Lifted engagement 44% and cut key research tasks by almost two days by replacing a generic dashboard with one that adapts to each role.
Context
Role
Design Lead.
Timeline
Discovery through shipped redesign took roughly 4–6 months; impact was measured a year later comparing Q2 2023 to Q2 2024.
Team
Design lead plus a PM and a small engineering team for build; sales reps, professional services, and other PMs/engineers were also pulled in as stakeholders for a 2-day design sprint.
Key constraint
Usage was already low going in: only 12% of customers who landed on the dashboard engaged with it, and most had already formed the habit of bypassing it for the test list page. Any redesign had to overcome an existing habit, not just improve a page nobody trusted yet.
Where things stood
The Workspace Dashboard was built as a landing page summarizing a customer's recent tests, drafts, and highlight reels. Engagement data told a different story: customers found it confusing, couldn't easily locate what they needed, and mostly ignored it in favor of navigating directly to the test list.
The problem
The one-size-fits-all model was the actual failure.
Understood at the start
The dashboard's content and layout weren't engaging enough: a presentation problem, fixable by redesigning what was shown and how.
What it turned out to be
The real question wasn't "what should a redesigned dashboard look like," it was "what problems are customers actually facing, and how do we improve their first landing experience to drive real engagement and business value." Research also surfaced that a solo researcher, a team collaborator, and an admin needed fundamentally different things. The one-size-fits-all model was the actual failure, not the visual design.
Process
01 Reframing the problem before designing anything
What was tried. Translating research findings into job stories ("When I log in, I want to quickly resume my most recent work...") instead of jumping straight to layout options.
Why. The team's instinct was to redesign; the research suggested the problem was deeper than layout.
What it taught us. Shifting the question from "what should this look like" to "what job is this failing to do" changed the entire solution space that got explored next.
Considered and rejected
Going straight into visual redesign concepts without this reframing step. The original dashboard was arguably a product of exactly that shortcut.
[Diagram: research findings mapped to job stories]
02 Testing three distinct concepts instead of converging early
What was tried. A personalized-research-hub concept, a team-centric concept, and a fully customizable widget-based concept, tested with 11 participants via moderated and unmoderated research.
Why. Role-based needs (solo researcher vs. team collaborator vs. admin) were different enough that picking one direction on instinct risked repeating the original one-size-fits-all mistake.
What it taught us. No single concept won outright. People wanted role-based personalization, but disliked starting from a blank page, and still wanted team visibility. The findings pointed to a hybrid, not a single clean model.
Considered and rejected
Shipping any one of the three concepts as tested, each solved for one type of user at the expense of the others.
03 Letting users manually pin content, not just surfacing it automatically
What was tried. A pinning mechanism alongside dynamically surfaced recent tests and drafts.
Why. Personalization needed a manual layer, not just algorithmic recency. Testing showed people wanted a default view, but also control.
What it taught us. This became one of the most measurable parts of the whole redesign. See outcome below.
Considered and rejected
A purely automatic, recency-based surfacing model with no manual control, likely rejected because it repeats the "generic information that doesn't match my priorities" complaint from the original dashboard.
Research and evidence
[Research and evidence the decisions above rested on]
The trade-off
Gave up. The simplicity of one clean interaction model.
Testing showed that no single simple concept served solo researchers, team collaborators, and admins equally. Shipping a simple solution would have preserved the same one-size-fits-all failure that caused the original 12% engagement problem.
A hybrid, role-based dashboard combining personalization, team visibility, and manual control, more composite and complex than any single concept tested.
44%
More platform engagement. Users who engaged with the dashboard vs. those who didn't, Q2 2023 to Q2 2024.
~2 days
faster to complete key research tasks
13%
increase in dashboard engagement
922
users pinned content, July 2024
What shipped
Role-based customization, a priority-focused layout, pinned and recently-surfaced content, a workspace overview for teams, and more actionable notifications.
Among users who pinned content: 26% higher result consumption, and 5 days faster research completion.
What happened after
No major regressions surfaced after launch; iteration since has focused on extending pinning and role-based customization based on continued usage data.
What I'd do differently
Running three full concepts through independent testing before converging was thorough, but it's also the slowest possible path to a hybrid. Given the final answer combined elements of all three anyway, I'd now prototype a converged direction earlier and validate it iteratively, rather than validating three separate directions in parallel first.