Rolli 123
Mobile app that helps users discover, book, and manage flights with real-time updates and seamless UX.
Industry
SaaS · Media Tech
Scope of work
Product strategy · Research · End-to-end design across web and mobile · Web & Mobile design · Design system · Prototyping
Duration
6 Months
Introduction
A platform people only opened when they had to.
Rolli connects journalists with vetted experts. When I joined, it worked - journalists could search a directory, find someone to interview, and move on. The problem was right there in the usage pattern: journalists only needed Rolli when they had a story to file. Experts only checked Rolli when they remembered to. Both sides were using the product on a need basis, and need-based products are hard to grow.
The team was preparing to raise, and investors would want to see engagement, retention, and a path to becoming something more than a search box. My job, working directly with the founder and engineering team as the only designer, was to figure out what Rolli could become - not just how it could look.
The Brief vs The Problem
Nick asked me to improve the directory. I made the case for changing what the product was for.
Rolli connects journalists with vetted experts. When I joined, it worked - journalists could search a directory, find someone to interview, and move on. The problem was right there in the usage pattern: journalists only needed Rolli when they had a story to file. Experts only checked Rolli when they remembered to. Both sides were using the product on a need basis, and need-based products are hard to grow.
The team was preparing to raise, and investors would want to see engagement, retention, and a path to becoming something more than a search box. My job, working directly with the founder and engineering team as the only designer, was to figure out what Rolli could become - not just how it could look.
The real issues, drawn from user research, usability testing, and competitor analysis, were:
Limited search and filtering. The only filters were radius and location, not nearly enough for journalists looking to match expertise to a story.
A generic, transactional experience. The platform felt like a basic database, with no clear difference between verified and unverified experts.
No mobile access. Journalists couldn't use Rolli on the move, the moment they often needed it most.
Trust and credibility gaps. Visual inconsistency and the lack of a structured vetting process made the product feel less professional than the experts it was surfacing.
No personalisation. The system didn't remember anything between sessions, so every search started from scratch.
All real problems. But they were symptoms. The underlying issue was that Rolli was built for a low-frequency use case, and no amount of feature work fixes a low-frequency use case if you don't change what the product is for.
I reframed the brief with Nick. Instead of "make the directory better," what if we made Rolli a place both sides wanted to visit even when they didn't strictly need to? For journalists, that meant something worth checking in the morning with coffee. For experts, it meant a reason to log in beyond updating their bio.
That reframe shaped everything that came after.
THE DECISIONS
Four decisions to move Rolli from directory to destination.
DECISION 01
Turn news into the front door.
We built a Trending News surface tied to expert recommendations. If a major story broke, say, the Russia and Ukraine war, journalists would see the story alongside the foreign policy experts available to comment, with the ability to filter by tag or jump straight to search. The point wasn't the news itself. It was giving journalists a reason to open Rolli without a specific assignment in mind. Discovery becomes a habit, not a task.
DECISION 02
Make experts feel like they're building something.
Expert profiles went from static bios to something closer to a creator dashboard. Social media integration pulled in their existing posts, so experts didn't have to maintain Rolli separately from the platforms where they already had a presence. Profile metrics, views, clicks, and search appearances, gave them a reason to check in and a sense of progress. Trending tags showed them what journalists were searching for right now, so they could position themselves accordingly. This also opened the door to monetisation later, through profile boosts and promoted placements on top of the base subscription.
THE DECISIONS
DECISION 03
Make discovery feel personal, not transactional.
The old search had two filters: radius and location. We added language, availability, and expertise filters, plus a recommended-experts surface based on the journalist's prior searches and reading. The point wasn't filter completeness. It was making the system feel like it knew them. We also surfaced preferred contact methods directly on expert profiles, so journalists on deadline didn't have to guess how to reach someone.
DECISION 04
Build the system that makes the product feel credible.
Rolli didn't have a design system. It had a loose UI library with inconsistent components, no documentation, and no rules for how things should behave. For a product whose entire value proposition is trust between journalists and experts, that's a real problem. A platform that looks inconsistent feels unreliable, and an unreliable-looking platform is a hard sell to a journalist deciding whether to quote one of your experts.
So alongside the new features, I built the foundations: a proper component library, type and colour systems, documentation for how things behave, and a visual language that signalled "professional tool" rather than "early-stage startup directory." It wasn't a side project. It was part of the credibility argument. Consistency is credibility. A product that holds together visually is one users trust to hold together functionally.
TESTING & THE PIVOT
Our first hypothesis was wrong. Users told us in week two.
