
Farbound
Farbound
AI-Powered Accessible Travel Platform
Booking that proves the trip will fit before you fly.
Farbound is a booking platform for travelers with disabilities. Set your needs once and every hotel, flight, and activity is checked against them, with the evidence shown in writing.
Booking that proves the trip will fit before you fly.
Farbound is a booking platform for travelers with disabilities. Set your needs once and every hotel, flight, and activity is checked against them, with the evidence shown in writing.


Role
Role
Role
End-to-end product design.
Research, strategy, naming, brand, design system, 50+ screen prototype.
End-to-end product design.
Research, strategy, naming, brand,
design system, 50+ screen prototype.
Context
Context
Context
Independent product case study solo · 2026
Independent product case study
solo · 2026
Type
Type
Type
Mobile app and responsive web
Mobile app and responsive web
Tools
Tools
Tools
Figma, FigJam, design tokens, AI tools (Claude)
Figma, FigJam, design tokens,
AI tools (Claude)
AI-Powered Accessible Travel Platform
THE SCALE OF THE PROBLEM
Travel fails disabled travelers at a scale that is hard to look at directly
Travel fails disabled travelers at a scale that is hard to look at directly
What struck me reading through the research is that these numbers describe a market actively spending money on travel while being systematically failed at it. The gap between $58 billion in annual spending and a 54% arrival mismatch rate is where this project lives.
What struck me reading through the research is that these numbers describe a market actively spending money on travel while being systematically failed at it. The gap between $58 billion in annual spending and a 54% arrival mismatch rate is where this project lives.
96%
96%
96%
of wheelchair users hit
accommodation problems
of wheelchair users hit
accommodation problems
of wheelchair users hit
accommodation problems
54%
54%
54%
arrive to a room that does
not match the booking
arrive to a room that does
not match the booking
arrive to a room that does
not match the booking
4 in 10
4 in 10
4 in 10
mobility aids damaged
in airline transit
mobility aids damaged
in airline transit
mobility aids damaged
in airline transit
$58B
$58B
$58B
spent yearly on travel by US mobility-disabled travelers
spent yearly on travel by US mobility-disabled travelers
spent yearly on travel by US mobility-disabled travelers
CONTEXT
CONTEXT
How booking works today
How booking works today
Listings say "accessible" and mean nothing specific. A roll-in shower, a bed at the right height, a quiet room away from the elevator — none of these are guaranteed by checking a filter. So travelers do the verification themselves: they call ahead, get a verbal promise, keep no record of it, and still arrive to rooms that do not match. This is treated as a normal part of traveling with a disability, and the industry has not been particularly motivated to change it.
Listings say "accessible" and mean nothing specific. A roll-in shower, a bed at the right height, a quiet room away from the elevator — none of these are guaranteed by checking a filter. So travelers do the verification themselves: they call ahead, get a verbal promise, keep no record of it, and still arrive to rooms that do not match. This is treated as a normal part of traveling with a disability, and the industry has not been particularly motivated to change it.
Unverified listings
Unverified listings
No evidence, no recourse
No evidence, no recourse
Caregivers ignored
Caregivers ignored
Nothing for sensory needs
Nothing for sensory needs
I looked at six competitors across specialists, rentals, maps, and mainstream online travel agents (OTAs) and found one pattern across all of them. The gap between what travelers need to know and what any platform actually confirms is where the whole market fails. Nobody keeps evidence, nobody serves sensory or cognitive needs, and nobody treats caregivers as users.
I looked at six competitors across specialists, rentals, maps, and mainstream online travel agents (OTAs) and found one pattern across all of them. The gap between what travelers need to know and what any platform actually confirms is where the whole market fails. Nobody keeps evidence, nobody serves sensory or cognitive needs, and nobody treats caregivers as users.
6 PLATFORMS AND WHAT THEY MISS
Wheel the World
Coverage limited to their verified inventory. Strong on mobility, thinner on sensory and cognitive needs. No full trip planning.
AccessibleGO
Hotel-centric. Little support for the full trip — flights, transit, activities, day-by-day planning are not covered.
Becoming Rentable
Rentals only, small inventory. US-focused.
No trip planning layer beyond the accommodation itself.
