EMPIREYO
← Insights
Online marketingupdated 12 November 202617 min read

Conversion optimisation: test small, win big

Attracting more visitors costs money and effort. Looking at why the visitors you already have do not request, call or buy is often the cheapest step. This is how to go about it, even with a small site.

In short

Conversion optimisation means that more visitors do what they came for: book an appointment, request a quote, sign up or buy something. First choose one goal and take a baseline measurement. Find where visitors drop off and fix what is clearly better without a test. A/B testing only makes sense with enough visitors and conversions to see a real difference.

When people hear conversion optimisation, many think of a button that turns from green to orange, and a chart that shoots up afterwards. In practice it starts much earlier and is much more ordinary. You ask yourself what a visitor comes to your site to do, you look at where they drop off, and you remove what holds them back.

Below you will read how to go about it, when an A/B test makes sense and how to keep it honest. The examples come from sites we built ourselves. How many extra requests they brought in, we have not measured; so that is not in here either.

What is a conversion?

A conversion is an action by a visitor that counts for your business. That is by no means always a purchase. For a dentist it is a registration, for a lawyer a first message, for a taxi company a ride request and for a repair shop an appointment. Google Analytics calls such an action a key event: an event that measures an action that is important to the success of your business.

On top of that there are intermediate steps: someone opens the pricing page, starts a form or taps the phone number. These show where things go wrong, but they are not the goal.

The conversion rate is the number of conversions divided by the number of visits in the same period. A worked example: if a month brings 600 visits and 12 requests, then 1 in 50 visits leads to a request. That figure only becomes useful when you compare it with your own site last month, per page or per device.

Which conversion do you choose as your main goal?

Choose the action with which your visitor solves their problem and you gain a customer. That is often not what you would most like to see yourself. At Daily Phones most visitors arrive with a broken screen, not a shopping plan. That is why the repair planner comes first on that site and buying is the second step. At Atiq Tandartsen registering is the most important step, and it is one click away on every page.

How do you measure before you change anything?

Without a measurement, you do not know after a change whether things got better, worse, or whether it was chance. So start with a baseline: a fixed starting point you compare against later. This is how:

  1. Write down your goal. One main conversion, in plain words: “a completed quote form” or “a booked appointment”.
  2. Check that the measurement works. Submit a request yourself and see whether it shows up in your statistics, once and not twice. Compare the count with your inbox or your calendar. A measurement that counts double or misses half sends you in the wrong direction.
  3. Measure the steps before it too. How many people open the form, and how many send it? Without those intermediate steps you can see that it leaks, but not where.
  4. Choose a fixed period. Take at least a few full weeks, so that a busy Monday or a quiet weekend does not decide the picture.
  5. Note what else was going on, such as a holiday or an advertising campaign. Later, that sometimes explains more than your change.

No statistics, or many visitors who do not consent to measurement cookies? Then your own records are a measurement too: the number of requests in your inbox, appointments in your calendar or calls per week. How to set up a measurement that adds up, you can read in GA4 done right and Cookieless measurement.

Where do visitors drop off?

Put your visitor’s steps one below the other, from arrival to confirmation, and look per step at how many people carry on. The biggest drop is your first candidate. It is often one of these places:

  • The first page after a search or an ad. The visitor does not immediately find what they were looking for and goes back.
  • The price or the terms. Someone wants to know what it costs or what happens after sending, and cannot find it anywhere.
  • The form. Too many fields, a required field that does not feel necessary, or an error message that does not say what is wrong.
  • The phone. On a small screen a button falls below the fold, or a drop-down list becomes awkward.
  • After sending. No confirmation, or no idea when you will hear back. You do not see that as a drop-off in your figures, but it can lead to duplicate requests and extra phone calls.

Figures tell you where it leaks, but not why. You only see that when you watch. Ask a few people from your audience to make a request on their own phone, without help, and see where they hesitate. If you want that done thoroughly, a UX audit fits. The US government’s plain language guide puts it briefly: a comparative test tells you whether something works, but not why. Eight common points of friction, each with its fix, are in the UX mistakes that cost customers in 2026.

Checklist: what do you check at each step of the visitor?

Go through this list on a phone, for your most important page. Anything you answer no to is a candidate for improvement.

