How looking at the data differently — screen by screen instead of milestone by milestone — uncovered the drop-off no one knew about and improved end-to-end conversion by 25%.
Patients who completed onboarding and connected with their care team loved the platform. The problem was that too many weren't making it through. Onboarding completion felt stalled — but no one had a clear picture of where patients were actually dropping off.
The team was tracking milestone events: profile created, HCP connected, appointment booked. Those milestones told a story, but it was an incomplete one. They showed that patients were dropping off, not where or why.
I pushed the engineering and data teams to instrument every screen in the onboarding flow — not just the milestone events, but every tap, every button, every point where a patient could stop. That analysis hadn't been done before. The team was skeptical it would reveal anything the milestone data hadn't already shown.
It revealed three things.
Three changes: update the copy on the first screen, make the button look like a button, and cut the introduction down. No flow restructure, no new screens, no engineering complexity. The diagnosis told us exactly where the problem was — which meant we didn't need to guess at a solution.
The restraint was intentional. A small, targeted change meant we'd have a clean signal. If the funnel improved, we'd know why.
The change shipped in May 2025. In the months that followed, every downstream stage improved. The fix at the top of the funnel compounded all the way to appointment booked.
The real work here happened before I drew anything. Advocating for a different way of looking at the data, convincing engineering and product to instrument screens they'd never tracked, and holding the thread on what we were actually trying to find out — that's what made the design fix possible.