Wheelmap + AccessNow
Point-of-interest ratings only — not planning or booking. Data freshness and depth vary significantly by city.
Expedia, Booking, Airbnb
Filters are host-declared and unverified. A query can return 95,000 results that travelers must verify one by one. The ADA room guarantee sits with the hotel, not the platform.
Google Maps
Mostly binary and entrance-level. Nothing about rooms, bathrooms, sensory environments, or cognitive load.
Not a booking tool.
PLATFORM
PLATFORM
TYPE
TYPE
WHAT THEY DO
WHAT THEY DO
WHAT THEY MISS
WHAT THEY MISS
Wheel the World
Wheel the World
Specialist
Specialist
Verified accessible booking for homes, experiences, and multi-day trips. Deep data on the properties they cover.
Verified accessible booking for homes, experiences, and multi-day trips. Deep data on the properties they cover.
Coverage limited to their verified inventory. Strong on mobility, thinner on sensory and cognitive needs. No full trip planning.
Coverage limited to their verified inventory. Strong on mobility, thinner on sensory and cognitive needs. No full trip planning.
AccessibleGO
AccessibleGO
Specialist
Specialist
Human-backed confirmation service: a team member calls each vendor before booking to confirm accessibility.
Human-backed confirmation service: a team member calls each vendor before booking to confirm accessibility.
Hotel-centric. Little support for the full trip — flights, transit, activities, day-by-day planning are not covered.
Hotel-centric. Little support for the full trip — flights, transit, activities, day-by-day planning are not covered.
Becoming Rentable
Becoming Rentable
Specialist
Specialist
Vacation rentals filtered by 43 adaptive elements with room-by-room checklists and a certification badge for verified properties.
Vacation rentals filtered by 43 adaptive elements with room-by-room checklists and a certification badge for verified properties.
Rentals only, small inventory. US-focused. No trip planning layer beyond the accommodation itself.
Rentals only, small inventory. US-focused. No trip planning layer beyond the accommodation itself.
Wheelmap + AccessNow
Wheelmap + AccessNow
Crowdsourced maps
Crowdsourced maps
2M+ rated places via Wheelmap, 30k+ places with photos via AccessNow. Active community of disabled travelers adding data.
2M+ rated places via Wheelmap, 30k+ places with photos via AccessNow. Active community of disabled travelers adding data.
Point-of-interest ratings only — not planning or booking. Data freshness and depth vary significantly by city.
Point-of-interest ratings only — not planning or booking. Data freshness and depth vary significantly by city.
Expedia, Booking, Airbnb
Expedia, Booking, Airbnb
Mainstream OTAs
Mainstream OTAs
Massive inventory and price competition. Accessibility filters exist across all three.
Massive inventory and price competition. Accessibility filters exist across all three.
Filters are host-declared and unverified. A query can return 95,000 results that travelers must verify one by one. The ADA room guarantee sits with the hotel, not the platform.
Filters are host-declared and unverified. A query can return 95,000 results that travelers must verify one by one. The ADA room guarantee sits with the hotel, not the platform.
Google Maps
Google Maps
General platform
General platform
Accessibility features integrated into global places data at enormous scale — entrance-level info for most locations.
Accessibility features integrated into global places data at enormous scale — entrance-level info for most locations.
Mostly binary and entrance-level. Nothing about rooms, bathrooms, sensory environments, or cognitive load. Not a booking tool.
Mostly binary and entrance-level. Nothing about rooms, bathrooms, sensory environments, or cognitive load. Not a booking tool.
THE QUESTION THAT FRAMED THE WORK
THE QUESTION THAT FRAMED
THE WORK
How might we let travelers verify their needs are met before they book, instead of after they arrive?
How might we let travelers verify their needs are met before they book, instead of after they arrive?
THE CHALLENGE
THE CHALLENGE
Moving verification from something travelers do, to something the platform does
Moving verification from something travelers do, to something the platform does
A search tool with accessibility filters, where the traveler calls ahead, carries the risk, and has no recourse if the room does not match.
→
→
A system where your needs are set once, every result is pre-checked against them, every claim carries its source, and the platform holds the evidence if something goes wrong.
