Conversion rate optimisation is about getting more of your existing visitors to act. This guide covers how to measure a conversion rate, what to fix or test first and how to run a fair test.
Conversion rate optimisation is the practice of increasing the share of website visitors who take a wanted action, such as sending an enquiry or buying. The conversion rate is the number of people who completed the action divided by the number of visitors, multiplied by 100. Fix clear faults first, then test one change at a time.
What conversion rate optimisation is
Conversion rate optimisation, shortened to CRO and spelt optimization in US English, is the practice of raising the share of visitors who take an action you want. The action is called a conversion. For a shop it is usually a purchase. For a service business it is an enquiry, a booking, a call or a message.
You choose what counts. A main conversion is the action that earns money, such as a quote request. Smaller ones sit on the way to it, such as adding a product to a basket or opening the contact page. Define the main one first, because every later number depends on it.
CRO sits after the work that brings people to the site. Search, adverts and social media bring visitors. CRO improves what happens once they arrive. A site that already has visitors but few enquiries has two options: bring more visitors, or help more of the current ones to act. Doing the second can be the quicker route, and it also makes every later visitor worth more.
It is not a set of tricks. False scarcity, fake countdown timers and hidden extras may lift a figure for a while, but they damage trust and may break consumer rules, which differ by country. Honest, clear pages are what the work aims for.
Measuring conversions starts with your analytics. Google Analytics lets you mark important actions as key events, which is explained in the guide to Google Analytics for beginners.
How to calculate a conversion rate
The formula is the number of conversions divided by the number of visitors, multiplied by 100. The result is a percentage. Before you calculate, fix two choices and keep them: what counts as a conversion, and what counts as a visitor. The visitors can be sessions, users, or only the people who reached one page. A rate worked out on different bases is not comparable from month to month.
An invented example shows the arithmetic and is not a benchmark. If 400 people visit a service page in a month and 8 of them send an enquiry from that page, the page's conversion rate is 8 divided by 400, multiplied by 100, which is 2 percent.
Work the figure out in slices as well as for the whole site. Calculate it for each important page, for each source of visitors and for each type of device. The site-wide figure averages away the problems. A page may do well for people who arrive from search and badly for people who arrive from social media, or the reverse, and only the slices show it.
Keep the data clean. Count each enquiry once. Remove spam and test submissions. Filter out visits by your own team if your tools allow it. Analytics can report key events against sessions or users, but the names of its rate metrics change from time to time, so check the current documentation for the exact definitions in your version.
Find out why visitors do not act
Look for the reason before changing anything. Three kinds of evidence help, and a small business can collect all of them without special software.
The numbers
Look at where people enter, which page they leave from and which step of a form or checkout loses them. Compare phones with computers, and compare sources. A page that converts well on a computer and badly on a phone points to a layout or speed problem on small screens.
What people do
Many tools record sessions or draw heatmaps showing where people tap, scroll and stop. They can show a button that nobody finds or a form field that makes people quit. They also raise privacy questions. Read the tool's documentation on hiding form fields and personal data, and ask a lawyer what applies to you.
What people say
Read the messages and enquiry forms you already receive. Ask recent customers what almost stopped them. Watch a few people who have not seen the site try to do one task on it, such as finding a price or booking a visit, and do not help them. What they struggle with is rarely a surprise afterwards, but it is often a surprise before.
Check the basics too: that every form and button works, that the pages load promptly on a phone, and that the site behaves in more than one browser.
What to fix or test first
Separate faults from questions. A form that fails to send, a button that does nothing and a page that breaks on a phone are faults. Repair them without testing. A headline, an offer or a page layout is a question on which reasonable people disagree. Test those.
Start with the pages that matter most: the ones with the most visitors that also sit closest to an enquiry or a sale.
| Area | Typical problem | First change to try |
|---|---|---|
| Headline and opening | It names the business but not what the visitor gets | State the offer, who it is for and where you work, in plain words |
| Call to action | Several competing actions, or vague wording such as "Submit" | One main action per page, with wording that says what happens next |
| Forms | Too many fields, unclear errors, no confirmation | Remove fields you do not use, show clear error messages and confirm receipt |
| Trust and proof | Nothing on the page shows that the business is real and reliable | Add real examples, reviews you hold, contact details and policies |
| Mobile and speed | Slow loading, tiny buttons, content hidden by pop-ups | Compress the heaviest images and make the main action easy to reach with a thumb |
| Message match | The advert promises one thing and the page says another | A page that repeats the promise the visitor clicked on |
| Checkout or booking | Costs appear late, or an account is forced | Show costs early and offer a way to continue without an account |
Wording of the main action deserves care, and the guide to writing a call to action covers it in detail. Pages built for a single campaign are covered in the guide to landing pages and websites.
