Kiwi —
an AR navigation app
for air travel

Kiwi was a collaborative project undertaken by four Interaction Design students using the Lean UX framework over eight weeks, from October to late November 2023. The brief was open-ended but focused: design a prototype for an AR app that makes the air-travel experience seamless, inclusive, and personalized — targeting the Apple Vision Pro as the development platform.

Air travel is a high-pressure, high-complexity environment where passengers routinely struggle with navigation, time management, and finding contextual information in unfamiliar airports. AR technology offered a meaningful advantage: integrated, customizable visual guides layered directly over the physical environment. It also gave us the challenge of designing something original and future-facing at a moment when AR applications were still largely uncharted territory.

Our team identified a clear pattern in existing air travel products: widespread failures in accessibility, accuracy, and overall user experience. AR technology offered a structural advantage that conventional apps couldn't replicate. Integrated, customizable visual guides that work in the context of the user's physical environment rather than requiring them to step away from it. It also gave us the challenge of designing something genuinely novel at a moment when the platform itself was still new.

For our design method, we used an adaptation of the Lean UX framework — a process centered on cross-disciplinary collaboration within short timeframes, with continuous learning through iterative design as its core operating principle. In the words of Jeff Gothelf and Josh Seiden, Lean UX "brings the nature of a product to light faster, in a collaborative, cross-functional, and user-centered way."

Our timeline was eight weeks: two three-week sprints, followed by two weeks of refinement. Working in person was ideal but challenging — our team was composed of full-time students with varying schedules. Frequent virtual communication within Microsoft Teams became essential to maintaining a proactive working environment within those constraints.

The Lean UX Canvas guided the team through a structured set of collaborative brainstorming exercises before any design work began. Its purpose is to surface and articulate a team's initial assumptions — establishing a shared foundation for hypotheses, design plans, and testable MVPs. This allows teams to validate ideas quickly while minimizing wasted effort.

An early example of speed with Lean UX
We initially assumed that Kiwi's value would be concentrated in airports the user was unfamiliar with. Through interviews and testing, we found the need was consistent regardless of how familiar a traveler was with a given airport. This shift would greatly affect how we scoped the product.

At the start of each sprint, the team worked through the Lean UX Canvas together — conducting research and brainstorming exercises for each component. Meetings were facilitated in person, supplemented by ongoing coordination in Microsoft Teams. From this work, we built a product backlog to track proposed solutions and hypotheses, and a sprint backlog defining which items would be built and tested in Sprint 1.

The remaining two weeks of Sprint 1 were dedicated to user research interviews and iterative MVP development. We scheduled five interviews with participants matching our projected user type. Due to our time constraint, the first session was conducted without any MVP artifacts — focusing instead on open-ended inquiry to learn about the target audience. In subsequent interviews, participants began with open-ended questions before being guided through the current MVPs. Designs were updated between each session, and we completed four iterations in total.

One significant shift emerged near the end of the first week. While working on Navigation Assistance, I realized that a critical part of the flow — route creation — had been assigned to another teammate, Michelle Hayden. Route creation would show the user journey to set a custom route originating from the homepage, and would incorporate the maps feature. In our initial iteration, we had designed separate screens for each. Recognizing the overlap, Michelle and I decided to develop our MVPs as a pair. We still kept Navigation Assistance and Route Creation separate for testing purposes, but worked together from that point forward. My responsibilities remained focused on the user journey screens and interface styling — but this pairing also meant taking on significantly more depth with the interactive map feature.

The retrospective was conducted at the end of the second week through a virtual meeting, using a template provided by the course instructor. Team members shared thoughts anonymously via digital post-its — cameras and microphones off — to keep responses honest and unbiased.

What went well: The team maintained strong chemistry, communication, and work ethic throughout the sprint. Our iteration pace stood out — the MVPs consistently reflected feedback from the prior interview session, showing that research was being actively integrated into the design.

What could have gone better: While the interviews were productive, we all identified a need for better organization and structure in how walkthroughs were moderated, and cleaner transitions between MVP segments. UI consistency was also raised — we needed to establish and apply a defined style system across the product.

What we would try next: The team aligned on one goal for Sprint 2 — a functional prototype. We wanted a dynamic, interactive experience ready for the next round of users, which meant narrowing our focus to complete, consistent interactions for the core features, integrated into a coherent flow.

Entering Sprint 2, the team revisited the Lean UX Canvas with new insights gathered from Sprint 1 research and user feedback. The business problem was updated to reflect newly surfaced struggles — navigation near car rental facilities and accounting for TSA wait times. The proto-persona was expanded to address additional needs: assistance with trip planning and timeframe management, access to live store-hours information, and guidance on exit routes.

These updates cascaded into the outcome-to-impact mapping, which was revised with new leading and lagging outcome paths. We added three new hypothesis statements — introducing a trip planner feature, TSA wait time integration, and in-app pop-up screens for restaurants and stores — and removed three prior statements that had become redundant. After reordering prioritizations and updating the risk and MVP tables, we built the Sprint 2 backlog.

Sprint 2 followed the same rhythm as the first: continuously learning from interviews and refining designs in response. The key difference was that the product backlog and style system were now solidified, giving the Figma work significantly more consistency. Design plans became more narrowly scoped, so iterations focused on improving interactions rather than revisiting visual design; rough screens were converging into a final prototype.

The most technically intensive work in Sprint 2 was the Kiwi Maps feature. Navigation Assistance and Route Creation had evolved from separate MVP flows into a unified product feature — and building it required an interactive, scrollable map component with actionable elements. To achieve this, I built a component set called map filtering with 32 variations covering every possible layout for map filter combinations. I constructed a nested filter bar component — also with 32 variations — that enabled click interactions to switch between map states. We configured over 130 interactions in total to support full navigational functionality within the feature.

Once the map component was complete, I designed the Kiwi Maps Interface: a component set containing both map filtering and a sidebar component with all sidebar navigation variations — providing the full interface context in which the map would be used.

The final retrospective served both as a reflection on Sprint 2 and a setup for the remaining refinement phase. Consistent team collaboration and continuous learning remained our clearest strengths.

I identified two priorities going into refinement: completing screen connections across all prototype walkthroughs, and improving the visual waypoints frames — which had not yet reached the same design quality as the rest of the project. The team also agreed that visual waypoints would be more effective as an AR experience if the frames contained a video background.