What is CRO (Conversion Rate Optimization)

How it works

The core cycle keeps decisions grounded in evidence rather than opinion or the loudest voice in the room.

• Research: analytics, session data and user feedback
• Hypothesise: a specific, testable change and expected effect
• Test: an A/B or multivariate experiment with enough sample
• Learn: implement winners, document losers, repeat

Common mistakes

Running tests without enough traffic to reach significance, calling winners too early, changing several things at once so you cannot attribute the result, or optimising a single metric while harming downstream ones like retention or refunds.

Why it matters

Conversion rate optimisation is the discipline of systematically increasing the share of visitors who take a desired action, and its appeal is pure leverage: it grows revenue from traffic you already have, so its gains compound on top of every acquisition channel at once. A well-run CRO program often returns more than an equivalent investment in more traffic, because it costs less to convert existing visitors than to acquire new ones — and unlike a traffic spike, a conversion-rate gain is permanent and lifts the ROI of SEO, paid, email and everything else simultaneously.

• Systematically lifting the conversion rate of traffic you already have
• Compounds on top of every acquisition channel at once
• Usually cheaper per incremental customer than buying more traffic

How it is done properly

Real CRO is a research-and-experimentation loop, not a bag of "best practice" tweaks. It starts with research — analytics, heatmaps, session recordings, surveys and user testing — to find where and why visitors drop, forms prioritised hypotheses grounded in that evidence, and validates them with statistically-rigorous A/B tests so a "winner" is genuinely a winner and not noise. The two failure modes are copying someone else's "winning" layout (which ignores your specific users) and calling tests too early on too little traffic (which produces false positives). Done right, it builds a compounding library of validated learnings about your users.

• A loop: research → prioritised hypotheses → rigorous A/B test → iterate
• Research-led, not copied "best practices" that ignore your users
• Statistically-valid tests — under-powered tests produce false wins
• Each cycle compounds into a growing understanding of your users

Worked example

A team is tempted to copy a competitor's checkout layout that 'converted well'. Instead they run the proper loop. Research first: analytics shows heavy drop-off on the shipping step, and session recordings show users hesitating at an unexpected shipping cost revealed only at the end. That evidence forms a specific hypothesis — surfacing shipping cost earlier will reduce abandonment — which they test with a statistically-powered A/B test rather than shipping it on faith. The variant wins with significance, lifts completed checkouts, and the learning ('cost transparency reduces late-stage abandonment') feeds the next test. Copying the competitor would have addressed a problem their own users did not have. The example captures why real CRO is a research-and-experimentation discipline: the wins come from fixing your users' actual friction, validated by rigorous tests, not from importing someone else's 'best practice'.