Health questions before a single location
A Fortune 100 company’s clinical research team was losing most of its trial candidates partway through the pre-screener. Across four previous campaigns, abandonment averaged 65%, even with targeted, high-intent traffic. In every one of those campaigns, people couldn’t see where the trial sites were until they’d finished a medical screener. That meant answering sensitive health questions before knowing whether taking part was even realistic.
Constraint: six weeks to fall enrollment.
Three decisions behind the new flow
Show them where, up front
Reviewing the prior campaigns’ data, my theory was that people won’t hand over health information, or even spend the time, if they’re not sure a test site is within reach. I inverted the standard industry flow by putting a location directory first, so people could confirm the study was geographically feasible before facing the medical screener. It was a low-risk bet: if I was wrong, switching back to the old flow would take minimal changes.
Cities, not contact information
Operations pushed back hard at first. In the past, publishing facility details had turned test sites into makeshift support desks, with phones ringing all day from people who hadn’t even been screened. My recommendation was a directory of cities with active sites, keeping every facility unnamed until the final consent step. Users got the one thing they needed to decide, and the sites didn’t get the calls. Test sites reported no increase in premature contacts.
Three steps in plain language
I streamlined the flow into three steps (city directory → branching screener → consent and facility match), replaced clinical jargon with everyday labels, and added plain-language privacy notes like “We don’t store your information until you consent.” The branching screener changes which questions it asks based on earlier answers. I built the flow and brought in an engineer to pair with me on the clinical screener logic; with health data and a client this size, a second set of eyes felt like the right move.
Results
| 65% → 37% | 35.3% | 2 more studies |
|---|---|---|
| Screener abandonment | End-to-end conversion, up from ~13.3% | Have used the flow as their template |
| Metric | Before: avg of 4 prior campaigns | After: first 30 days (1,086 sessions) | Change |
|---|---|---|---|
| Bounce rate | 62% | 44% | −18 pts |
| Screener starts | 38% of sessions | 56% (608) | +18 pts |
| Screener abandonment | 65% | 37% (225) | −28 pts |
| Screener completion | 35% of starts | 63% (383) | +28 pts |
| End-to-end conversion | ~13.3% of sessions | 35.3% (383 / 1,086) | +22 pts |
| Average time on task (screener) | 4:20 | 2:49 | −1:31 |
Sources: Google Analytics, Mixpanel, form telemetry.
At its peak, the study ran across 94 facilities nationwide, and about a dozen hit their enrollment quotas early. The client has since used the flow as the template for two more studies.
Takeaways
- The order was the problem. Following an industry standard to the letter wouldn’t have fixed this.
- Find the minimum that answers the user’s question. When operations is worried about exposure, look for the least you can show: here, cities instead of contacts.
- Make the case airtight. These results compare against past campaigns, not a controlled test, so with more time I’d A/B test where the location step sits.