How to run a test
A test shows two versions of a page at the same time, splits visitors between them at random and measures which gets more of the wanted action. Follow these steps.
- Write a hypothesis in one sentence: because we saw this evidence, changing this element in this way should increase this action.
- Change one thing, or one tightly related group of things, so that you know what caused any difference.
- Choose the single measure that decides the winner, and decide how long the test will run, before it starts.
- Run both versions at the same time. Comparing this month with last month is not a test, because seasons, offers and the mix of visitors all change.
- Do not stop because one version is ahead. Early gaps often close.
- Run the test for whole weeks, since behaviour differs between weekdays and weekends.
- Record the result, including tests with no clear winner. A flat result is still information.
- Apply the winner and remove the test code.
Testing and search engines
Google has published guidance on running tests without harming search. It says not to show Googlebot different content from what visitors see, which counts as cloaking. It advises using a canonical tag on alternate addresses that points to the original, using a temporary redirect and not a permanent one when sending visitors to a variant, and ending the experiment once you have enough data, removing the test elements promptly. A test that runs far longer than needed can look like an attempt to deceive search engines.
How long should a test run?
A test should run until it has enough conversions to tell a real difference from chance, and for whole weekly cycles at least. There is no universal number of days. The time depends on how much traffic the page gets, how often it converts today and how big a difference you want to be able to detect. A small expected difference needs a lot more data than a large one.
Most testing tools can estimate how much data they need or tell you when a result is reliable. Use that guidance, set the end point before you start and keep to it. Stopping the moment one version leads is the most common way to fool yourself.
Do not let a test run on without end either. A long test picks up seasonal shifts and promotions that muddy the result, and Google advises running experiments only as long as necessary.
Do you need a lot of traffic to test?
For a formal split test, yes, and often more than a small site has. What matters is the number of conversions, not the number of visits. A site with a handful of enquiries a month cannot reliably detect the effect of a changed button colour. The test would run for a very long time and still be unclear.
Low traffic does not rule out improving the site. It changes the method.
- Repair obvious faults without testing them.
- Make bigger changes, such as a clearer headline and a stronger offer, instead of small tweaks. Large differences show up in less data.
- Compare before and after with care. Note every other change and any seasonal effect, and treat the result as a hint, not proof.
- Lean on qualitative evidence: watch people use the page, read their messages and ask customers what held them back.
- Check the quality of visitors. If the wrong people arrive, no change to the page will convert them. The guide to lead generation for small businesses covers how to attract the right ones.
- Work on the busiest pages first, since they give the most evidence soonest.
Common conversion rate mistakes
- Testing small details on a page whose offer is unclear.
- Copying a winning test from another site. Their visitors, offer and layout differ from yours.
- Changing several things at once and not knowing which one worked.
- Counting spam submissions or your own visits as conversions.
- Chasing more form submissions while ignoring whether the people are a good fit.
- Using pressure tactics such as false scarcity or countdown timers.
- Making a change that harms readability, or the use of the site by keyboard or screen reader, for the sake of a small gain.
- Running one test and stopping, instead of keeping a log of what was tried and learnt.
A team can help set up measurement, make the page changes and run the tests. The digital marketing service page describes what Liftup does in this area.
Questions and answers
What is conversion rate optimisation?
Conversion rate optimisation is the practice of increasing the share of website visitors who take a wanted action, such as sending an enquiry, booking or buying. It works on the pages visitors already reach, by fixing faults, clarifying the offer and testing changes. It is often shortened to CRO.
How do I calculate a conversion rate?
Divide the number of conversions by the number of visitors, then multiply by 100. Decide in advance what counts as a conversion and what counts as a visitor, and keep both the same from month to month. Work it out for each page, source and device as well as for the whole site.
What should I test first?
Repair plain faults first, such as forms that fail or pages that break on a phone. Then test the headline and opening, the main call to action, the form and the proof on your busiest pages. Choose changes that come from evidence about why visitors leave, not from guesswork.
How long should a test run?
Until it has enough conversions to separate a real difference from chance, and for whole weeks at least. No fixed number of days suits every site. Set the end point before the test starts, use your testing tool's guidance and do not stop because one version is briefly ahead.
Do I need a lot of traffic to test?
For a formal split test, yes, because the number of conversions matters more than the number of visits. A site with few enquiries cannot detect small effects. In that case, repair obvious faults, make larger changes, watch real users and compare before and after with care.
What is a good conversion rate?
There is no single good figure. It depends on the offer, the traffic source, the device and how well known the business is. Compare your rate with your own earlier figures for the same page and source, and judge success by whether enquiries or sales rise.
Does testing affect my search rankings?
It can if it is done carelessly. Google advises against showing search engines different content from visitors, recommends a canonical tag on variant addresses, a temporary redirect when sending visitors to a variant, and removing test elements once the test is over.