← Back to work
Case Study 02

TraveList

2024 · UX Designer, 3-Person Team · Figma

01 · Overview

An AI-powered travel planning app consolidating itineraries, packing lists, and bookings using a supervised learning recommendation engine.

Role
UX Designer, 3-Person Team
Duration
Oct – Dec 2024
Course / Tools
INFO693 · Figma
02 · Problem Framing

Existing productivity apps can't generate contextual, trip-specific suggestions; niche travel apps solve only a sliver of the problem.

03 · Key Decision: Explainable AI
Problem
A black-box AI model would undermine the app's core promise of reducing travel-planning anxiety.
Decision
Chose supervised learning over unsupervised/reinforcement approaches specifically because it keeps recommendations traceable to the inputs a user actually gave.
Grounded in Norman's principle that interactive systems need transparent feedback loops.
04 · Core Flow
Login→ Home→ Packing List (AI)→ My Trips→ Explore
05 · Ethical Design Decisions

Minimal data collection with explicit consent, anonymized/secured training data, bias audits on diverse datasets, and user-facing feedback/error reporting.

06 · Research & Validation

Heuristic evaluation surfaced 3 issues: system status under poor connectivity, an ambiguous "Connections" menu label, and a visually busy Explore page. Concept validation interviews (4–5 participants) surfaced two themes, each confirmed by a second, independent signal (Figma comment threads).

Decision: Offline Support
Problem
Connectivity anxiety surfaced repeatedly in interviews.
Evidence
4–5 concept validation interviews plus independent Figma comment threads converged on the same concern.
Decision
Offline mode + local storage prioritized as core, not a later add-on.
Decision: Menu Clarity
Problem
"Connections" menu label caused confusion about what it contained.
Evidence
Confirmed independently by both interview feedback and Figma comment threads.
Decision
Label rewritten and menu restructured for clarity.
07 · Impact

Directional, not quantified: no task-based measurement was run. The strongest signal: two independent research methods converged on the same two problems, giving real confidence the fixes addressed genuine friction.

08 · Recommendations

True offline mode, inline error handling and validation, and accessibility improvements: contrast, readable fonts, and voice assistance.

09 · Reflection

Strongest asset was decision reasoning and ethical framing; with more time, the next step would be structured task-based usability testing to replace directional findings with measured outcomes.

Next case study: PawPal → Say hello →