Web Redesign
Information Architecture
UX Research
Amtrak (National Railroad Passenger Corporation)
Rebuilt Amtrak's navigation around how travelers actually think
Over three months, I worked with a team of four to run user interviews, card sorting, tree testing, and usability testing across 20 participants to diagnose exactly where and why users failed. I personally moderated research sessions, synthesized findings into information architecture redesign decisions, and designed the navigation bar and mega menu in the mid-fidelity prototype, rebuilding the site's structure around how travelers actually think about a trip.
Timeline
Feb – May 2026 (3 months)
My Role
UX Researcher and Designer
Skills
Interviews, Qualitative Research, Card sorting, Affinity mapping, Tree Testing, Usability Testing, Prototyping
Team size & Members
James Huang (Me), Shelly Guan, Katherine Cajamarca, Betty Yang
My Role in the Project
I led two critical areas: personally moderating research sessions across all four research phases and designing key interface components in the mid-fidelity prototype, while contributing to the broader team effort on research synthesis, IA decisions, and site map restructuring.
Primary responsibilities:
•
Personally moderated 2 user interviews, 2 card sorting sessions, 2 tree testing sessions, and 1 usability testing session across all research phases
•
Redesigned the navigation bar and mega menu in the mid-fidelity prototype, translating IA decisions directly into interface structure
•
Designed the radio button interaction in the Manage Trip flow, replacing a confusing dropdown that caused form errors in usability testing
Collaborative contributions:
•
Synthesized findings from 4 research methods into 7 concrete IA redesign decisions as a team
•
Built and iterated on the proposed site map based on card sorting and tree testing results
•
Contributed to all project phases from initial discovery through final prototype delivery
A navigation bar rebuilt from the ground up
The original navigation crammed booking tools, trip management widgets, and utility options into a single row. Users couldn't tell what was a tool and what was information. The redesigned bar introduces a two-level layout: a top utility row for Need Help?, English, and Log In, and a main row that separates browsing content on the left from trip actions on the right.
"Plan" and "My Trip" were renamed to "Travel Info" and "Manage Trip", labels that better communicate what each section actually contains and what users can do there. Book is styled as a distinct call-to-action button, not just visually separate from informational links but also serving as a prominent action entry point consistent with how other travel platforms present booking.
"Plan" and "My Trip" were renamed to "Travel Info" and "Manage Trip", labels that better communicate what each section actually contains and what users can do there. Book is styled as a distinct call-to-action button, not just visually separate from informational links but also serving as a prominent action entry point consistent with how other travel platforms present booking.
A mega menu organized around how a trip unfolds
Inside Travel Info, content is organized around three trip phases: Before You Go, At the Station, and Onboard the Train. This helps users navigate by thinking about when they need information rather than how Amtrak categorizes it, making the menu faster to scan and easier to act on.
Pre-trip topics like baggage, trip changes, and special items are grouped under Before You Go so users can find preparation details without hunting through unrelated content. Accessibility and Special Assistance was pulled out of the planning section and given its own category, creating a more direct path for users who need it.
Pre-trip topics like baggage, trip changes, and special items are grouped under Before You Go so users can find preparation details without hunting through unrelated content. Accessibility and Special Assistance was pulled out of the planning section and given its own category, creating a more direct path for users who need it.
One place for everything booking-related
The original homepage mixed booking tools, trip management widgets, and informational links together in a single navigation row with no visual separation. The redesigned booking widget consolidates Book, Manage Trip, and Train Status into a prominent tabbed widget on the homepage, giving users immediate access to the three core functions without navigating through the menu.
The same three actions also appear on the right side of the sticky navigation bar, serving as a persistent second entry point accessible from anywhere on the site.
The same three actions also appear on the right side of the sticky navigation bar, serving as a persistent second entry point accessible from anywhere on the site.
From a dead end to a clear path forward for users who get stuck
The original Need Help? page opened to a nearly empty landing page. The redesigned Help and Support page surfaces Ask Julie as an integrated widget at the top, followed by category cards for Frequently Asked Questions, Chat Live Agent, Email Us, and Call/Write Us, each with a scannable summary. Users who get stuck now have a clear, immediate path to answers.
Amtrak's Website Had the Information.
The Navigation Didn't.
The Navigation Didn't.
