Information Architecture · UX Research
Kayak
An information architecture evaluation of Kayak.com: auditing how travel content is organized, restructuring navigation around how people actually plan trips, and validating the new structure with users.
Methods
Content inventory · Card sorting · Tree testing · Heuristic evaluation
Testing tool
UXtweak
Deliverables
Annotated sitemap · Mid-fidelity wireframes
01
One trip, many separate bookings
Kayak is a high-traffic global travel platform for booking flights, hotels, cars, and cruises. Its visual design is consistent, but the end-to-end booking experience can feel fragmented, especially when people want to book flights, hotels, and transportation together in one session.
This project evaluated Kayak's information architecture and proposed improvements to simplify trip planning while preserving flexibility for individual bookings.
Analyse
Audit the current IA, navigation structure, and content organization of Kayak.com.
Improve
Propose and iterate on structural improvements, optimizing for clarity and task completion.
Validate
Test improvements through card sorting, tree testing, and heuristic evaluation.
Who we designed for
PRIMARY USERS
Frequent travelers
- Travel often for work or personal needs
- Compare options quickly using filters
- Value speed, flexibility, and real-time updates
PRIMARY USERS
Occasional travelers
- Travel less often and need clearer guidance
- Prefer simplified booking flows and decision support
- Rely on reviews, recommendations, and structured navigation
Two use cases framed the work throughout: booking a single trip component, and planning a full trip.
02
Where the existing structure breaks down
Before restructuring anything, we needed to understand what was already there. A content inventory audited Kayak's primary navigation, major booking flows, and support content in a spreadsheet of pages, labels, and hierarchy.
Redundant content
Similar content appeared across multiple sections, such as Deals, Explore, and Flights, and multiple entry points led to the same actions.
Overlapping labels
Different labels described similar actions, with no clear distinction between planning content and booking content.
Scattered support
Help, account, and informational content lacked a unified structure.
How the audit informed the proposed structure
Removed
- Exploratory paths that duplicated booking intent
- Overlapping "deal" entry points scattered across sections
Merged
- Deals consolidated under each category: Flights, Hotels, Rental Cars, Cruises
- Booking flows unified under core categories
- Support grouped into a single Help & Support section
Re-labeled
- Packages given intent-based sub-labels such as Destination and Seasonal Packages
- Travel Ideas organized into Popular Destinations, Weekend Getaways, and Seasonal Travel
These findings shaped content consolidation and set up the structure for card sorting.
03
Letting users organize the content
Three rounds of closed card sorting, a pilot and two full rounds, tested whether people would group Kayak's content into the categories we proposed. Each round tightened the cards and the category labels.
Pilot
Vague categories and overlapping cards meant no card reached full agreement. Ambiguous labels kept participants from placing cards where expected, and secondary pages like airline fees, low-fare tips, and price alerts could be consolidated.
Change: cards reduced from 40 to 34 for Round 1.
Round 1
Participants understood the main travel flows, such as Flights, Hotels, and Rental Cars, but didn't cleanly separate planning from booking, research from transaction, or guidance from support.
Change: abstract categories like "Plan your trip" and "Search and Booking" were replaced with explicit ones such as Hotels, Packages, and Cars.
Round 2
Product-based buckets worked well: Find flights, Flight deals, Find car rentals, Car rental deals, Package deals, and Seasonal travel. Flight content was often pushed into Packages, booking management did not map cleanly (suggesting a possible "Trip planner" category), and corporate content split across About, Help/Support, and Not sure.
Change: a dedicated "Flights" category was added to the navigation. The suggested "Trip planner" category was not incorporated into the final sitemap.
04
A structure built around tasks
The final sitemap brings the research together: seven top-level sections organized by travel task, with each product pairing a search page and its deals, plus utility navigation for Sign In, My Trips, and About KAYAK.
How the research shaped it
Content inventory
Exposed fragmented flows and redundancies, leading to a simplified structure, the removal of filters, and stronger core categories.