Before we built anything, I ran concept tests on the first hypothesis: Trending News integrated into the search experience. The thinking was that journalists searching for an expert would also see what was breaking in that topic area, combining the two surfaces a journalist needed.
Users didn't buy it. The pattern was consistent. When journalists were on deadline, they wanted to search and get out. When they were exploring, they wanted to browse and discover. Mixing those modes in one surface created friction in both directions. Search felt cluttered, discovery felt buried.
Same content, two distinct modes, each doing one job well.
So we separated them. Trending News moved to the homepage as a discovery surface, the first thing a journalist sees, designed for the morning-coffee scroll rather than the deadline rush. Filtering by tag and direct search sat at the top of the same surface for journalists who arrived knowing what they wanted.
This was the pivot the rest of the product hinged on. Without it, we'd have built a feature-rich search page that nobody used the way we intended.
ROLLOUT
The launch strategy was as much of a design decision as the product itself.
A redesign this large carries a real risk. Existing users come back to find the product they knew has changed, can't find what they used to use, and leave. Especially dangerous for a platform whose entire business model depends on returning users.
I made the case for a staged rollout rather than a single big-bang launch.
DECISION 01
Experts first.
We redesigned and shipped the expert side before touching the journalist experience. Experts are the supply side of the marketplace. Getting their profiles richer, more credible, and more complete first meant that when journalists arrived at the new experience, the experts they found were already presenting themselves better.
DECISION 02
Journalists next, in stages.
Rather than reveal every new feature at once, we layered them in. The new search experience first, then Trending News, then the events surface, then personalised recommendations. Each release was small enough that users could absorb it without feeling the product had changed under them.
DECISION 03
Each release as a research moment.
Every rollout gave us another round of feedback. One example I keep coming back to. A key "add" action was sitting in the tertiary hierarchy of the page, and we kept hearing from users that it took them too long to find. Promoting it to a primary action made an immediately visible difference in adoption. Small change, real impact. The kind of thing you only catch if you're watching the product after launch, not just before.
The rollout strategy traded short-term polish, a single dramatic relaunch, for long-term retention. For a platform whose growth depended on bringing existing users back, that was the right trade.
CHALLENGES
The hard calls weren't about what to build. They were about what to hold back.
Balancing user feedback with vision.
Some users were attached to older workflows, and we had to balance innovation with familiarity. The staged rollout helped. But there were still moments where I had to push for the new pattern even when early feedback was mixed, because the long-term direction was clearer than any single piece of feedback.
Balancing simplicity and functionality.
Early versions of the advanced search filters overcomplicated the UI. The temptation was to keep adding controls because we could. The harder discipline was to strip it back to the filters that mattered most, and put the rest behind progressive disclosure.
OUTCOME
Journalists started opening Rolli without a reason. That was the point.
By the time the full experience was live, the product was doing something it hadn't done before. Bringing users back without a specific task in mind.
Journalists started checking Rolli the way they'd check a news app, opening it in the morning to see what was trending, who was available on which topics, and what conversations were worth pitching into. Experts had a reason to log in beyond a profile update: their visibility metrics, their position in trending topics, the social proof their existing content was building inside the platform.
The team I worked with reported strong engagement and retention gains post-launch. The specific metrics belong to the client, but the directional outcome was clear. Rolli moved from a directory to a destination.
REFLECTION
Three things I'd do differently if I started this project today.
Push for measurement infrastructure earlier.
We made strong product decisions, but a lot of the post-launch story relied on qualitative read rather than instrumented data. I'd want clear baseline metrics defined before the first release, so each staged rollout was answering a sharper question than "did this land?"
Build the design system at the start rather than alongside the features.
We paid a real tax in rework for components built before the system existed. For an eight-month engagement on a startup it's not always feasible to front-load it fully, but I underestimated how much that foundational work would pay back.
Shorten the cycle between concept and real users.
Eight months is a long time to be designing without journalists and experts actually using the new experience. Some of our best decisions came from post-launch usage, and I'd want more of those earlier.
LESSONS
Three lessons I've carried into every project since.
Push for measurement infrastructure earlier.
We made strong product decisions, but a lot of the post-launch story relied on qualitative read rather than instrumented data. I'd want clear baseline metrics defined before the first release, so each staged rollout was answering a sharper question than "did this land?"
Build the design system at the start rather than alongside the features.
We paid a real tax in rework for components built before the system existed. For an eight-month engagement on a startup it's not always feasible to front-load it fully, but I underestimated how much that foundational work would pay back.
Shorten the cycle between concept and real users.
Eight months is a long time to be designing without journalists and experts actually using the new experience. Some of our best decisions came from post-launch usage, and I'd want more of those earlier.