Amtrak's website serves millions of travelers at every stage of a trip: researching routes, preparing for departure, and navigating unexpected changes. But when users arrived with specific questions, they consistently hit dead ends. Someone looking for the pet policy searched under "Plan" and found booking options. Someone looking for station lounges clicked "Onboard" and found seating upgrades. Someone trying to bring a bicycle ended up three clicks deep inside a catch-all link with no indication it was the right place to look.
The content was there. But the navigation mixed pre-departure decisions with in-journey logistics, buried critical information behind unindicative labels, and grouped content in ways that made sense to no one outside the organization. Users didn't fail because Amtrak hadn't published the answers. They failed because the structure never let them reach those answers.
The challenge wasn't to create new content. It was to figure out why the existing structure kept failing users, and rebuild it around how they actually think.
The content was there. But the navigation mixed pre-departure decisions with in-journey logistics, buried critical information behind unindicative labels, and grouped content in ways that made sense to no one outside the organization. Users didn't fail because Amtrak hadn't published the answers. They failed because the structure never let them reach those answers.
The challenge wasn't to create new content. It was to figure out why the existing structure kept failing users, and rebuild it around how they actually think.
Three challenges drove this failure
Categories don't reflect how users think about their trip
Information groupings and labels don't match user expectations, forcing users to browse multiple sections before finding what they need.
Critical information is buried too deep
Content requires too many clicks or sits behind vague, unindicative labels that give no hint of what's inside.
Labels don't reflect the actual content of the page
Ambiguous wording leads users to look in the wrong place entirely.
Project Goals
Understand how travelers navigate and find information
Identify where and why users fail to find content on Amtrak's website through multi-method user research.
Redesign the information architecture around users' mental models
Regroup, restructure, and relabel Amtrak's navigation so content is placed where users expect it, surfaces at the right moment, and is described in language that matches how travelers think.
Validate the redesign through prototyping and testing
Build and test a mid-fidelity prototype to confirm the restructured navigation resolves the failures identified in research.
How might we redesign Amtrak's navigation to close the gap between how Amtrak organizes its content and how travelers expect to find it?
/ Step 1: User Interview
Listening to where users get stuck
We started with user interviews to understand where and why users struggled on Amtrak's website. Rather than asking users to evaluate the site directly, we probed how they approached finding specific information and what they expected to find where.
One thing stood out immediately: booking a train ticket was not the problem. Participants described the core booking flow as straightforward. Auto-populating station names, results sorted by departure time, and clear confirmation emails all contributed to a smooth experience. Users were satisfied with completing the primary task. The problem started the moment they needed anything beyond that.
The interviews also surfaced a second category of pain points around interface design: the "Guest Rewards" label being used in place of a standard login entry point, and date picker behavior that didn't align with travel industry conventions. These fell outside our scope but confirmed that the navigation problems were not isolated issues.
The interviews also surfaced a second category of pain points around interface design: the "Guest Rewards" label being used in place of a standard login entry point, and date picker behavior that didn't align with travel industry conventions. These fell outside our scope but confirmed that the navigation problems were not isolated issues.
Insights
1.
Information is not found in users' expected locations.
Users looked for content in places that made intuitive sense to them, only to find it wasn't there.
2.
Critical information is buried too deep.
Reaching key content like baggage allowances, refund policies, onboard food options, and station lounge availability required too many clicks, often behind generic catch-all labels.
3.
Labels don't match how users think.
Vague or ambiguous wording led users to look in entirely the wrong sections. Some participants gave up on the site entirely and turned to Google or AI to find the answers they needed.
/ Step 2: Affinity Mapping
Synthesizing what we learned into themes
After interviews, we grouped observations and quotes into an affinity diagram to identify patterns across participants. This process helped us make a critical scoping decision: while interface design issues existed, the majority of pain points clustered around one area: finding specific information and details within Amtrak's site, most of which lived under the PLAN category. This became the focus of the redesign.The affinity diagram pointed to four redesign opportunities: relabeling confusing items, reorganizing content categories, surfacing buried information, and adding tools to aid the information-finding process.
/ Step 3: Card Sorting
Understanding how users naturally group travel information
With a clear problem focus, we ran a card sorting study to understand users' mental models. Participants organized 33 labels selected from level 4 of Amtrak's current site map within the PLAN category, some relabeled to better reflect their actual content, and with new labels added to cover content areas not represented in the original site map. Participants grouped them however made sense to them, then named each group.The card sorting results directly informed how we regrouped content and structured the proposed site map that would go into tree testing.
Insights
1.
Two distinct mental models emerged.
