Conversion Rate Optimisation: Analyse Funnels, Test Changes and Measure Results

Introduction

Conversion rate optimisation is the practice of finding out why visitors don’t act on your website, then testing changes that help more of them do so. Plenty of sites get steady traffic and still lose most of it. Sometimes the cause is a form that fails on certain phones. Sometimes it’s a checkout step nobody has questioned in years. And sometimes the tracking is simply wrong, so the team spends its time fixing problems that don’t exist.

A sound optimisation process follows a clear order. You confirm the data can be trusted, study the funnel, investigate where people struggle, rank the opportunities, run tests that suit your traffic and judge the outcome by what it did for the business. This guide walks through each step with the technical detail you’ll need to apply it.

Establish Conversion Goals and Reliable Measurement

Define Primary and Secondary Conversion Actions

Start by deciding what counts as success. On a lead-generation site, the primary action is usually a submitted enquiry or a booked call. For an online shop, it’s a completed purchase. Supporting actions such as a brochure download, a product view or a newsletter sign-up matter too, but they sit lower down. If every click is called a conversion, the rate looks healthy and tells you very little about revenue.

Calculate a Consistent Baseline Conversion Rate

The basic formula is conversions divided by visitors, multiplied by 100. The catch is what you count as a visitor. A user-based rate divides by people, while a session-based rate divides by visits, so one person returning three times counts three times. Neither is wrong. Pick one, write it down and use it every time. Without a steady baseline, you can’t tell a real improvement from a change in how the number was worked out.

Validate GA4 Events and Tracking Implementation

Before analysing anything, check that your conversion tracking records what actually happens. GA4 calls these key events, which is what it used to label conversions. Open GA4’s DebugView and complete a test enquiry or purchase to watch each event arrive in real time. Use Google Tag Manager’s Preview mode to confirm which tags fire, and when. You’re looking for events that never fire, events that fire twice and thank-you pages that count a conversion every time someone refreshes them.

Then compare GA4’s totals with your CRM or order system. A small gap can be normal. A large one usually points to a fault in your GA4 conversion tracking, and it needs fixing before any test begins.

Map the Conversion Funnel and Diagnose Friction

Build and Interpret a Step-Based Funnel in GA4

Funnel exploration in GA4 lets you define the steps of a journey, such as viewing a service page, starting a form and submitting it. The report shows how many users complete each step and where they leave. You can build an open funnel or a closed one. In an open funnel, people can join at any step. In a closed funnel, they only count if they enter at the first. That difference changes the numbers, so choose deliberately. Conversion funnel analysis works best when every step is something a person can actually see and do, not a vague stage like “consideration”.

Segment Drop-Offs by Device, Traffic Source and Landing Page

An overall completion rate can hide the real problem. Break the funnel down by device and you may find desktop users finish comfortably while mobile users stall at the same step. Compare traffic sources too, since visitors from paid ads often behave differently from those arriving through search. Landing page conversion optimisation usually starts here. Find the entry pages where people drop out fastest and look at those first, rather than rewriting the whole site.

Combine Analytics With User Research and Usability Testing

A drop-off tells you where to look. It doesn’t explain why. Session recordings can show hesitation and repeated taps, provided your consent setup allows them. Customer emails and sales call notes often reveal confusion the data can’t. Even five short usability sessions can expose a muddled sentence or a hidden button.

Treat each source as a clue, and look for the same issue turning up in more than one. A single recording is an anecdote. Three sources pointing the same way is evidence, and that’s what should drive website user experience decisions.

Check Mobile Forms, Accessibility and Page Performance

Try your forms on a real phone. Look for fields that bring up the wrong keyboard, error messages that vanish and buttons too small to tap. Test navigation with a keyboard and a screen reader as well, since inaccessible interactions block real customers.

For speed, Core Web Vitals give useful signals. As of 2026, Google’s good thresholds are an LCP of 2.5 seconds or less, an INP of 200 milliseconds or less and a CLS of 0.1 or less, measured at the 75th percentile. Treat them as clues, not automatic explanations.

Prioritise CRO Opportunities and Develop Testable Hypotheses

Rank Issues by Impact, Evidence, Effort and Risk

Once the research is done, you’ll have more ideas than you can test. A simple scoring sheet helps. Rate each issue on likely commercial value, strength of evidence, number of users affected, build effort, technical risk and how reliably you could measure the result. A one-to-five scale is enough.

Treat the total as a prompt for discussion, not a prediction, because the scores are still judgement calls. A fix that affects most mobile visitors and has evidence from two sources should outrank a bold redesign backed by a hunch. That ordering sits at the core of any sensible CRO strategy.

Turn Findings Into Testable Hypotheses

A hypothesis links an observation to a change and a measurable result. “Let’s try a shorter form” is weak. Try this instead: “Funnel data and recordings show mobile users abandoning our enquiry form at the phone number field. If we make that field optional, completion among mobile visitors should rise, measured by submitted enquiries per mobile session.”

Notice what it contains: the audience, the evidence, the change, the expected outcome and the success metric. A good hypothesis is specific enough to be wrong. That makes even a losing test useful, because you’ll know which assumption failed.

Design and Run Reliable CRO Experiments

Choose Between Controlled Experiments and Direct Fixes

Not everything deserves an experiment. If a form breaks in one browser, fix it. Nobody needs a test to prove that a working form beats a broken one. Save A/B testing for choices where you honestly don’t know which version will perform better, such as page layouts, offers or headline approaches. Testing obvious defects only delays the repair and uses traffic you could spend on real questions.