StepWhat the visitor wants to knowCheck
ArrivingAm I in the right place?Does the heading say in plain words what you do and for whom? Does the page match the search or ad the visitor came from?
First screenWhat can I do here?Is the main button visible without scrolling? Does the button say what happens when you tap it?
DoubtWhat does it cost, and can I trust this?Is the price, or the way you price, shown before the form? Can address, phone number and genuine reviews be found?
FormHow much work is this?Do you only ask for what you really use? Are the labels above the fields, not just grey text inside them? Is it clear what is optional?
A mistakeWhat should I do differently?Does the error message say what is wrong and how to fix it? Does what the visitor already entered stay in place?
SendingDid it work, and what now?Is there a confirmation on screen and by email? Does it say when they will hear back?
LaterCan I stop again?Is cancelling or unsubscribing as easy to find as signing up?

What can you improve without a test?

Many improvements do not need a test, because they remove an obvious barrier: a field nobody needs, an error message that says nothing, a price that is nowhere to be found. Fix it and measure whether the number of requests moves with it. That is what we call measuring before and after.

Forms

The form is the last stretch before the conversion, and so a place where every barrier counts. Three rules help:

  • Only ask for what you need. The UK government only lets its teams add a question once they know why they need the answer and what they will do with it. A question you cannot justify goes. At MijnbouwSubsidie three buttons lead to one short form: name, email, phone, place of residence and your message.
  • Start with the visitor’s question, not their details. The booking wizard of Sittard Taxi does not ask for your details first, but for what you need. At Ikiz the form first asks what it is about. The UK government also advises starting with the questions that show whether someone is eligible for the service at all, so you do not waste anyone’s time.
  • One thing per screen, if the form is long. The UK government advises starting with one thing per page: one question, one decision or one piece of information. That makes errors easier to fix and shows you per question where people stop.

Watch the fields themselves too. Grey example text in a field is no substitute for a label, the W3C writes: that text disappears as soon as you start typing, and screen readers do not read it as a label. Put the label above the field and indicate which fields are required or optional.

Buttons and copy

A button that says “Submit” says what the computer does. A button that says “Request a quote” or “Book your appointment” says what the visitor gets. The same goes for the line under the button: when will you hear back, and does it cost anything? We wrote a separate article about those small words: Microcopy: small words, big difference.

Speed

A page that stays blank for a long time on a phone loses visitors before they have seen your offer. Measure on a phone, not just on your laptop, and look especially at the pages where your conversion starts. How to do that and what Google calls fast, you can read in Why every second of load time counts.

Trust

Someone making a request wants to know who they are dealing with. Show where you are based, how to reach you and what happens after the request. Show the price, or the way you price, before the form, not after it; the Dutch Authority for Consumers and Markets (ACM) writes that the price must be clear right at the start of an offer. And only show genuine reviews: according to the ACM, whoever shows reviews must also say whether and how they check that they are genuine.

“A test with too few visitors is not proof. It is an opinion with a chart attached.”

When does A/B testing make sense?

In an A/B test there are two versions of a page, A and B, and chance decides which version a visitor gets. Because chance does the splitting, the two groups are alike, and you can attribute a difference in conversion to the difference in the page.

An A/B test makes sense when three things are true. You have enough visitors and enough conversions on that page. You test one clear change with a reason behind it, for example: “If we put the price above the form, more people will request a quote, because now they first have to call to hear it.” And you agree in advance how long the test runs and what you will do with the result.

What is sample size, and how many visitors do you need?

The sample size is the number of visitors you need in your test to be able to trust a difference. Think of a coin. If you toss it ten times and it lands heads seven times, you do not yet know whether the coin is biased; that may be chance. After a great many tosses you do know. Conversions work the same way. With five requests for version A and eight for B, the difference may be pure chance.

How many visitors you need depends mainly on two things: how many conversions you have now, and how large the difference is that you want to be able to see. The smaller the difference, the more visitors. Guidance from the UK Office for Health Improvement and Disparities (OHID) simply names the drawback of A/B testing: you need a lot of users to make a result statistically reliable, and you base the length of the test on your number of visitors per month and the change you expect. So we do not give a fixed number. Have it calculated in advance with a sample size calculator, or by someone who knows statistics.

Two mistakes are tempting with small tests. Stopping as soon as version B is ahead, when that may be a chance spike. And changing several things at once, so you do not know what made the difference.

What do you do if you have too few visitors?

