The problem
The Adversent dashboard had a structural conflict baked into its original design. It was trying to support two fundamentally different user behaviours at the same time — open-ended browsing for users exploring what was available, and structured progression for users working through a course they'd enrolled in.
The result was an interface that did neither well. New users felt lost because there was no clear starting point. Returning users couldn't quickly find where they'd left off. Completion rates were suffering because the experience didn't give users enough momentum to keep going.
Online learning has a well-documented drop-off problem — most platforms lose the majority of enrolled users after the first session. The dashboard redesign was specifically focused on solving that.
The insight was that discovery and progression aren't competing behaviours — they happen at different moments. The dashboard needed to know which moment the user was in.
Research and reframing
Rather than treating discovery and progression as two modes that needed their own interfaces, I reframed the problem. The real issue was that the dashboard didn't know anything about the user's context — it showed the same interface to a brand new visitor and someone halfway through their third course.
Analysed how early users were actually moving through the platform — where they entered, where they dropped off, and what patterns distinguished users who completed courses from those who abandoned them after one or two sessions. The data showed that completion was strongly correlated with how quickly users found their next step after logging in.
Mapped the full user journey from first visit through to course completion across three distinct user states: new users exploring the platform, enrolled users returning to continue a course, and users between courses deciding what to do next. Each state had completely different needs from the dashboard.
Built and tested prototypes with real users before committing to development. Validated the adaptive dashboard concept — where the interface prioritised different content based on user state — against the original design. The adaptive version consistently reduced time-to-next-lesson for returning users.
The design solution
The redesigned dashboard used a modular card system that adapted to user context. Returning users saw their active courses and their next lesson immediately — no searching, no scrolling. New users were guided through curated starting points based on their stated goals rather than being dropped into a content library with no direction.
- Adaptive homepage layout — the primary content surface changed based on whether the user had active courses, completed courses, or was new to the platform.
- Modular card system built to surface different content types — courses, progress states, recommendations and achievements — in a consistent visual language that could be extended as the platform grew.
- Progress indicators designed to encourage continuation without creating anxiety — visible momentum rather than a countdown to completion.
- Continue watching as the primary CTA for returning users — borrowed directly from the Netflix reference point, reducing friction to re-engagement to a single tap.
- Design system established for the full platform — not just the dashboard — so future features could be built consistently without revisiting visual decisions each time.
The outcome
Early students achieved a 35% course completion rate — a significant result in a category where most platforms consider anything above 10-15% a success. The modular design system also made it substantially faster to add new courses and content types without reworking the core interface.
The platform was eventually closed due to funding running out rather than any product failure. What the dashboard work demonstrated was that completion rates in online learning are a UX problem as much as a content problem — getting users back to their next lesson quickly, consistently and without friction is the single most important design challenge the dashboard needed to solve.
