Hwang Yoo-Yup · Digital Legacy Platform · Reflection & Learnings

What This Project Actually Was

Most UX projects involve designing something that doesn’t exist yet. This one involved designing something that should have existed already — a platform for an artist whose legacy is historically significant but digitally invisible. That asymmetry shaped everything. The design problem wasn’t “how do we make this attractive?” It was “how do we make this credible enough that the right people take it seriously, while making it accessible enough that new audiences even try?”

That’s a harder problem than it sounds. Reflecting on it honestly means separating what I got right from what I would do differently — and being specific about both.

What Worked

The dual-audience architecture held. The decision to design for two fundamentally different user modes — Discovery and Research — rather than one composite persona was the right call, and the most consequential one. Every major structural decision traces back to it: the dual navigation, the two homepage CTAs, the progressive disclosure system, the Archive Mode toggle. Testing confirmed it. The professional users went straight to Archive Mode. The discovery-mode participants explored through themes and images. Neither felt like they were on the wrong site. That was the goal.

The Motifs section exceeded expectations. It was the riskiest section — the one with no direct precedent in comparable platforms, and the one that required the most interpretive writing and structural design. All four industry professionals responded to it more strongly than any other section. Alice described it as contextual scaffolding that institutions rarely achieve digitally. Nelly said it was the first thing on the platform that felt written for the work rather than about it. The section was built because research said it was needed. Testing confirmed it was the right call — and the strongest differentiator.

Real participants produced real insight. Conducting interviews with actual industry professionals — an art advisor, a gallery dealer, a museum director, an artist — rather than recruited general-population participants produced qualitatively different data. These were people with formed professional opinions and specific, articulable standards. They did not hedge. They told me exactly what they expected and exactly where the design fell short. That directness made the synthesis cleaner and the design decisions easier to justify.

The research was traceable all the way through. From the initial interview questions to the synthesis clusters to the personas to the POVs and HMWs to the IA to the individual component decisions — there is a clear documented line from what participants said to what was built. That traceability mattered in prototype testing: when something worked, I knew exactly why. When something failed, I could trace it back to a gap in the research.

What I Would Do Differently

I would have validated the navigation labels earlier. The “Journey” label confusion was predictable in hindsight. The word carries a metaphorical weight that creates ambiguity in a navigation context — and navigation labels are one of the easiest things to test without a working prototype. A simple card-sort or label-matching exercise in the research phase would have surfaced this before wireframes. Instead, it was caught in prototype testing and committed as a post-testing change. That’s a later, more expensive fix than it needed to be.

I would have built the motif cross-linking into the initial prototype. The observation that Nelly couldn’t navigate from an artwork detail page back to the Motifs section (D3) was entirely avoidable. The cross-linking between artwork pages and motif pages was part of the original design intent — it just wasn’t implemented in the prototype. Testing an incomplete interaction and then labeling the fix “In Review” is accurate, but it’s a gap that should have been present from the start. If cross-linking is a core feature, it needs to be testable.

I underestimated how much tone matters to professional users. Alice flagged the two tonal outliers in artwork descriptions within minutes of exploring the Works Archive — and she did it without prompting. I had built a tone guide and applied it, but not rigorously enough. Tone consistency across a large content system requires more than a guide — it requires a content audit process, and ideally a second reader on every piece. Building that process should have been a deliverable alongside the design, not a post-testing action item.

The usability test plan used simulated participants where real ones would have been more valuable. The four industry interviews were real. The five usability test participants were simulated — constructed personas based on the established Maya and Samuel models. This was a practical constraint, not a principled choice. The results were directionally useful but lack the specificity and surprise of real-participant data. For a project at this level of ambition, real testing participants throughout — not just in the initial research phase — would have produced sharper validation.