Card sorting
Grouped content the way people think about travel tasks, separating booking from exploration and informing clearer top-level categories.
Tree testing
Iterated the structure across rounds, leading to task reframing, improved labels such as "Build your Package", and a clearer, task-driven hierarchy.
05
Testing whether people could find their way
Tree testing in UXtweak evaluated the navigation structure across a pilot and two rounds of four tasks. Findings from each round fed back into both the tree and the tasks.
Pilot
59.4%
Round 1
75%
Round 2
81%
Pilot
Filters in the tree distorted navigation
Change before Round 1: filters were removed from the tree, top-level categories simplified, labels clarified, and tasks reframed to better reflect user intent.
Round 1
Core categories worked; deals and bundles didn't
Change before Round 2: the deals task was reframed to align with "Flight Deals", and Packages sub-categories were relabeled, adding "Build your package" to reduce ambiguity.
Round 2
Clearer labels, better alignment
87.5%
75%
87.5%
75%
Flight deals and bundled booking improved, indicating better alignment with user expectations. The first two tasks showed slightly lower success rates in Round 2, which also involved a larger participant sample.
What tree testing showed
Flights and Hotels performed consistently as top-level categories across all rounds, and overall task success rose from 59.4% in the pilot to 81% in Round 2 alongside task reframing and clearer labeling. Clearer labeling reduced mismatches with users' mental models, though bundled booking still needed stronger guidance - confusion in exploratory and bundled tasks revealed the need to better separate booking, planning, and discovery pathways.
06
From structure to interface
Navigation is only part of the experience. A heuristic evaluation of two core booking tasks, scored independently by two researchers and averaged, showed where the flows themselves add friction.
Book a round-trip flight
3.5
Strong for single-task booking.
Book a full trip: flight + hotel + car
2.36
Lowest on supporting mental models, preventing errors, and helping users recover from mistakes (2.0 each).
- Fragmented experience for multi-step trip planning
- High cognitive load from excessive filters and no unified flow
- Limited visibility of progress, feedback, and the overall trip summary
The wireframes below are proposed improvements, not changes implemented on Kayak.
Proposed · Task 1
Booking a round-trip flight
Finding: too many filters and options at once; feedback such as price changes isn't always visible, and users lack a sense of progress.
- Progressive disclosure for filters, with a strong default (price)
- A step-by-step flow (Departure, Return, Review) with a progress indicator
- A persistent trip summary where users can override a selection, acting as an undo
Supports visibility of system status and recognition over recall.
01 · Flight search
02 · Departure selection
03 · Return selection
04 · Booking review
Mid-fidelity wireframes: representative screens for each stage of the proposed flow. Trip details differ between screens.
Proposed · Task 2
A unified Trip Builder
Finding: each component is booked separately, with no unified workflow or central summary, so users track selections, pricing, and details themselves.
- One linear flow treating flight, hotel, and car as a single package
- Hotel dates pre-populated from flight dates, preventing misinterpretation
- Components always visible and editable, with the option to skip parts of a trip
Optimizes for recognition over recall and flexibility.
07
What the research showed
Research insight
People approach travel sites through tasks, such as booking flights, hotels, or packages, rather than conceptual categories like planning or discovery.
Research insight
Labeling strongly influenced success. Several navigation failures came from unclear labels rather than structural issues.
Research insight
Consolidated trip booking is inherently complex and needs stronger guidance than navigation placement alone.
What we'd do differently
- Test more diverse task framings in the tree test, particularly discovery and flexible-planning scenarios with different entry points
- Test interactive prototypes of key flows, such as building a package or an all-inclusive trip, rather than evaluating the IA alone
Future directions
- Further test and refine the Packages and bundled-booking entry point
- Explore clearer integration of the AI Trip Planner within navigation and planning workflows
- Validate the IA across mobile and desktop navigation patterns