OVERVIEW
Artworks Discovery is the first unified entry point for art content the company built. It lets users explore works from galleries, Artnet Auctions, partner auction houses, and historical sale records within a single interface, across a database of more than 50 million items.
On the surface this is a horizontal integration of content. Underneath, it held three thornier challenges: data structured differently across heterogeneous sources (data schema inconsistency), market semantics that conflict with one another (market semantics mismatch), and a difference in how different kinds of users search (search intent variance).
Some users arrive with a specific goal, looking for a particular work. Others start from an artist, a style, or a market type. So the scope grew from a single page into an exploration system that has to support several search intents at once while keeping information consistent.
The project spans two core surfaces: the Discovery list page (browse and filter across sources) and the Artwork detail page (a single work's information, brought together).
THE PROBLEM
Problem statement
When browsing artworks, I want to find everything I need in one search system, so I don't have to jump between different sites to gather it.
Art content was scattered across four sources: galleries, Artnet Auctions, partner auction houses, and historical sale records. Each had a different data structure and market meaning, and they were not consistent with one another.
Under the existing setup, users had to switch between systems and piece the information together themselves to understand a work's full market context.
HMW
How might we present artwork data from different sources in one harmonious design language, without creating conflict between them?
And avoid treating the primary market (gallery sales) and the secondary market (auction results) as directly, like-for-like comparable data.
THE SOLUTION
I redesigned two core surfaces: the Discovery list page handles cross-source integration, browsing, and filtering; the Artwork detail page presents a single work's context. Here they are separately.
Surface 01 — Discovery list page
An aggregated entry point that lets galleries, auctions, and sale history sit together in one smoothly browsable list.
Every work uses the same card design, but keeps its own market meaning: gallery works are marked "Price on Request," live auctions show "Live Now" with an estimate, and closed sales show "Results Available." One layout holds different market states without flattening them.
Filters split by data source and state, including Available Now, Artnet Auctions, Galleries, Auction Houses, and Ending Soon, to work through more than 50 million works. The same layout extends from phone to desktop, so a work's reading structure doesn't shift with screen size.
Surface 02 — Artwork detail page
The detail view for a single work, carrying its market context clearly by source.
The redesign gives source information, market data, and recommended works each a block of their own, instead of crowding into the same layer.
Galleries and auctions carry different fields and market meanings, so the detail page keeps one layout hierarchy while adjusting per source: gallery works lean toward exhibition and background context, auction works toward results and estimates, to avoid visual fragmentation.
Gallery work: exhibition and background context.
OUTCOMES
Because this is a newly built Discovery entry point, the following are early signals (early behavioral observations), not a strict before/after comparison. Early observations: average browsing depth rose to about 1.6×, users shifted more from single lookups to continuous browsing, and cross-source switching dropped.
1.6×
Browsing depth
Average browsing depth rose to about 1.6×, with users shifting from single lookups to continuous browsing.
~30%
Monthly return rate (early data)
Used to watch whether one-stop exploration is starting to become a habit, not to represent final growth.
LEARNINGS
After launch, I ran usability testing with four participants tied to the art market: two senior gallery professionals, a gallery founder, and a high-net-worth first-time collector. Each session combined a 10-minute interview with 20 minutes of hands-on tasks across four scenarios, all conducted remotely.
I analyzed behavior on two axes, familiarity with technology and familiarity with the art market, and arrived at three main user types: the Visual Explorer (browses by style and visual instinct), the Market Strategist (searches by work name and market info), and the Gallery Promoter (focused on artist exposure and visibility).
Usability was stable overall (4.5 / 5), but one task showed a clear drop. The key problem concentrated on finding a specific work, the only clear break point across the four scenarios.
The most severe problem was a mental-model error: users assumed the header's global search equaled the Discovery page search, but the two have different scopes. That led to wrong search paths, missing results, and failed tasks.
The full research report walks through, in order, the method, the three user personas, the core insights, and the four contextual test scenarios.
1 / 13 Background: research goals and method.
The testing confirmed the core issue was a misread of the system boundary. So the priorities were:
The four participants leaned toward gallery-side users, so the results describe a specific professional group, not the whole market. The value of the research is in clearly defining the scope its conclusions apply to.
And the deeper takeaway: designing a cross-source system is, at its core, designing how users understand information sources, and what those sources mean in different market contexts.