Turning an A/B Test into a Permanent Lead-Gen Funnel
Dayan Ramly Ramadhan
Overview
Role: Frontend Developer, Experimentation. Stack: Optimizely, Nuxt/Vue, Google Tag Manager, GA4.
A comparison brand running several insurance verticals depended on one thing above all, getting users from landing page to completed lead. I owned the experimentation layer, from building variants and tracking to making sure the data could be trusted across multiple domains.
The Problem
The existing funnel showed several questions per page. It was functional, but it asked a lot of users at once, and there was no reliable way to know which interactions caused drop-off. Two further issues made testing hard.
Users moved between domains within a single funnel, so experiment data could fall out of sync.
Tracking stopped at page-level events, which hid what happened inside each step.
The Approach
1. Hypothesis, one question at a time
I proposed changing the funnel from multiple questions per page to a single question per page, aiming for less cognitive load and higher completion.
2. Build and run the experiment
I implemented the variants in Optimizely, with audience targeting driven by query params, storage, and other signals depending on each test's needs.
3. Consistent data across domains
I made sure traffic splitting stayed accurate and consistent when users crossed domains, so results reflected real behavior instead of split or duplicated sessions.
4. Granular tracking
Working with the data team, I set up interaction-level events, such as dropdown selections and input completion, so we could see exactly where users hesitated, not just where they left.
5. Collaboration
QA. I provided a stable control experience for regression testing and explained how to read results.
Design and UX. I advised on what could be implemented within the existing system and codebase, and what needed adjusting.
Data. We aligned event definitions so analysis stayed consistent.
The Results
10% lead uplift on the latest experiment.
Single-question-per-page consistently outperformed the original across verticals.
Promoted from experiment to the permanent default funnel structure.
The experimentation setup was extended to additional verticals and new platforms.
Key Takeaways
A winning test only matters if it becomes the new standard, so ship it permanently.
Trustworthy experiment data across domains is as important as the variant itself.
Interaction-level tracking explains the why behind a conversion number.
Like this project
Posted Oct 8, 2026
Rebuilt a multi-vertical insurance funnel as one question per page in Optimizely, with cross-domain GTM/GA4 tracking. Leads rose 10%.