THE SOLUTION
THE SOLUTION
The profile is the filter. The evidence is the product.
The profile is the filter. The evidence is the product.
The central idea is that your profile replaces the filter panel. You describe your needs once, in plain language, and every result you see has already been checked against them. An AI agent handles the verification work you would otherwise do by phone, and saves each written reply from a property as evidence attached to the booking.
The central idea is that your profile replaces the filter panel. You describe your needs once, in plain language, and every result you see has already been checked against them. An AI agent handles the verification work you would otherwise do by phone, and saves each written reply from a property as evidence attached to the booking.
If a listing cannot prove it fits you, Farbound says so rather than surfacing it anyway.
If a listing cannot prove it fits you, Farbound says so rather than surfacing it anyway.

Fit, not filters
Fit, not filters
Nothing hidden, everything explained
Nothing hidden, everything explained
Results are checked against your profile before you see them, and what does not fit stays on the page with its reason rather than disappearing. I decided early that hiding failures would repeat the same problem the product is trying to solve:
giving travelers results that look complete but are not.
Results are checked against your profile before you see them, and what does not fit stays on the page with its reason rather than disappearing. I decided early that hiding failures would repeat the same problem the product is trying to solve:
giving travelers results that look complete but
are not.


The AI agent
The AI agent
The verification work, done for you
The verification work, done for you
Ask the agent to confirm a detail and it contacts the property, then saves the written reply word for word with a date and reply ID. That reply is not a chat summary; it is a verbatim record, and if the room at check-in does not match what was confirmed, it is already attached to your booking as evidence.
Ask the agent to confirm a detail and it contacts the property, then saves the written reply word for word with a date and reply ID. That reply is not a chat summary; it is a verbatim record, and if the room at check-in does not match what was confirmed, it is already attached to your booking as evidence.
The worst day
The worst day
Designed for arrival mismatches
Designed for arrival mismatches
The 54% arrival mismatch statistic was the clearest brief I had. I wanted a traveler who checks in to a wrong room to have recourse without needing to argue.
A thirty-second room check at arrival: if something does not match, the report attaches your written confirmation automatically, notifies the hotel and support at the same moment, and a verified backup room is already on hold. The confirmation does the work that would otherwise fall to the traveler.
The 54% arrival mismatch statistic was the clearest brief I had. I wanted a traveler who checks in to a wrong room to have recourse without needing to argue.
A thirty-second room check at arrival: if something does not match, the report attaches your written confirmation automatically, notifies the hotel and support at the same moment, and a verified backup room is already on hold. The confirmation does the work that would otherwise fall to the traveler.

THE HARD DESIGN DECISION
THE HARD DESIGN DECISION
What the AI agent is and is not allowed to say
What the AI agent is and is not allowed to say
THE CONSTRAINT IS THE FEATURE
There is a version of this product where the AI agent answers any question about any property, whether or not it has verified data to support the answer. That version demos well. But in this context a confident wrong answer about a roll-in shower is not just a bad user experience; it means a traveler arrives to a room that does not work for them, with no evidence and no recourse.
So I constrained the agent to answer only from verified data. Where the data is unverified, it says so and offers to go find out. The restraint is the feature, and it also happens to be what makes the product honest rather than just useful.
BUILDING BLOCKS
BUILDING BLOCKS
The components that carry the promise
The components that carry the promise
Fit badge
Fit badge
Three states: fits you, one thing to check, does not fit. Always an icon plus a label plus a background tint, never color alone, so the signal holds for colorblind users and in grayscale and in bright sunlight on a phone screen.
Three states: fits you, one thing to check, does not fit. Always an icon plus a label plus a background tint, never color alone, so the signal holds for colorblind users and in grayscale and in bright sunlight on a phone screen.
Evidence card
Evidence card
The property's written reply, quoted verbatim, with a date and a reply ID. This one component is what makes verification, the booking guarantee, and the mismatch report all possible. It is the record the whole trust model rests on.
The property's written reply, quoted verbatim, with a date and a reply ID. This one component is what makes verification, the booking guarantee, and the mismatch report all possible. It is the record the whole trust model rests on.
