Landing Page Optimization

You already paid for the traffic. Stop letting it leak out of the funnel.

Most landing pages bleed conversions at a step no one is watching — a form field that kills mobile signups, a deploy that quietly broke the CTA, a headline that confuses more people than it converts. The data to find it is sitting in PostHog or Mixpanel right now. Nobody has the hours to dig.

This worker digs. Every morning it scans your funnels, ranks the leaks by how many conversions you could actually recover, and tells you whether a problem is chronic or something that just broke. It watches the session recordings of the people who didn't convert and comes back with the specific friction — not "the page underperforms," but "47% drop at the form step, mobile, abandoning at the phone field in 18 of 30 sampled sessions."

Then it turns that evidence into tests worth running: one falsifiable hypothesis per change, the win defined before the test starts, sized so you know it'll reach significance, ranked cheapest-impact-first. When a deploy ships, it checks Vercel and confirms with real sessions whether the change helped or hurt. And when a test wins, it doesn't die in a slide — it files the implementation ticket in Jira, Linear, or GitHub with the exact variant to ship and the lift to expect.

It's disciplined where it counts. It won't call a winner before the math is there, it rejects a "win" that breaks a guardrail metric, and below your minimum sample size it stays quiet instead of inventing a finding. It works the conversion path so you can work everything else.