. Delete the menu was only ever an implementation of it, for a page with no site attached.
Message match travels the same way. The version everyone teaches is match your landing page headline to your ad headline. The version that applies to you is match it to whatever brought them: the query they typed, the anchor text they clicked, the subject line of the email. Same principle, different source.
That's the filter for everything below. Where a rule was measured on paid campaign traffic, it's named as such, and what's left after you take the ad away is spelled out.
The test behind that 304% was run on a page selling a recipe and ingredient delivery service to busy families. The control did what you'd expect. Offer at the top, sign-up form beside it, everything a visitor needed inside the first screen. The treatment moved the form to the bottom and spent the space above it explaining what the service actually was.
The treatment won by 304%.
Michael Aagaard, who ran it, attached a caveat almost nobody repeats. The two versions differed in more than the position of the form, so the whole lift can't be pinned on the fold by itself. Worth carrying that. It's a strong result and it isn't a clean single-variable test.
It isn't the only one either. MECLABS recorded a 20% lift from moving a call to action below the fold in one test, and 220% in another where the extra room let the page make its argument before asking for anything.
Then the result that keeps the rest of this honest. MECLABS also ran three different layouts of a Boston Globe subscription page against each other and found no meaningful difference between any of them. Sometimes the fold decides nothing, because the visitor already knew what they came for and would have found the button wherever you put it.
Which is the actual finding here, and it isn't "put the button lower." Above the fold is right for a page whose offer explains itself in a sentence. A free download, a product people already understand, a returning customer coming back to buy. If someone can weigh the whole proposition in five seconds, making them scroll before they can act is friction and nothing else.
What decides it is how much a person needs to know before the answer could be yes.
So write those questions down. Not the ones you'd like to answer, but the ones that stand between a stranger and an enquiry. What is this. Is it for someone like me. Roughly what does it cost. How long does it take. What if I don't like it. What happens if I need a change after launch. If there's one question, the button goes at the top. If there are five, the button goes after the fifth, and every question you answer on the way there is one that can't stop them at the bottom.
One thing doesn't move. Your value proposition stays above the fold whatever you do with the button, because a visitor who can't tell what you sell won't scroll far enough to find out. The argument is about where the action sits, not where the explanation starts.
One last thing about that 304%, since the number is going to keep turning up. Before you rebuild a page around a statistic, check which way it was pointing when somebody measured it.
Everybody who lands on your page arrives holding a sentence. It's whatever brought them there, and they're carrying it in working memory for the first few seconds after the page loads.
Somebody typed shopify store setup cost into Google. Somebody clicked a link that said where CTAs actually belong on a page. Somebody opened an email with the subject line the quote you asked about. Each of them arrives with a specific expectation about what they're going to see next.
Then the page opens with Ecommerce Solutions That Scale, and none of them can tell whether they're in the right place.
That's the whole failure. It isn't that the page is bad, or that it doesn't eventually answer the question. It's that the answer to am I in the right place should take no effort at all, and instead the visitor spends their first three seconds reconciling two different sentences. Some of them do it. Some of them go back to the results page, which was one tap away and cost them nothing.
In paid advertising this is called message match, and the standard version of the advice is to echo your ad headline in your landing page headline. Fine advice, if you're running ads. The principle underneath it doesn't need the ad: the first thing on the page should repeat the thing that brought them.
Which is a different job depending on where they came from:
From search. The query. Open Search Console, sort the page's queries by impressions, and read the top five. Your first line should contain the language in them.
From your own site. The anchor text, which matters more than people think. If a link on another page promises where CTAs actually belong, the destination needs to say that near the top, in those words. Internal links make promises too.
From email. The subject line. Someone who clicked the quote you asked about should see the quote, not your homepage.
One caution, because this is where people overcorrect. You're matching the dominant entry point, not stuffing every query into the first paragraph. If two entry points want two different opening sentences, you don't have one page with a headline problem. You have two pages.
There's a related job the first line is doing at the same time, which is saying plainly what you sell that's the first thing a hero section has to get right, and it's a separate test from this one.
Let me save you some money. Stop testing button colours.
Almost every list of conversion tips has a section on it, usually with a story about an orange button beating a green one somewhere in 2010. Colour does one job: it makes the button findable against everything around it. Once it's clearly the most prominent thing on the screen, the rest is decoration, and the argument about which shade wins is an argument about a variable that has already finished paying out.
The words on it haven't.
Michael Aagaard, the same person who ran the fold test, changed a button from Start your free 30 day trial to Start my free 30 day trial and recorded a 90% lift in click-through. One possessive. Nothing else on the page moved.
That figure is from 2013 and it lives in a blog post rather than a paper, so treat the size of it as directional. The mechanism holds up better than the number does, and the mechanism is the useful part.
A button written in the second person is you giving an instruction. Start your trial. Get your quote. Book your call. It's the page telling the visitor what to do next, which puts the reader in the position of being asked for something.
