26
NYC Tourism Redesign
From IA Research to Applied Design
NYCTourism.com has two meaningfully different user groups with conflicting mental models: tourists looking for things to do, and businesses looking to participate in the tourism ecosystem. The site's IA was built almost entirely for the former, leaving the latter to navigate a structure never designed for their goals. Business-related pathways were buried, inconsistently labeled, and written in language only internal teams used.
Phase one: recruited five business owner participants, ran semi-structured interviews to map mental models, card sorting to surface categorization expectations, and tree testing to validate specific navigation paths with quantified success rates. Phase two: built a Figma prototype introducing dual entry pathways — one for visitors, one for businesses — unified under a shared visual system, iterated through three feedback rounds, and validated the revised structure through second-round usability testing.
- Dual-entry navigation system with clear audience routing established at the homepage level
- Business-facing pathways surfaced to primary navigation, removing the need to hunt through tourist content
- All revised labels drawn directly from language used in participant interviews
- Second-round usability testing showed 80%+ task success rate across both audience types
A site can't serve two audiences well if it treats them as one
The moment you have two user types with genuinely different goals and mental models, you have an IA problem that content updates alone can't solve. You need separate entry points and separate structural logic. Trying to serve everyone through the same architecture serves no one particularly well.
Research is only useful when it drives decisions, not just documents problems
The midterm study surfaced the gap clearly. The final redesign was the test of whether those findings could be translated into something real. Every design decision that couldn't be traced to a specific research finding got questioned. That constraint is uncomfortable, but it's what makes the output trustworthy.