Conversion rate optimisation (CRO) is the practice of improving the share of visitors who take a desired action, such as sending an enquiry or buying. It works as a repeating process: define the action, find where and why people drop out, form hypotheses, prioritise them, test, and learn. Lists of “10 button colours that convert” skip the first steps, which is why they often disappoint.
Step 1: Define the conversion and measure it
Pick one primary action per page or journey. In Google Analytics 4, a key event is an event measuring an action particularly important to your business, and any collected event can be marked as one. Track it correctly before changing anything, otherwise you cannot tell whether a change helped. Our guide to what makes a business website convert explains typical enquiry and sales paths.
Step 2: Find where people drop out
Look at quantitative data first: which pages lose the most people, on which devices, from which sources. Then go and find out why. Speed is a common cause. Google’s guidance on Core Web Vitals suggests aiming for Largest Contentful Paint within 2.5 seconds, Interaction to Next Paint of 200 milliseconds or less, and Cumulative Layout Shift of 0.1 or less, measured with real-user data. Our Core Web Vitals business guide explains how to read those numbers.
Step 3: Learn why, not only where
Numbers show where. People explain why. Nielsen Norman Group argues that A/B testing works best when grounded in research, and describes three useful methods:
- Short surveys asking why visitors came and whether the visit succeeded, which can uncover objections that differ from assumptions such as price.
- Usability testing, where about five users can reveal most major problems and flows that no button tweak would fix.
- Tree tests to confirm people can find things before you test navigation.
Step 4: Write hypotheses
A good hypothesis states the cause, the change and the expected effect: “Visitors do not see how quotes work, so adding a three-step explanation above the form should increase completed enquiries.” The same source warns that testing every idea at random wastes time and exposes users to a weaker design for longer.
Step 5: Prioritise
Score each idea on a simple scale and start with the best ratio of impact to effort.
| Criterion | Question |
|---|---|
| Evidence | How strong is the research behind it? |
| Impact | How many people does it affect, and how close to the conversion? |
| Effort | How long will it take to build and check? |
| Risk | Could it harm trust, accessibility or brand? |
Fix obvious problems (a broken form, an unreadable page on mobile) directly rather than testing them. Those are repairs, not experiments.
Step 6: Test fairly
A/B testing compares a variation with the original over the same period. Practical guardrails:
- Change one meaningful thing at a time, or accept that you will not know which change worked.
- Decide the success measure and the minimum run time before starting, and do not stop the moment a result looks good.
- Make sure the variation works properly. NN/g notes a sound concept can fail because it was executed poorly.
- Check results by device, since a gain on desktop can hide a loss on mobile.
Small sites often lack the traffic for firm results. In that case, rely on before-and-after comparisons, research and common sense, and label the evidence honestly.
Step 7: Record and repeat
Keep a log: hypothesis, change, dates, result, decision. Over time it becomes an asset that stops the team re-testing old ideas. Landing pages for campaigns are a good place to apply the process, as covered in landing pages that turn campaign traffic into enquiries.
Common mistakes
- Copying another business’s winning test, without knowing why it won.
- Optimising a step far from the real problem.
- Ignoring page speed and accessibility.
- Running many tests at once on the same page.
- Judging by conversion rate alone when lead quality matters.
Where Radiance Impact fits
Radiance Impact’s Growth and Technology divisions work on landing pages, websites and measurement, which are where CRO usually starts. To discuss a project, contact us or see Growth.
Sources
- A/B Testing and UX Research, Nielsen Norman Group
- Key events in Google Analytics 4, Analytics Help (Google)
- Web Vitals, web.dev (Google)