Some users organized content by when it happens in a trip (before, during, after). Others organized by function. "Onboard Dining Options" landed under "Onboard the Train" for some and "Food and Dining" for others, reflecting genuinely different ways of thinking about the same content.
2.
Planning, booking, and ticket info blurred together.
Some participants viewed planning as part of booking, while others grouped all ticket-related items together regardless of whether they required action or were purely informational. This pointed to a need to more clearly separate transactional content from informational content.
3.
Tickets and fares felt separate from trip planning.
Most participants did not group booking-related content alongside pre-trip information, suggesting Tickets and Fares deserved its own navigation category.
4.
Accessibility content was personal, not categorical.
Participants placed accessibility services under vastly different groups depending on personal relevance, ranging from Accessible Travel to Help and Info to Policies and Regulations. This pointed to a need for a more prominent, dedicated category.
5.
Some content required multiple entry points to serve different mental models
Labels like "Bringing Your Bicycle" and "Reporting Lost Items" split roughly evenly across Baggage, FAQs, and Policy groups. This was a signal that multiple entry points would be needed for edge-case content.
/ Step 4: Tree Testing
Testing whether users could find what they needed in the new structure
Armed with card sorting insights, we built a proposed site map and tested it using tree testing. The goal was to validate whether the restructured categories and labels actually helped users find content faster and more directly. Participants were given 8 tasks and asked to navigate a simplified, text-only version of the restructured site map. We measured success rate, directness, and time to evaluate findability.Overall, the restructured site map performed well for most tasks, confirming that organizing content by trip phase matched how users naturally navigate. But the results also revealed where the structure still had gaps.
Insights
1.
Users expected to navigate by "when," not "what."
All participants sought information based on when it happens during their journey, reinforcing the case for trip-phase grouping.
2.
Schedule was expected to live inside Plan, not as a standalone item.
Participants found it odd that Schedules existed as a separate navigation item, associating it with pre-trip planning.
3.
Some content needed multiple entry points.
Bicycle policy, lost items, and station amenities produced split navigation behavior, with users approaching from different starting points depending on their mental model.
4.
Booking-related widgets were misinterpreted as content pages.
Many participants expected to find information under Book, Train Status, and My Trip. But these are widgets, not content pages, and the current navigation gave no indication of that distinction.
5.
Previewing more content improved navigation confidence significantly.
Most participants expressed visible satisfaction when they realized the navigation branched further than the initial options. Once they discovered the subcategories under Plan A Trip, their ability to navigate and explore improved drastically.
What the research added up to:
Seven IA Decisions That Rebuilt the Navigation
Seven IA Decisions That Rebuilt the Navigation
Four methods, 20 participants, one core insight: the structure was the problem, not the content. Users weren't failing because information was missing. They were failing because the categories, labels, and hierarchy didn't match how they thought about traveling. With that understanding, we had everything we needed to make our first round of redesign decisions. We translated the findings into seven specific information architecture decisions, each one grounded in what users showed us.
Reorganized PLAN by trip phases
Users expected information organized by when it happens during their trip, not by topic category. We restructured the PLAN mega menu around three trip phases: Before You Go, At the Station, and Onboard the Train.
Leveled up content pages in PLAN
Accessibility and Special Assistance, At the Station, and Special Items were moved up in the content hierarchy so they became directly accessible subcategories in the mega menu, reducing the number of clicks required to reach them.
Regrouped support-related content
Users consistently pushed support-related information into its own category, separate from trip planning content. We combined the Help and Support category with the existing Need Help? page to give support-related content a dedicated, clearly labeled home in the navigation.
Separated Tickets and Fares from PLAN
Most participants felt tickets and fares were distinct from trip planning content. Tickets and Fares became its own top-level navigation category, making transactional content clearly separate from informational browsing.
Moved Schedules under PLAN
Participants consistently considered finding a schedule part of pre-trip planning, not a standalone task. The Schedules widget was moved under PLAN and grouped with Plan Your Trip content.
Multiple paths for different mental models
Certain content like bicycle policy, lost items, and station amenities produced genuinely split navigation behavior. Rather than forcing a single path, we planned to provide multiple entry points to accommodate users who approach the same content from different starting points.
Regrouped booking-related widgets under a newly added Book button
Participants were confused when action-oriented tools appeared mixed with informational navigation links. We regrouped all booking-related widgets under one call-to-action button: Book, making it immediately clear what is a tool and what is content.
We Thought We Had It Figured Out.
