Conversion Rate Optimisation in London

Honestly depends on starting point and testing rigour. Sites with significant existing CRO gaps (poor landing pages, broken funnels, fundamental UX issues) can see 30-100%+ improvements over 6-12 months of structured testing. Well-optimised sites with mature programmes typically see 5-15% annual improvements from incremental testing — still highly valuable when compounded across all traffic sources. We agree realistic targets during discovery based on audit findings rather than promising specific percentage improvements upfront.

Statistical significance requires sufficient sample size — typically you want 1,000+ conversions per test variant to reliably detect meaningful improvements. For sites with low conversion volume, testing major changes with bigger expected effect sizes can still work; testing small changes requires much more traffic. Below ~10,000 monthly conversion-relevant visitors, A/B testing becomes slow and we typically focus on bigger redesign-driven changes plus qualitative research rather than continuous A/B testing programmes.

Critical distinction. Redesigning the page produces a change but doesn't isolate what specifically caused the improvement (or regression). A/B testing splits traffic between control and treatment so you can attribute changes specifically to the treatment. Most CRO 'wins' reported by agencies that don't A/B test are actually a mix of real improvements, seasonal variation, and confirmation bias. Without proper testing, teams systematically overestimate impact while spending budget on changes that don't actually matter.

Depends on the engagement. VWO and Convert are our default A/B testing platforms — both offer robust testing capabilities at reasonable price points. Server-side testing (Optimizely, LaunchDarkly) for engineering-heavy implementations. Platform-native A/B testing (Shopify's Theme Studio, Webflow native testing) where it covers the requirements. For research and analysis: Hotjar, Microsoft Clarity (free, excellent), FullStory, plus GA4 for quantitative funnel data.

All three. eCommerce CRO focuses on product page conversion, cart and checkout funnels, and post-purchase retention paths. SaaS CRO focuses on signup conversion, trial-to-paid activation, and self-service expansion. B2B lead gen CRO focuses on landing page conversion, form optimisation, and the prospect-to-MQL handoff. The methodology is the same across all three — research, hypothesis, test, document — but the specific opportunities differ. We are explicit about which conversion goals matter to your business during discovery.

Yes — and sometimes redesign is the right answer when research shows the page has fundamental issues that micro-tests won't fix. We coordinate with our bespoke website design and e-commerce development teams for full redesign builds, then run A/B testing programmes on top of the new design to compound the initial improvement over time.

Test prioritisation using a framework like PIE (Potential impact, Importance for business, Ease of implementation) or ICE (Impact, Confidence, Effort). Each potential test gets scored against these dimensions and ranked. We focus testing on high-traffic high-leverage flows (checkout, primary landing pages, signup) rather than scattered cosmetic changes that won't move the needle even if they win.

A losing test is one where the treatment variant performed measurably worse than control. They're typically 30-50% of tests in mature programmes and they're not failures — they're informative. We document the hypothesis, what we learned about why it underperformed, and use it to inform future testing. Programmes that report only winners are either testing only obvious changes, peeking at results, or only running for short durations — none of which produce reliable learnings.

Yes. We audit existing testing setup, historical test documentation, current funnel performance, and current tooling. Then propose either targeted improvement (where the foundations are sound) or structural restart (where testing rigour has been absent). Common findings include tests run without proper sample sizes, peeking and early stopping, lack of test documentation so learnings haven't compounded, and prioritisation focused on easy changes rather than high-impact ones.