Conversion rate optimisation: where to start

CRO is a process, not a list of tweaks

Conversion rate optimisation is the work of getting more of your existing visitors to take the action you want, whether that is buying, booking or sending an enquiry. It is tempting to start with button colours and headline swaps. The programmes that work start somewhere less exciting: accurate measurement, then research into what is actually stopping people. Here is the order to work in.

Step 1: fix measurement first

If your conversion data is wrong, every decision built on it is wrong too. Before you change anything:

• Define the primary conversion. A purchase, a qualified enquiry or a booking, not a page view.
• Define the steps before it. Product view, add to cart, checkout start, form start. These micro conversions show where people fall out.
• Check that events fire once, and only once. Duplicate or missing events are common after site changes.
• Reconcile with your back end. Compare analytics conversions with orders in your store or leads in your CRM. Some gap is normal; a large or changing gap needs fixing.

In GA4, a funnel exploration lets you lay out up to 10 steps and see how many users move through each one, broken down by device, channel or other dimensions.

Step 2: find where people drop off

Use your analytics to locate the problem before you try to explain it. Look at:

• Landing pages with high traffic and low conversion.
• The funnel step with the biggest fall in users.
• Differences between mobile and desktop. A large gap often points to a usability problem on one of them.
• Differences by traffic source. Poor conversion from one channel may be a targeting issue rather than a site issue.

The goal is a short list of pages and steps where improvement would matter most.

Step 3: find out why

Analytics tells you where. Qualitative research tells you why. Use several sources, because each one shows a different part of the picture.

• Heatmaps show where people click and how far they scroll.
• Session recordings show real visitors hesitating, backtracking or hitting errors. Microsoft Clarity offers both heatmaps and recordings free, which makes it an easy place to start.
• On-site surveys ask visitors a single question at the right moment, such as what nearly stopped them buying.
• Customer feedback from support tickets, sales calls, reviews and returns reveals objections and confusion in customers' own words.
• Usability testing, watching a handful of people from your audience try to complete a task, often surfaces the obvious problems everyone inside the business has stopped noticing.

Step 4: write proper hypotheses

Turn each finding into a hypothesis you can test or act on. A useful format:

Because we saw [evidence], we believe that [change] for [audience] will cause [outcome]. We will know by measuring [metric].

For example: because recordings show mobile users repeatedly tapping the delivery information link at checkout, we believe showing delivery costs on the product page will reduce checkout abandonment, measured by checkout completion rate on mobile.

A hypothesis tied to evidence teaches you something whether it wins or loses. A random idea does not.

Step 5: prioritise

You will have more ideas than capacity. Score each one on a few simple factors and work from the top:

• Impact: how much traffic the page gets and how close it sits to the conversion.
• Confidence: how strong the supporting evidence is.
• Effort: how much design, development and approval it needs.

Frameworks such as ICE (impact, confidence, ease) formalise this. The exact scoring matters less than being consistent.

Step 6: test properly

An A/B test splits traffic between the current version and a variation, then compares results. The basics that keep tests honest:

• Pick one primary metric before the test starts.
• Calculate the sample size in advance with a tool such as Evan Miller's sample size calculator. It depends on your current conversion rate and the smallest lift you want to detect. Halving the lift you want to detect roughly quadruples the visitors you need, so small improvements on low-converting pages can need tens of thousands of visitors per variation.
• Do not stop early. Checking results repeatedly and stopping when they look significant sharply raises the chance of a false winner. Run to the planned sample size.
• Run in full weeks so weekday and weekend behaviour are both represented.
• Plan any segment analysis upfront. Slicing results by device or channel after the fact will almost always turn up a "winner" somewhere by chance.
• Record every result, including losses and flat tests. A shared log stops the same idea being tested twice and builds a picture of what your customers respond to.

Google Optimise was shut down on 30 September 2023, so you will need a dedicated testing tool or a testing feature in your site platform.

When not to A/B test

• When traffic is too low to reach a reliable result in a reasonable time.
• When something is simply broken. A failing form or a checkout error should be fixed, not tested.
• When the change is required anyway, such as legal, compliance or accessibility updates.
• During unusual periods like major sales events, when visitor behaviour does not reflect the rest of the year.
• When you will not act on the result. If the decision is already made, a test only adds delay.

What to do with lower traffic

Low traffic does not rule out CRO. It changes the method. Lean harder on research and usability testing, ship well-evidenced fixes and monitor before-and-after results with care for seasonality. When you do test, test bolder changes, since larger effects need fewer visitors to detect, and consider measuring an earlier step such as add to cart, which happens more often than a purchase.

Sources

• Google Analytics Help - Funnel exploration
• Microsoft Learn - About Microsoft Clarity
• Evan Miller - Sample size calculator
• Evan Miller - How not to run an A/B test
• Google Optimise Help - Optimise sunset