Our Peers Disagreed.
Our Peers Disagreed.
Before building the low-fi prototype, we presented the proposed site map and IA decisions to peers and our professor for a design critique. The feedback pushed us to reconsider several decisions and surfaced new opportunities we hadn't fully addressed.
Navigation structure needed adjustment
“Manage Trip” was suggested as a candidate for inclusion directly in the navigation bar. “Tickets and Fares” fell outside our project scope and should be removed.
Labeling and widget clarity needed work
Special Items needed a clearer label such as "Restricted" or "Irregular" to better communicate what kinds of items it covers. "Find Trip" was being read as "Manage Trip," signaling that the label needed to more clearly communicate what the widget actually does.
The mega menu needed to reduce cognitive load
The mega menu felt visually overwhelming with too much content at once, signaling a need to reorganize and simplify content within each phase group.
Mixing tools and booking actions created confusion
“Train Status” and “Find Trip” didn't feel like they belonged alongside “Book”. Grouping them together made it harder to distinguish between starting a new booking and quickly accessing trip tools.
Putting the Redesign in Users' Hands to See if It Worked
With the seven IA decisions defined, we built a paper prototype covering 15 screens across two user flows. We chose these two tasks because they represent distinct interaction modes: one tests the site's information architecture, the other tests its transactional workflow.
Task 1: Finding baggage policy information without a booked trip
Task 2: Rescheduling a booked trip as a guest user with no Amtrak account
Both tasks were completed by all 4 participants. But the sessions surfaced specific friction points within each task.
Task 1: Finding baggage policy information without a booked trip
Task 2: Rescheduling a booked trip as a guest user with no Amtrak account
Both tasks were completed by all 4 participants. But the sessions surfaced specific friction points within each task.
/ Task 1: Finding baggage information
No participant found baggage information on their first attempt. The path to it was indirect and unintuitive every time.
The Booking Widget Dominates Attention
The widget's size dominated visual attention, pulling users away from other navigation paths. 3 of 4 participants started with the homepage booking widget and missed the navigation bar entirely.
"Plan" Doesn't Signal General Information
Users expected "Plan" to be about booking, not general travel information. 4 of 4 participants did not associate "Plan" with baggage information.
Mega Menu Overwhelm
The mega menu felt overwhelming once opened. Participants who did find their way to the "Plan" mega menu struggled to scan it quickly due to the volume of nested options.
/ Task 2: Rescheduling a booked trip
All participants correctly identified "Manage Trip" as the starting point, confirming the navigation structure worked for this task. The issues surfaced once they started filling in the reservation lookup form.
"Search By" Dropdown Confusion
The default state didn't communicate that a selection was required. 2 of 4 participants missed the "Search By" dropdown and clicked "Find Trip" without selecting a search method, triggering an error.
Duplicate "Manage Trip" Entry Points
"Manage Trip" appeared in two places on the homepage and are visually close on the page, which caused the confusion. 2 of 4 participants were unsure whether the one in the navigation bar and the one in the booking widget led to the same place.
"Date & Time" vs. Action-Oriented Labels
1 participant found "Date & Time" inconsistent with the other two labels. "Cancel" and "Upgrade" are action-oriented, while "Date & Time" is descriptive. Participant expected to see a label like "Reschedule" to match the action-oriented pattern.
From What Broke to What We Fixed: The Thinking Behind the Mid-Fidelity Prototype
Each finding from the low-fi usability testing and design critique drove a specific set of changes in the mid-fidelity prototype. Here are what changed and why.
1.
Restructuring the navigation bar into two levels: separating browsing from actions
A single cluttered row mixing tools, information, and utilities with no visual hierarchy
Everything lived in one row. Users couldn't distinguish between tools and informational links, and the booking widget overpowered the navigation bar entirely.
A two-level layout separating browsing content on the left from quick-access trip tools on the right
A top utility row holds Need Help?, language selector, and Log In. The main row places Travel Info, Deals, and Guest Rewards on the left for information browsing, and Train Status, Manage Trip, and Book on the right for quick action.
2.
Relabeling key navigation items to match user expectations
Vague labels like "Plan" and "My Trip," and a login hidden behind "Guest Rewards"
"Plan" was associated with booking, not information. "My Trip" didn't clearly communicate post-booking actions. "Guest Rewards" was used as the login entry point, which confused users expecting a standard Log In.
