Conversion Rate Optimization: Why Redesigning Your Website Won't Fix Low Conversions (And What Will)

A Canadian e-commerce client came to us right after a full website redesign — new visual identity, new photography, a sleeker layout — expecting conversion rate to climb. It didn't move. The redesign had fixed everything the team could see was 'wrong' visually, but nobody had actually watched a session recording of a real customer trying to check out. Once we did, the issue was obvious within the first ten sessions: the shipping cost only appeared after entering a full address on the second-to-last checkout step, and a large share of visitors abandoned right there. No amount of better photography fixes a pricing surprise at the worst possible moment. This is the pattern behind most CRO failures: teams optimize what they can see (the look) instead of what they can measure (the actual point of abandonment).

Finding the Real Friction Points Before Touching a Single Pixel

A proper friction audit starts with funnel analysis in analytics to locate exactly which step loses the most visitors — landing page to product page, product page to cart, cart to checkout, checkout to confirmation — because 'low conversion rate' is not a diagnosis, it is a symptom with a specific location. From there, session recordings and heatmaps reveal what the numbers alone can't: rage clicks on elements that look clickable but aren't, users scrolling straight past a value proposition because it's buried below a hero image, or repeated attempts to fix a form field that's silently rejecting valid input. Mobile and desktop should be reviewed separately, since friction points rarely match — a form that works fine on desktop can be nearly unusable on mobile if fields aren't sized for touch input. In bilingual markets such as Canada, this audit needs to be run separately by language too, since a translated checkout flow inherits whatever bugs or unclear copy existed in the version it was copied from, often unnoticed for months.

Once friction points are identified, they should be turned into hypotheses, not opinions — 'we believe removing the phone number field from checkout will reduce abandonment because session recordings show hesitation there' is testable; 'the button should be blue' is not. Prioritize the resulting list with a simple impact/confidence/ease score so the team tests the changes most likely to move the needle first rather than whatever the most senior person in the room prefers. Run tests with enough traffic and time to reach statistical significance — a common and costly mistake is calling a test after three days or a few hundred visitors, which produces noise dressed up as a result. Not every test needs to target the final purchase: micro-conversions (added to cart, email captured, product configurator completed) are valid intermediate signals, especially for lower-traffic B2B or high-consideration purchases where the final conversion event is too rare to reach significance quickly.

• Map the full funnel in analytics and identify the single step with the highest drop-off rate.
• Watch at least 20-30 session recordings at that step before forming any hypothesis.
• Score every test idea on impact, confidence and ease before prioritizing the roadmap.
• Run tests to statistical significance, not to a fixed calendar deadline.
• Audit mobile and desktop, and each language version, separately.
• Track micro-conversions alongside the final conversion goal for low-traffic or high-consideration funnels.