A/B testing guide: how to run website tests that hold up, from hypothesis to decision
Hypotheses, sample size, duration and significance: how to run A/B tests on your website without drawing false conclusions.
How to arrange a campaign page with a single goal: promise, proof, objections, form and the mistakes that make customers cost more.
A good landing page structure serves a single goal and follows a clear order: a promise that continues the ad, the benefits, real proof, the next steps, answers to objections and a short form, with the call to action repeated. No full menu and no links that lead the visitor away. Unlike your homepage, the visitor here has just one decision to make.
A landing page, or campaign page, is the page a visitor reaches after clicking an ad, a link in an email or a QR code. It has one job: to prompt a specific action, such as a quote request, a booking, a sign-up or an order.
A homepage does something else. It has to serve several audiences: new customers, existing customers, job candidates, partners, journalists. That is why it has a menu, varied sections and many ways out. If you send paid traffic there, you are paying for the visitor to get lost.
The difference shows in the decision. The homepage answers “who are you?”, while the landing page answers “why should I do this, now?”. As a principle, every element that does not lead towards the desired action is a distraction and needs a good reason to stay.
Not every service page is a landing page. A service page can also be the destination of a campaign, if it has a clear message and a single goal, but it normally carries information for a wide audience, including search engines. A campaign page is built for one message and one audience.
The order below works for most offers. It is not a fixed recipe; the page gets shorter for simple offers (a free ebook) and longer for expensive or complicated ones, where people need more answers.
| Section | Its role | What it contains | Typical mistake |
|---|---|---|---|
| 1. Minimal header | Shows who you are, without distractions | Logo, perhaps a phone number; no full menu | The site’s entire menu |
| 2. Promise | Confirms the visitor is in the right place | Benefit headline, subheading, image, button | A vague headline, or just the company name |
| 3. Problem and benefits | Shows you understand the customer | 3–5 benefits in the customer’s language | A list of technical features |
| 4. How it works | Reduces uncertainty | 2–4 simple steps, what happens after the click | Vague steps, or none |
| 5. Proof | Builds trust | Real reviews, verifiable figures, permitted logos, guarantees | Invented testimonials, or ones with no source |
| 6. Objections | Answers “but what if…” | FAQs, prices or ranges, conditions | Dodging the obvious questions |
| 7. Form and button | Captures the action | Few fields, a verb on the button, a note about data | A long form, a button saying “Submit” |
| 8. Minimal footer | Compliance and trust | Company details, privacy policy | Links that take people off the page |
The main button is normally repeated a few times, at natural intervals: next to the promise, after the proof and at the end. All of them lead to the same goal. If the page offers two competing actions, for example “call now” and “download the guide”, they steal attention from each other.
The first seconds decide whether the visitor stays. The area visible without scrolling must answer three questions: what do I get, why does it matter to me and what do I do now. The headline carries the weight: it states the concrete benefit, not the name of the offer. “Monthly accounting for limited companies, at a fixed price” says more than “Professional accounting services”.
The subheading adds the detail that makes the promise believable: who it is for, what result and in what timeframe, without promising what is not in your hands. The image helps if it shows the product, the result or real people; generic stock photos tend to have little effect.
The button should be visible from the first screen, with text that says what comes next: “Get a quote”, “Book a 15-minute call”, “Download the guide”. Avoid “Submit” and “Learn more”, which say nothing about the outcome. If the offer is urgent, a tap-to-call phone number on mobile can beat the form.
The most common cause of a weak landing page is not the design but the break between ad and page. If the ad promises “mobile car valeting at your door” and the page opens with “Complete cleaning services”, the visitor no longer feels they have arrived in the right place and leaves. The same words, the same offer and the same price should appear in both places.
In Google Ads, landing page experience is one of the components that influence Quality Score, so a well-matched page can help with click costs as well as with conversion. If you have ad groups for different intents, build different pages, or at least adapted headlines. Traffic quality also depends on exclusions: negative keywords in Google Ads keep you from sending to the page people who are not looking for what you offer.
With social media ads, the person has not searched for anything, so the page has to build desire more carefully, with images, proof and explanation. We discuss the difference between the two channels in the article Google Ads vs Facebook Ads.