Then an A/B test does not give a reliable result. First improve what demonstrably gets in the way, and:

  • Watch real people. A few sessions in which someone makes a request while thinking aloud show why people hesitate. You do not see that in figures.
  • Measure before and after over a longer period, and write down what else changed in that time. That is not hard proof, but it is an honest picture.
  • Ask your customers. One question in the confirmation email, such as “What almost made you hesitate?”, brings answers that no statistic gives.

If you do test, mind Google. Google asks you to show search engines the same pages as visitors, to have test versions point to the original (canonical, or a temporary 302 redirect) and to remove the test as soon as it is finished.

How do you keep it honest?

Conversion optimisation is about removing barriers, not about adding pressure. The Dutch consumer authority ACM sets out where the line lies in its guidelines on protecting the online consumer. Influencing is allowed, the ACM writes, but goes too far if you steer the average consumer towards a choice they would not have made had they been able to choose in a considered way.

What the ACM names in those guidelines, and so what you do not do:

  • False urgency or scarcity. No messages such as “only a few left” or “offer valid for a limited time” if that is not true.
  • Pre-ticked boxes for something the consumer has to pay for. And no pre-ticked newsletter for people who are not yet customers.
  • Messages about other customers that are not accurate, such as a number of viewers that is really the total number of visitors on the site.
  • Fake reviews. And if you paid a customer for a review, you say so.
  • Making cancelling harder than signing up. Whoever takes out a subscription online must also be able to cancel it online, without a string of questions or messages such as “We are sorry to see you go”.
  • Misusing A/B tests to find the version in which required information stands out as little as possible.

The ACM also gives businesses two pieces of advice that apply to every designer. Do not steer your designers and marketers only on more conversions, but also on an informed and suitable choice by the customer. And do not only measure sales, but also whether customers would have made the same choice without your influence.

An example from our own work: at TaalPro you can enrol online and also cancel online, just as easily as enrolling. The eight tools on that site, such as the level test, do not ask for an email address.

When this does not apply

This article is about sites that want a request, appointment, registration or order. If your site is mainly information for existing customers, you measure something else, for example whether people find their answer without calling.

The explanation of A/B testing is a simple explanation, not a statistics lesson. How many visitors your test needs depends on your own figures; have it calculated before you start. What the ACM writes, we pass on as general information, not as legal advice. The ACM is the Dutch regulator, and its guidelines are aimed at those who sell online to consumers or design for them. If you only sell to businesses or work outside the Netherlands, they are not aimed at you, though they remain a useful yardstick.

What we do not know: our case pages show no conversion figures, and we have not measured how many extra requests a short form, a button or a faster page brought these clients. The examples show what was built and why, not how much it yielded.

Frequently asked questions

What is a conversion?+
An action by a visitor that counts for your business. That is not only a purchase, but also a quote request, a registration, an appointment or a phone call. Choose one as your main goal and measure the steps towards it.
How many visitors do I need for an A/B test?+
That depends on how many conversions you have now and how large the difference is that you want to see. The smaller the difference, the more visitors. Have the sample size calculated before you start; there is no fixed number for every site.
What can I improve without testing?+
Obvious barriers: fields you do not use, error messages that say nothing, a price that is nowhere to be found, a button that does not say what happens and a slow page on the phone. Fix them and measure before and after.
What is a baseline measurement?+
A fixed starting point: how many visits and conversions you had over a set period, before you change anything. First check with a test request of your own that the measurement is right, and note what else was going on in that period.
May I use urgency or scarcity on my site?+
Only if it is true. The Dutch ACM writes that you may not create a false impression that someone has to decide quickly, such as with an unjustified message that only a few are left or that an offer is valid for a limited time.

Wondering where your site leaks? Send us the address and your goal, and we will look at where visitors drop off and what to tackle first. More about how we look at that is under UX audit. You work directly with Nawid, with a fixed team of five specialists.

Sources

Consulted on 3 November 2026.

NR
Nawid RasaFounder of Empireyo · in design, development & strategy since 2014

Also worth reading: Marketing that pays off in 2026.

Written with the help of AI, reviewed and edited by Nawid Rasa, who is responsible for the content. Sources are listed below the article. Spotted an error? Email hi@empireyo.com.

Ready for measurable growth?

We set up a CRO programme that keeps learning, and let the numbers do the work.

What do you have in mind?
What budget do you have in mind?
When do you want to launch?
Almost there. How do we reach you?
Request sent.

Your request has reached us. You will usually hear back within two working days. Prefer direct contact? Email hi@empireyo.com or message us on WhatsApp with the chat button at the bottom right.

Reply within two working days · No obligation · privacy