Exact-room booking
Exact-room booking
You book room 204, not a room type. The room leaves inventory the moment you reserve it, so what you verified is what you will arrive to. This is the structural answer to the 54% problem rather than a softer promise layered over the same underlying issue.
You book room 204, not a room type. The room leaves inventory the moment you reserve it, so what you verified is what you will arrive to. This is the structural answer to the 54% problem rather than a softer promise layered over the same underlying issue.
Verified badge
Verified badge
An apricot dot inside a plum outline, repurposed from a passport-stamp logo concept that did not survive the selection process. It marks data that a human has confirmed against a specific claim, as opposed to data that a listing asserted on its own.
An apricot dot inside a plum outline, repurposed from a passport-stamp logo concept that did not survive the selection process. It marks data that a human has confirmed against a specific claim, as opposed to data that a listing asserted on its own.
Managed profiles
Managed profiles
Caregivers can plan and book for someone else as a first-class flow, with the person being planned for able to view their own profile and take it over at 18. This needed to be in the architecture from the start rather than added as a workaround later.
Caregivers can plan and book for someone else as a first-class flow, with the person being planned for able to view their own profile and take it over at 18. This needed to be in the architecture from the start rather than added as a workaround later.
ICONS
ICONS
Apricot appears in exactly one icon, the gateway logo. The brand gesture stays reserved.
Apricot appears in exactly one icon, the gateway logo. The brand gesture stays reserved.
COMPONENTS · TOKEN-BOUND
COMPONENTS · TOKEN-BOUND
Label
Label
Label
Label
Fits you
Fits you
1 thing to check
1 thing to check
Doesn't fit
Doesn't fit
Verified by Farbound
Verified by Farbound
Need
Need
Need
Need
Eighteen icons on a 24px grid and five component sets, all bound to the semantic token system.
ICONS
Apricot appears in exactly one icon, the gateway logo. The brand gesture stays reserved.
COMPONENTS · TOKEN-BOUND
Label
Label
Fits you
1 thing to check
Doesn't fit
Verified by Farbound
Need
Need
Eighteen icons on a 24px grid and five component sets, all bound to the semantic token system.
THE SYSTEM
THE SYSTEM
Built to be legible
Built to be legible
I chose Atkinson Hyperlegible as the primary typeface because it was designed by the Braille Institute specifically for low-vision readers, with each character drawn to be uniquely distinguishable. That felt like the right starting point for an accessibility product. The type system has a 16px floor across all screens. Underneath, a two-layer token architecture — primitives mapped to semantic roles — makes the high contrast mode a real mode switch rather than a separate design: the same components re-theme themselves when you flip the setting, which is what makes the accessibility settings screen feel like architecture rather than decoration.
I chose Atkinson Hyperlegible as the primary typeface because it was designed by the Braille Institute specifically for low-vision readers, with each character drawn to be uniquely distinguishable. That felt like the right starting point for an accessibility product. The type system has a 16px floor across all screens. Underneath, a two-layer token architecture — primitives mapped to semantic roles — makes the high contrast mode a real mode switch rather than a separate design: the same components re-theme themselves when you flip the setting, which is what makes the accessibility settings screen feel like architecture rather than decoration.
farbound. design system
farbound. design system
farbound. design system
Semantic tokens · 2 modes · icon-paired signals · 16px floor
Semantic tokens · 2 modes · icon-paired signals · 16px floor
Semantic tokens · 2 modes · icon-paired signals · 16px floor
COLOR · PRIMITIVES → SEMANTIC ROLES
COLOR · PRIMITIVES → SEMANTIC ROLES
COLOR · PRIMITIVES → SEMANTIC ROLES
plum
plum
plum
plum
plum
plum
lilac
lilac
lilac
apricot
apricot
apricot
mist
mist
mist
ink
ink
ink
green
green
green
amber
amber
amber
red
red
red
brand/primary → plum/600 · High contrast → plum/800
brand/primary → plum/600 · High contrast → plum/800
brand/primary → plum/600 · High contrast → plum/800
surface/tint → lilac/100 · High contrast → white
surface/tint → lilac/100 · High contrast → white
surface/tint → lilac/100 · High contrast → white
signal/fits · signal/check · signal/nofit
signal/fits · signal/check · signal/nofit
signal/fits · signal/check · signal/nofit
TYPE · ATKINSON HYPERLEGIBLE + FRAUNCES DISPLAY
TYPE · ATKINSON HYPERLEGIBLE + FRAUNCES DISPLAY
TYPE · ATKINSON HYPERLEGIBLE + FRAUNCES DISPLAY
Go far. Arrive sure.
