Northwind — CRO — A funnel rebuild that nearly tripled conversion

1. We started by watching, not guessing

Before touching the funnel, we watched 50 real session recordings and layered them against GA4 and Hotjar. Three problems showed up in almost every drop-off: unclear promise at the top, a form that asked for too much too early, and a pricing page that read like an internal spec.

Every change after that was tied to one of those three problems. Nothing shipped because it looked nicer — it shipped because it moved a number we agreed on up front.

2. One promise, one CTA, one job per page

We rewrote the hero around a single, testable outcome instead of a feature list, cut the primary CTA to one action, and moved the strongest customer proof directly under the fold. Bounce rate on the homepage dropped by nearly a quarter in the first two weeks.

• Hero rewritten around the outcome, not the feature set.
• Signup form cut from seven fields to three.
• Pricing page reframed around who each plan is for.
• Sticky CTA on mobile, always one thumb away.

3. Testing what actually mattered

We ran three sequential A/B tests — hero copy, pricing layout, signup form — instead of shipping everything at once. Each test locked in a real, attributable lift before we moved to the next, so by week six we knew which changes were driving what.

4. 1.9% to 5.6% — same ads, same audience

Blended conversion moved from 1.9% to 5.6% in eight weeks, revenue per session climbed 3.9x, and the team stopped needing to raise ad spend to hit their number. The traffic didn't get better. The funnel finally respected it.