Select a Primary Metric and Guardrail Metrics

Pick one primary metric before the test starts, such as submitted enquiries per session. Then add guardrail metrics that warn you about side effects. Suppose a shorter form lifts submissions, but the sales team reports more unqualified enquiries. Without a guardrail like qualified-lead rate, you’d call the test a win and move on.

Plan Sample Size, Minimum Detectable Effect and Stopping Rules

Sample size depends on four inputs: your baseline conversion rate, the smallest improvement worth detecting, the significance level and the statistical power. Teams commonly use 95% significance and 80% power, though those are conventions rather than laws.

The smaller the effect you want to detect, the more visitors you need. A page converting at 2% needs far more traffic to detect a 5% relative lift than a 30% one. Use a sample-size calculator before launch, and write down your stopping rule, such as a fixed number of visitors per variant across full weekly cycles. There’s no universal duration. Decide it in advance, then leave the test alone until it ends.

Avoid Common Experiment Design Errors

Stopping early is the classic mistake. Results swing wildly in the first few days, and a variant that looks brilliant on Tuesday often fades by the following week. Other errors are quieter. Teams run overlapping tests on the same page, change the design halfway through, split traffic unevenly by accident or ignore an outside event such as a sale or a press mention. Any of these can leave you with a result that looks decisive and means nothing.

Interpret Results and Measure Commercial Impact

Calculate Conversion-Rate Change and Relative Uplift

Here is a hypothetical example, using made-up figures for illustration. A page receives 10,000 visits and produces 200 conversions, a baseline rate of 2%. After a change, the rate becomes 2.5%. That is a 0.5 percentage-point increase. At the same traffic, you’d see 250 conversions, which is 50 more. Relative to the baseline, the uplift is 25%. People often report a 0.5% gain when they mean 0.5 points. State which one you mean every time.

Interpret Statistical Uncertainty and Test Limitations

The arithmetic above proves nothing about cause. A 25% uplift in a sample could come from chance, a seasonal bump or a different mix of visitors. Statistical tests estimate how likely a difference is to be noise, but they don’t remove the doubt entirely.

If a test ends inconclusive, say so. It still teaches you something, namely that the change probably wasn’t large enough to matter. Presenting it as a quiet win damages trust in the whole programme.

Assess Qualified Leads, Revenue and Profitability

A higher conversion rate isn’t automatically a better result. For a lead-generation business, count qualified leads and, where your CRM allows it, closed deals. For an online shop, look at revenue per visitor and average order value, because a discount that boosts orders can still shrink profit.

Consider acquisition cost as well. If a change lifts conversions but pulls in lower-value customers, the headline number flatters you. Tie every test back to the measure the business cares about, then treat conversion rate as supporting detail.

Validate Changes After Implementation

A winning variant still needs checking once it’s live. Re-run your DebugView and Tag Manager checks to confirm events fire on the permanent version, not just the test build. Then monitor the primary and guardrail metrics for a few weeks, since some effects fade once the novelty wears off. If results drift back toward the old baseline, you’ll want to know before you build the next idea on top of it.

Conversion Rate Optimisation Process

Maintain a Prioritised CRO Backlog

Keep one shared list of issues. For each, record the problem, the evidence, its priority score, the proposed fix and its status. The point is that ideas don’t live in people’s heads or scattered emails, and a new team member can see why something ranks where it does.

Document Experiment Decisions and Outcomes

Log every test, including the failed and inconclusive ones. Note the hypothesis, dates, result, limitations and the decision you made. Teams that skip this tend to repeat old experiments a year later, with the same disappointing outcome. A short entry takes five minutes and saves far longer.

Review Performance and Reassess Assumptions

Set a regular review, monthly or quarterly depending on traffic. Check whether earlier assumptions still hold, whether new data changes your priorities and whether business goals have shifted. A plan that never gets revisited slowly drifts away from what the business needs.

Conclusion

Conversion rate optimisation works when each step builds on the one before. Check that your measurement is sound, find where visitors struggle, rank what you’ve found, test where traffic allows and judge results by revenue and lead quality, not just a rate. Start with the tracking audit, because everything else depends on it. If you’d like experienced help with any stage, Midland Marketing’s conversion rate optimisation service covers research, testing and reporting.

Frequently Asked Questions

What is a good website conversion rate?

It depends. Industry, how you define a conversion, traffic source and business model all shift the figure. Compare your rate against your own past performance and similar sites, rather than a single headline number.

How long does conversion rate optimisation take?

There’s no fixed timeline. It depends on how much research is needed, how quickly changes can be built, your traffic levels and how long reliable measurement takes. Direct fixes can pay off quickly, while tests on modest traffic can run far longer.

Can a low-traffic website benefit from CRO?

Yes. Without enough visitors for a reliable A/B test, rely on analytics, usability sessions, customer feedback and direct fixes to obvious problems. These methods need far less traffic and often uncover the biggest issues.

What should a business test first when website conversions are low?

Start by checking your analytics and tracking. Then look for the pages or steps where visitors leave most often. Review mobile forms, page clarity, checkout steps and calls to action. Use customer feedback and usability tests to understand the cause before deciding what to change.

How can you tell whether a website change has improved conversions?

Compare the results against your original conversion rate using the same measurement method. For an A/B test, choose a primary metric and set the sample size and stopping rules before launch. Check lead quality or revenue as well as conversion numbers. This helps you judge whether the change delivered a meaningful business benefit.

Written by - Lauren Davison

Introducing Lauren – one of our content writers who has a flair for SEO and creative strategy!

With a Master’s Degree in Creative Writing, Lauren has niched down into SEO and content writing.

Outside of work, she loves watching the darts, reading and the pub on the weekend.