What is CRO? A guide to conversion rate optimisation

What is CRO?

CRO (conversion rate optimisation) is the systematic improvement work carried out so that a larger share of the visitors to your site or app take the action that is valuable to you. That action may be a purchase on an e-commerce site, an appointment request at a clinic or a quote form on a corporate site. Conversion rate is the ratio of visitors who take that action to all visitors.

CRO is not a matter of design taste. It does not progress through isolated ideas such as changing a colour or a button, but by researching why users give up and measuring changes based on the findings. In this article we outline the logic and steps of CRO; the details of our measurement infrastructure and testing method are on our CRO and analytics service page.

Why does CRO come before advertising?

The ad budget brings visitors to the site; whether those visitors become customers is decided by the site. Sending more traffic to a site with a low conversion rate means losing more people through the same leak. Conversely, an improvement in conversion rate means more sales with the same budget and directly lowers customer acquisition cost.

There is an indirect effect too. The automated bid strategies of platforms such as Google Ads and Meta learn from conversion data. More, and more consistent, conversions give the algorithm a stronger signal. That is why the site is one of the first places businesses that want to increase advertising returns should look; we also cover this in our article on ROI-focused performance marketing.

How is CRO research carried out?

Good CRO work takes time to understand the problem before making changes. Different research methods answer different questions; the most reliable picture emerges when their findings confirm each other.

Analytics data

Analytics tools such as GA4 show which page users enter on and at which step they leave. Funnel reports and device and traffic source breakdowns tell you where the problem is; but they do not tell you why.

Heatmaps and session recordings

Heatmaps show click and scroll behaviour, while session recordings show anonymised replays of individual visits. They reveal a button the user cannot see, an image they think is clickable or a field where they get stuck in a form.

User testing

Asking a few people who resemble your target audience to complete a specific task while thinking aloud quickly reveals confusion that analytics data cannot explain. Even with a small group of participants, the most obvious problems usually come to light.

Surveys and customer feedback

Short on-page surveys ask people who leave without buying what was missing. Questions to customer service, live chat logs and product reviews are also valuable sources for learning users' language and objections.

How do you write a hypothesis?

Research findings are turned into testable hypotheses. A good hypothesis contains the observation, the change to be made and the expected result together: "In session recordings we saw users go as far as the checkout page to see the shipping cost and then leave. If we show shipping information on the product page, we expect the abandonment rate at the checkout step to fall."

Hypotheses are ranked by likely impact, ease of implementation and confidence in the finding; the team starts with the highest-value work. We describe how the same experimental logic applies beyond the site, across the whole customer journey, on our performance marketing service page.

What is A/B testing?

A/B testing is a controlled experiment in which visitors are randomly split into groups, some are shown the existing page (the control) and the rest a modified version (the variant), and the groups' results are compared. Random assignment ensures that external factors such as season, campaigns or traffic source are spread evenly across the groups; so the difference largely comes from the change made. Set-ups where several versions are tried at the same time are called A/B/n tests, and those where several elements are changed together are called multivariate tests.

What does statistical significance mean?

The difference seen in a test result may come from chance rather than a real effect. Statistical significance tells you whether the probability that the observed difference can be explained by chance falls below a threshold set in advance. The number of visitors and conversions needed for a reliable result should be calculated before the test starts. On low-traffic pages, telling small differences apart takes a long time or is not possible.

The most common mistake is to keep checking results while the test is running and stop it at the first positive sign. This habit markedly increases the chance of picking the wrong winner. Run the test for the period set in advance, covering different days of the week, and stick to a single primary success metric.

Which areas can be improved fastest?

Forms

Limit form fields to the information the sales team actually uses. Write field labels clearly, show error messages next to the field in plain language, and make sure the right mobile keyboard opens for phone and email fields. Long forms are often easier to complete when split into steps. Telling the person who submits the form what happens next also reduces abandonment.

Checkout page

In e-commerce, the most expensive losses happen at the checkout step. The following checks deliver quick wins on most sites:

• Show the shipping cost and delivery time before the checkout page.
• Offer guest checkout without creating an account.
• Support common payment methods and digital wallets.
• Keep trust signals such as secure payment, return terms and contact details visible.
• Make sure the discount code field does not send users off to other sites to hunt for codes.

Mobile experience

In many sectors most traffic comes from mobile, yet design and testing are still mostly done on desktop. Buttons within easy reach of the thumb, readable text size, a sticky call to action and checkout steps that work smoothly on mobile are the basic checkpoints. Examine the funnel separately by device; the overall average can hide a problem on mobile.

Page speed

A slow page loses the user before they see the content. Large images, unnecessary external scripts and slow server response are the most common causes. The Core Web Vitals report in Search Console and PageSpeed Insights are a good starting point for finding problem pages.

How is CRO applied on corporate websites?

On corporate and service sites, a conversion is usually not a purchase but a quote request, a meeting or a catalogue download. Traffic and conversion volume are often too low for A/B testing. In that case, improvements based on research findings and usability principles are applied, and before and after data are compared carefully.

The most common problems on corporate sites are home pages that do not explain at first glance what the company does, service pages with no call to action, a lack of trust signals such as certifications, process descriptions and team information, and uncertainty about what happens after the form. It is also necessary to connect CRM data to measurement so that you track not just the number of forms but the enquiries that turn into sales.

Common CRO mistakes

• Copying changes seen on competitor sites without doing research.
• Starting tests before measurement is verified; duplicated or missing events invalidate the result entirely.
• Running several interacting tests on the same page at the same time.
• Counting only intermediate steps such as clicks or add-to-basket as success, without looking at sales and enquiry quality.
• Not recording losing tests and retrying the same idea shortly afterwards.
• Publishing the winning version and not tracking its effect in later periods.

For the tools on the measurement side, see our Google platform page, and to review your site's conversion flow together, get in touch via our contact page.