Insights / Guide / Conversion rate optimisation (CRO)
GuideConversion rate optimisation (CRO): the complete guide
More orders from the same traffic. How to read your funnel, fix checkout and search, and run tests that produce evidence rather than opinion.
Winning a visitor costs money, through advertising, search, content or partnerships. Yet on a typical commerce site only a small share of visitors buy. The rest leave because they cannot find what they want, are surprised by costs, meet a confusing checkout or simply lose patience with a slow page. Conversion rate optimisation (CRO) is the disciplined work of finding and fixing those reasons.
Done well, CRO produces more orders from the same traffic, lowers the cost of acquiring each customer and replaces opinion with evidence. This guide covers how to analyse the funnel, understand behaviour, fix the most common leaks and run experiments you can trust.
Read your funnel
Start with the numbers. Define the steps a visitor passes through, from landing page to product page, basket, checkout steps and order confirmation, and count how many reach each one. The step with the biggest drop is usually where the greatest gains lie.
- Count each step: visits, product views, basket additions, checkout starts, each checkout stage and completed orders.
- Find the biggest drop, both in percentage and in the number of lost visitors.
- Segment the funnel by device, traffic source, new or returning visitor and region.
- Compare with the previous period to separate long-standing problems from new ones.
Segmentation matters because averages hide problems. Mobile visitors behave differently from desktop visitors: they browse more, buy less per visit and are far more sensitive to slow pages and long forms. A checkout that works well on a laptop may lose half its mobile users at the address step. Paid search visitors may convert well on one landing page and poorly on another.
Check that the numbers are right before acting on them. Tracking errors are common: a checkout step that is not recorded, duplicate order events or internal staff visits included in the data. A quick comparison between orders in the analytics tool and orders in the back office will reveal most problems. If the two differ by more than a few per cent, fix the tracking first.
Understand behaviour
Numbers show where visitors leave; behaviour shows why. Session recordings, heatmaps, on-site surveys and simple usability tests with five or six real customers reveal the confusion behind the drop: a delivery cost that appears only at the last step, a button below the fold on mobile, an error message that does not explain what went wrong.
- Watch recordings of sessions that reached checkout but did not buy.
- Ask a single question on exit or after purchase: what nearly stopped you, or what is missing?
- Review customer service contacts about ordering problems.
- Run short moderated tests where customers complete a typical purchase while thinking aloud.
Record each finding as a hypothesis: ‘Visitors abandon at delivery because they first see the cost there; showing it on the product page will reduce abandonment.’ Hypotheses make tests focused and results easier to interpret.
Prioritise what to fix
Research usually produces a long list of possible improvements. Rank them by the number of visitors affected, the likely size of the effect and the effort required. A fix on the checkout page affects every buyer; a change to a minor category page affects a few. Work on high-impact, low-effort items first, and keep the list visible so the team and stakeholders can see what is planned and why.
Fix the checkout
Checkout is usually the biggest leak, and many fixes are well established. Customers drop out when costs appear late, when they are forced to create an account, when forms are long or confusing and when their preferred payment method is missing.
- Show full cost, including delivery and taxes, as early as possible.
- Offer guest checkout, with an option to save details after the order.
- Reduce form fields to those you genuinely need, and use address lookup.
- Offer the payment methods your customers expect, including digital wallets.
- Show clear delivery options and dates.
- Keep the basket visible and editable throughout.
- Write error messages that say exactly what to fix.
- Show security and returns information near the payment step.
- Make every step work well on a small screen.
- Send a helpful reminder to logged-in customers who abandon their basket, where they have agreed to contact.
Consider a typical home goods retailer whose delivery charge appeared only on the final step. Showing an estimate on product pages and a free-delivery threshold in the basket reduced checkout abandonment noticeably and raised average order value, because some customers added items to reach the threshold.
Beyond the checkout
Product pages deserve the same attention. Clear images from several angles, honest descriptions, sizes and dimensions, delivery dates, stock status and reviews answer the questions that otherwise send visitors elsewhere. For higher-value items, comparison tables, finance options and an easy way to ask a question can make the difference. Each element is a candidate for research and testing.
Fix search, navigation and speed
Visitors who use site search often convert at a much higher rate, because they know what they want. When search fails them, they leave. Handle misspellings, synonyms and plurals, so ‘sofa’, ‘sofas’ and ‘couch’ return the same products. Show useful filters, such as size, price and availability. Review searches that return no results every month and fix them with synonyms, redirects or new products.
Navigation should reflect how customers think, not how the organisation is structured. Test category names with customers and keep menus shallow. Speed matters everywhere: slow pages lose visitors before they see anything. Compress images, limit third-party scripts and measure page speed on real mobile connections, not only on office broadband.
Test without fooling yourself
Structured experiments, such as A/B tests, show whether a change works. They also make it easy to fool yourself. Small samples mislead: a test that runs for three days with a few hundred visitors can show a large difference that disappears with more data.
- Plan the sample size before starting, based on current conversion and the smallest improvement worth detecting.
- Run tests for whole weeks, to include weekday and weekend behaviour.
- Change one thing at a time, or use a design that can separate the effects.
- Measure orders and revenue, not clicks or time on page.
- Record every test, including those that failed, so learning accumulates.
Be careful with tests during unusual periods, such as sales events, holidays or major campaigns. Visitor behaviour changes, and a result found then may not hold the rest of the year. Where you must test during such periods, note the context and confirm important results with a follow-up test later.
Not everything needs a test. Clear usability fixes, such as broken forms or missing payment methods, can simply be made. Reserve testing for changes where the outcome is genuinely uncertain and the traffic is sufficient to give a clear answer.
What you will achieve
A steady CRO programme delivers more orders from the same traffic, a lower cost per customer, better decisions based on evidence and a habit of learning that compounds over time. Each improvement makes every marketing pound work harder.
Make CRO a routine rather than a project. A monthly cycle of reviewing the funnel, adding new research findings, choosing the next fixes or tests and reporting results keeps progress steady. Share results beyond the digital team; buying, merchandising and customer service often hold the knowledge that explains why customers behave as they do.
Start by building the funnel report, segmented by device. Watch twenty recordings of abandoned checkouts. Between them, they will usually point to three or four fixes worth making this month.
Source: Intelliversal analysis, drawing on the briefings in this series.