A stranger will not take your word for it, so the page must give reasons. The best proof is the kind they can check: reviews with a name and a source (for example from your Google profile), examples of work, years in business, relevant certifications, clear guarantees, a returns policy.
Do not invent testimonials, do not use figures you cannot support and do not display client logos without their consent. Besides damaging trust, misleading advertising can have legal consequences. If you have no reviews yet, use other proof: a transparent process, examples, information about the team, the company’s address.
Objections are the reasons a person does not click. The most common are price, risk, time and effort. Answer them directly, in an FAQ section:
Each extra field usually reduces the number of people who finish the form. Ask only for what you need to take the next step. For a quote request, a name, a phone number or email and a free text box for details are often enough. If you need to qualify leads (budget, size), you can add one or two questions with simple answers, but each must justify its place.
A few practical rules:
Forms collect personal data, so state the purpose of processing clearly and link to your privacy policy; do not pre-tick marketing consents. This text is informative and is not legal advice; check your implementation with a lawyer or your data protection officer.
A large share of ad traffic often comes from phones, so design for small screens first: readable text without zooming, buttons that are easy to tap, a form that does not break. A page that loads slowly loses visitors before they see the promise; details are in our guide to website speed and Core Web Vitals. Compress images and remove scripts you do not need.
Without measurement, you do not know whether the page works. Track at least: visits from each source, conversion rate, the time taken to complete the form and the quality of enquiries. You will find how to set up events in the guide to conversion tracking, and how conversion rate is calculated in the article on conversion rate.
Changes are tested one at a time. You change the headline, or the button, or the form, not everything at once, and you let the test gather enough data; for the method, read A/B testing on your website. With low traffic, one well-reasoned change beats a formal test that never collects enough data.
For a quick check of an existing page, you can use the landing page audit in our Lab: it walks through the points above and shows where you are losing visitors.
When we build or review a campaign page, we start from the ad that feeds it and from the objections customers raise in real conversations. We write the copy before the design, because the structure follows from the arguments, not from a template. Then we measure and adjust on the basis of data, not of opinions within the team.
This is part of our conversion rate optimisation service. We do not promise a particular conversion rate, because it depends on the offer, the price, the audience and the traffic; we promise a page with a clear message, a simple path and measurement you can access.
As long as the offer needs. For something simple and cheap, such as a free guide, a short page is enough. For expensive or complicated services, people ask for more answers and more proof, so the page gets longer. The right test is whether each section answers a real question from the visitor, not the word count.
Usually not. A full menu offers ways out and distracts from the goal. Keep a minimal header with the logo and, if useful, a phone number. If visitors need more information before deciding, put it on the page, in sections, rather than on separate pages that take them away.
As few as possible: only what you need to take the next step. For many quote requests, a name, a contact detail and a free text box are enough. Each additional field usually reduces the number of people who finish, so add one only if it saves you real time or helps you qualify the enquiry.
Sometimes. If the page has a clear message, a single goal and a good match with the ad, it can work. Usually, though, a service page speaks to a wide audience and links to the whole site. A page dedicated to the campaign, with a reduced menu and an adapted message, normally converts better and is easier to measure.
Long enough to collect relevant data: usually a few weeks, depending on traffic. If the page is clearly broken (a faulty form, a button not visible on mobile), fix it at once. For optimisation, change one element at a time and compare periods, watching for season and for changes in the ads.
There is no threshold that suits everyone. It depends on the offer, price, channel and quality of traffic: a quote request for an expensive service has a different rate from the download of a free guide. Compare the page with itself over time, and with the cost you can afford to pay for a new customer.
Laborator
Lab tools that help you put this guide into practice.
Blog
Hypotheses, sample size, duration and significance: how to run A/B tests on your website without drawing false conclusions.
Match types, starter lists, the search terms report and a weekly routine that cuts spend on searches you never wanted.
Basic settings, key events, five useful reports and what to ignore in GA4, explained for people who are not analysts.
Tell us where you are now and what you want to achieve — you will get a concrete proposal with services, timeline and a transparent budget.