"Travel Info" for browsing, "Manage Trip" for post-booking actions, and "Log In" where users expect it
"Travel Info" signals general trip information accessible before booking. "Manage Trip" makes post-booking actions explicit. "Log In" sits in the utility row where users expect it, and Book is styled as a distinct call-to-action button consistent with travel industry standards.
3.
Restructuring the mega menu from functional categories to trip phases users actually navigate by
Broad functional categories that mixed pre-departure decisions with in-journey logistics
Content was grouped by department, not by journey. Users had to browse multiple unrelated categories before landing on what they needed.
Three trip phases organized around when users need information
Before You Go, At the Station, and Onboard the Train give users a clear mental map of the menu. Accessibility and Special Assistance was given its own category, no longer buried inside the planning section.
4.
Restructuring the booking widget to consolidate all booking-related actions in one place
Trip type selections hidden in a nav bar dropdown, static date fields, and a misplaced passenger count
To select a trip type, users had to interact with the navigation bar dropdown before the widget. The Return Date appeared even for one-way trips, and the Passenger count sat on a separate line labeled "Traveler."
All booking selections consolidated on the widget with dynamic fields and a relabeled passenger count
Trip types, date fields, and passenger count now live on the widget. The Return Date disappears for one-way trips. Manage Trip and Train Status are added as tabs, giving users one place for all three core functions.
5.
Redesigning Need Help? from a Dead End into a Full Support Hub
A landing page with sidebar links but no useful content in the main area
Users who got stuck had nowhere useful to go. The page offered no browsable content and no clear path to answers.
A full support hub where every option is visible and one click away
Ask Julie sits at the top for quick answers. Category cards for Frequently Asked Questions, Chat Live Agent, Email Us, and Call/Write Us give users an immediate overview of all support options.
6.
Replacing the hidden "Search By" dropdown with visible radio buttons to eliminate form submission errors
"Search By" dropdown so easy to miss that users submitted without selecting anything and causing errors
Participants consistently submitted the form without selecting a search method, triggering an error with no clear indication of what went wrong.
All lookup options visible upfront as radio buttons, reducing clicks and eliminating the hidden step
All lookup options are immediately visible. The default selection reduces the number of steps needed to complete the form.
7.
Renaming "Date and Time" to "Reschedule" to match the action-oriented pattern of all trip action buttons
Inconsistent button labels mixing descriptive "Date and Time" with action verbs "Cancel" and "Upgrade"
"Date and Time" described what was being modified, not what action was being taken, breaking the pattern set by the other two buttons.
Consistent action-oriented labels with "Reschedule" replacing "Date and Time"
All three buttons now follow the same verb-first pattern, making each action immediately clear.
Honest Takeaways from a Project That Pushed My Thinking Further Than Expected
working rhythm as a team
What I Learned
Every research method taught me something the others couldn't. Interviews surfaced the frustration. Card sorting revealed how differently people think about the same content. Tree testing showed whether our structure actually held up. And usability testing humbled us: even well-researched decisions still surprised us when real users got their hands on them. I came away with a much deeper respect for running the full research cycle rather than skipping steps.
I also didn't expect IA design to be this complex. Moving content around sounds simple. But every decision had consequences for the rest of the structure, and there was rarely one right answer. Learning to make deliberate tradeoffs and defend them with evidence was one of the hardest and most valuable parts of this project.
What I Would Have Done Differently
I wish we had involved Amtrak's team earlier as co-designers. Their internal knowledge of why the current structure exists could have saved us time and made our solutions more realistic.
I would also have pushed to prototype and iterate faster rather than perfecting each phase before moving forward. And I am curious what AI-assisted prototyping could have added to our early structural explorations, something I want to experiment with in future projects.
What's Next
For me, this project opened up a genuine interest in transportation and public service design. My next step is taking the redesign further into high-fidelity mockups and interactive prototypes to bring it fully to life.
I also want to explore how AI can accelerate this process, from generating design variations to building out prototype interactions faster, and see how an AI-assisted workflow could change the way I approach high-fidelity design work. Beyond this project, I want to keep learning about what it means to design for large-scale public systems where good navigation isn't just a usability improvement, it affects how millions of people experience getting from place to place.
I also want to explore how AI can accelerate this process, from generating design variations to building out prototype interactions faster, and see how an AI-assisted workflow could change the way I approach high-fidelity design work. Beyond this project, I want to keep learning about what it means to design for large-scale public systems where good navigation isn't just a usability improvement, it affects how millions of people experience getting from place to place.