Switch it to first person and the button stops being your instruction and becomes their sentence. Start my trial. Send my question. The visitor isn't complying, they're claiming something. Same click, different transaction.
Underneath that sits the more important test, and it works whichever pronoun you use: does the button name what they get, or what they have to do?
The effort | The payoff |
Submit | Send my question |
Sign Up | Get my project estimate |
Contact Us | Ask about my project |
Book a Call | See if we're a fit |
Every label in the left column describes work. Every one on the right describes a return on it. Neither column costs anything to write, and the difference between them is the difference between a form that reads like a task and one that reads like an answer.
Two things worth checking after you rewrite it. Read the button as the end of the sentence I want to… — if the result is something a person would say out loud, it's right. And keep it under about five words, because a button that needs a clause has stopped being a button and become a sentence with a border around it.
Two pieces of standard advice collide here, and working out which one is right tells you most of what you need to know.
The first says put your call to action in several places down the page, because people decide at different points and you want a button wherever that happens. The second says repeating the same button trains people to stop seeing it, usually called scroll blindness.
I went looking for the evidence behind the second one and there isn't any. What exists is banner blindness, which is real, well documented, and about something else. Research going back to a 1998 study at Rice University, and a decade of eye-tracking after it, found that people filter out page elements that look like advertising. Coloured strips running the full width. Starbursts. Boxes sitting where ads usually sit. The filtering is triggered by the shape of the thing, not by how many times it appears.
Which gives you a more useful rule than either version. Your call to action gets ignored when it looks like an ad, not when it repeats.
So repeat it. Same action, same destination, same colour and shape every time, so it reads as one recurring offer rather than four different products.
What changes is the sentence next to it.
Here I'm reasoning rather than citing, and it's worth saying so. Nobody has tested this the way the fold has been tested. But it follows from everything above. If the button's job is to arrive after the questions have been answered, then a visitor two paragraphs in and a visitor at the end of the page aren't holding the same objection. One line of copy can't answer both.
On a service page that usually looks like three positions:
After the value proposition. The reader knows what you do and nothing else. The line beside the button answers is this even for me: "Tell us what you're building and we'll say whether we're the right fit."
After the process or examples. They believe you can do it, and they're now wondering what starting costs them. "Send the details and we'll come back with a rough number. No call required."
After the pricing and proof. The questions are gone and what's left is inertia. Nothing clever needed: "Ask about my project."
Same button, three times, one destination. If you find yourself writing a genuinely different action at each position, quote here, call there, newsletter at the bottom, that's not this problem. That's competing calls to action, and it costs you in a different way.
The other rule you'll find everywhere is to keep it short. Clean layout, minimal copy, plenty of white space, nothing to distract. Long pages are treated as a failure of editing.
That advice is right for a particular kind of decision, and researchers have a name for the split. When something barely matters to someone, or they can't be bothered to think hard about it, they decide using shortcuts. Does this look professional. Have I heard of them. Is there a badge. When the decision does matter and they're willing to put effort in, the shortcuts stop working and they engage with the actual argument. They read.
Which route your visitor is on isn't a personality trait. It's set by what the decision costs them.
Someone downloading a free template is on the first route, and there's nothing to read because there's nothing at stake. Someone about to spend $4,000 with a company they found ten minutes ago is on the second, and a beautifully minimal page gives them nothing to engage with. They don't convert and they don't bounce in frustration either. They go and find a page that answers them, which is usually a competitor's.
So the length of the page isn't a style decision. It's a function of how much someone is risking.
Long is not the same as padded, and that's where people go wrong acting on this. A page can run 2,000 words and still answer nothing, because the words are describing your process in general terms, or restating the value proposition in four different ways, or explaining your commitment to excellence.
Here's the test. Every block on the page has to be retiring one of the questions you wrote down in Step 1. Put the question in the margin beside it.
If you can name the question, the block earns its space no matter how long the page gets. If you can't, cut it, and the page gets shorter without losing anything.
That list of questions is worth building properly, because most of them are the same ones stopping people from getting in touch at all what visitors are actually hesitating over covers where those come from.
The attention ratio numbers from earlier are the most quoted argument for removing things from a page, and they're solid. The trouble starts when people apply them to a page that isn't a campaign landing page.
HubSpot's own version of that test is the one worth knowing about. They stripped the links out of five landing pages, and conversions rose on four of them. Not five. One page did worse without its links, which is the result that gets left out of the summaries, and it's the reason "remove everything" isn't a rule.
So the question isn't how many links are on the page. It's what each one is doing there.
Two categories, and they behave completely differently.
Things that compete. These offer the visitor a different destination than the one you want, and they exist because somebody had a page that needed traffic. Related blog posts at the bottom. Social icons. A carousel of your other services. A footer holding thirty links. Every one of them is an exit dressed up as helpfulness, and on a page whose job is a single enquiry most of them aren't earning their place, though I'd keep anything a genuinely interested reader would have gone looking for anyway.
