Plot.gr v2
Problem
For a buyer, Plot.gr is a site you visit for a few weeks. For an agent, it is software you work in every day. Plot.gr is a Greek real estate marketplace with 360,000 listings. Agents get a free professional account and a dealer panel — listings, billing, subscriptions, credits, promoted placements, statistics — and that panel is where the supply side of the marketplace lives. By 2026 the numbers were moving the wrong way: sixth in Greek real estate by third-party estimates, down two positions, with country rank and traffic falling alongside. GA4 showed the decline was specific rather than general, and Clarity located where attention stalled. Sales and support had been carrying the same two requests for months — let an owner assign a property to an agent, and let a buyer request a viewing of one that agent manages. Neither existed. Search had been built for browsing a country, not for choosing a neighbourhood. The panel had accumulated thirteen sections and two parallel navigation systems. And the existing design system was partial and inconsistent, so nothing above it could be built at scale. None of it was only a visual problem.
Solution
A structural redesign of the two surfaces where the marketplace actually works: search, and the professional panel. For buyers, a map that answers what can I get here rather than how many properties exist, and filters that change with the category being searched. Map and list stay coupled, and the list can be held still to be read instead of re-querying on every pan. For agents, one navigation hierarchy instead of two, with the commercial functions gathered into a single family and the account page split so private administration and public profile are finally distinguishable. The two requests sales had been carrying became part of that structure rather than additions to it: an owner can assign a property to an agent, and a buyer can request a viewing of one that agent manages. Both delivered on a rebuilt design system, without which four listing types across three viewer roles and three platforms could not have shipped alongside a parallel build.
Rebuilding the design system
The existing system had grown by accretion — years of additions, third-party packs and one-off components layered on top of each other, with no shared tokens underneath. Nothing was reusable enough to build on and everything was similar enough to look intentional.
I rebuilt it before designing a single screen. Variables for colour, spacing and type, semantic tokens so a value carries a meaning rather than a hex, dark mode defined at token level rather than as a second set of components, and a component library reduced to what the product actually uses.
This was the schedule decision of the project. Four listing types across three viewer roles and three platforms is a combinatorial problem, and development was already building while design ran. Two designers could not hand-build those permutations and stay ahead of the build. The system was not documentation produced after the work — it was what let the work keep pace with it.
YEAR
February – July 2026
CATEGORY
Marketplace UX
SCALE
360,000+ listings · 150+ design frames
TEAM
1 Project Manager · 2 Designers · 3 Front-end · 2 Flutter · 2 Back-end
ROLE
I led design for Plot.gr v2 — direction, system architecture and quality bar across the full redesign.
A homepage instead of a front page
Plot.gr opened like a newspaper: everything the site contained, laid out at equal weight, with no view on what the visitor was there to do. Whole sections of the product existed but were effectively unreachable from the entry point.
The new homepage starts from intent. Search leads, and returning visitors get feeds built from what they have already searched for, so the page adapts to a buyer looking in one area rather than presenting the entire country. Services — the parts of the business that were buried — are surfaced as their own section rather than left to navigation.
A marketplace homepage is not an index of inventory. It is the first filter.