Things that orient. These tell the visitor where they are and what else exists, which is a requirement rather than a distraction. Your main navigation. A breadcrumb. A link to pricing, which is the thing a hesitating buyer will go looking for whether you provide it or not.
That second category is why the standard advice breaks on a service page. Strip the navigation off one of eight service pages and the visitor can't tell whether they're on a real website or something you built to sell them one. The nav probably isn't what's costing you the conversion. The eleven other things around it are.
Here's the test that separates them. For each element, ask whether removing it would make the decision harder for a serious buyer. Removing your pricing link makes it harder. Removing your Instagram icon does not.
One thing that isn't in either category, because it isn't a link at all: the number of actions on the page. Several buttons all asking for different commitments is a separate problem with a separate cost, and it's covered in ten design mistakes that quietly lose you customers.
The last thing every landing page guide tells you to add is urgency. A countdown clock, a note that only four spots remain, a line saying eleven people are looking at this right now.
There's good research on what those actually do, and it isn't what the guides assume.
A team at Princeton crawled about 11,000 shopping sites and catalogued the pressure tactics on them. They found several hundred countdown timers, and a substantial share of them were fake. The clock hit zero and the offer stayed exactly where it was, or the timer quietly reset when the page reloaded. They also found something more useful than the count: more than twenty separate companies selling these widgets as an off-the-shelf product, ready to drop into any site.
Sit with that second finding for a second. The widget on your page is the same widget your visitor has already seen on dozens of other sites, because a handful of vendors sold it to all of them. It doesn't read as this offer ends soon so much as this company bought the urgency plugin.
A 2023 study put numbers on the cost. Just over 200 people were shown versions of a site with and without limited-time pressure cues, and the versions carrying them produced measurably more frustration, stress and irritation than the version without. Not more purchases. More annoyance.
All of that was measured on ecommerce, which is the usual problem with conversion research, and here it works in your favour. You can't put a countdown on a $4,000 website build without it being visibly absurd. The fake version isn't available to you.
The real version is, and almost nobody uses it. For a service business the genuine constraint is capacity, and capacity is a fact rather than a tactic.
"We take on two builds a month and October is full." "The next available start is mid-November." "We're quoting for January now."
None of that is pressure. It's an answer to a question the buyer already had, sitting alongside the price and the timeline from the last section, and it does the urgency job without any of the machinery, because it's true.
Which gives you a clean rule. If the constraint would still exist whether or not you mentioned it, put it on the page. If it only exists because you put it there, take it off. That's the same standard the rest of your trust signals have to meet, and what makes a service site believable goes through the others.
The order is the part people get wrong. Almost everyone starts with the button, because it's the visible thing and moving it takes ten minutes. It's the last step, not the first.
We start with the questions. Whatever a buyer has to have answered before they'd get in touch, written down in their words rather than ours, usually pulled from the emails the owner is already answering one at a time. That list decides how long the page is, what order it runs in, and where the button can sit without arriving too early.
Then the entry points. What people searched, what the link that brought them promised, and whether the first line on the page repeats any of it.
Then the subtraction, which is quick and mostly obvious once the first two are done.
The button copy comes last, and by then it usually writes itself, because you know exactly what the person has just finished reading. If the page ends in a form, that has its own set of decisions.
If a page is getting traffic and no enquiries, this is the sequence we work through, and the answer is nearly always in the first step rather than the last.
As long as it takes to answer the questions standing between a buyer and an enquiry, and not a word longer. For a free download that's short. For a project costing several thousand, it's usually longer than the owner is comfortable with. Count the questions, not the words.
For most small service businesses the service page is the landing page, and building a separate one is work you don't need. A dedicated page earns its keep when you're running paid traffic to a specific offer, because then you control exactly what the visitor saw immediately before arriving. Without ads, you're better off making the service page do the job properly.
One action, repeated as often as the page is long, with different copy around each instance. The problem is never repetition. It's several different actions competing, which is a different mistake with a different fix.
Web Designer & Developer, AppDeel
Hafizur Rahman designs business websites at AppDeel. He has spent around five years on website design and redesign projects for brands across several industries, and works mainly on the practical side of design. Page structure, hierarchy, mobile behaviour, and the paths that lead a visitor to enquire or buy.
Read more posts by Hafizur Rahman →Why cutting fields is the most overrated fix, and what actually changes how many people finish your form.
Your site gets visitors but the form stays empty. These 7 reasons explain why people who could contact you don't and exactly what to change.
A complete free method to scan and clean WordPress malware by hand. No security plugin, no paid scanner, no AI. Includes the SSH path and the cPanel-only path.
How attackers actually break into WordPress sites, what they do with the access, and why most owners discover the problem months after it started.
Tell us about your business today — get a free quote within 24 hours. No pressure, no confusing tech talk.