Go far. Arrive sure.
Go far. Arrive sure.
H1 · Atkinson Hyperlegible Next Bold 28
H1 · Atkinson Hyperlegible Next Bold 28
H1 · Atkinson Hyperlegible Next Bold 28
Body · Atkinson Hyperlegible Next Regular 16 · Il1 0O rn m 8B3E
Body · Atkinson Hyperlegible Next Regular 16 · Il1 0O rn m 8B3E
Body · Atkinson Hyperlegible Next Regular 16 · Il1 0O rn m 8B3E
Designed by the Braille Institute for low vision. Every character uniquely identifiable. 16px body floor, buttons track at 3 percent.
Designed by the Braille Institute for low vision. Every character uniquely identifiable. 16px body floor, buttons track at 3 percent.
Designed by the Braille Institute for low vision. Every character uniquely identifiable. 16px body floor, buttons track at 3 percent.
LOGO · GATEWAY
LOGO · GATEWAY
LOGO · GATEWAY
farbound.
farbound.
farbound.
The apricot dot is the destination. It recurs as the wordmark period, so even text-only carries the idea.
The apricot dot is the destination. It recurs as the wordmark period, so even text-only carries the idea.
The apricot dot is the destination. It recurs as the wordmark period, so even text-only carries the idea.
ICONS
ICONS
ICONS
Apricot appears in exactly one icon, the gateway logo. The brand gesture stays reserved.
Apricot appears in exactly one icon, the gateway logo. The brand gesture stays reserved.
Apricot appears in exactly one icon, the gateway logo. The brand gesture stays reserved.
COMPONENTS · TOKEN-BOUND
COMPONENTS · TOKEN-BOUND
COMPONENTS · TOKEN-BOUND
Label
Label
Label
Label
Label
Label
Fits you
Fits you
Fits you
1 thing to check
1 thing to check
1 thing to check
Doesn't fit
Doesn't fit
Doesn't fit
Verified by Farbound
Verified by Farbound
Verified by Farbound
Need
Need
Need
Need
Need
Need
HIGH CONTRAST
HIGH CONTRAST
HIGH CONTRAST
Label
Label
Label
Fits you
Fits you
Fits you
1 thing to check
1 thing to check
1 thing to check
Verified by Farbound
Verified by Farbound
Verified by Farbound
Deeper colors, darker text, white surfaces.
Deeper colors, darker text, white surfaces.
Deeper colors, darker text, white surfaces.
The full design system: primitives to semantic roles, type and logo rules, the icon set, token-bound components, and the high contrast mode rendered from the same component instances.
Zoom in to see the full design system: primitives to semantic roles, type and logo rules, the icon set, token-bound components, and the high contrast mode rendered from the same component instances.
Seven colors with one job each: plum carries the brand identity, apricot is the destination dot, and three signal colors mean fits, check, and does not fit.
Seven colors with one job each: plum carries the brand identity, apricot is the destination dot, and three signal colors mean fits, check, and does not fit.
Gateway · the final mark
An open gate, and the traveler through it. One idea, chosen over four documented rounds of alternatives. The dot is the traveler and the destination at once, and it recurs as the wordmark's period.
Mark alone
App icon, avatars, watermarks.
farbound.
Full lockup
Marketing, splash, documents.
Reversed on plum · the dot stays apricot
Minimum sizes
96 · 60 · 40 · 28 · 20 · 16px. Mark holds to 20px.
Below that, the favicon rule applies.
Selection record
The logo went through four rounds and thirteen concepts before this one. The first round started with route-to-dot marks, until a check against the travel category found the same gesture used across too many existing products. The second round revised toward a letterform. The third started over with six new directions. Gateway — an open gate with the traveler through it — was chosen because the meaning is immediate and the shape is specific to this product. The two readings that came up in the audit were keyhole and doorway, both acceptable for an accessibility product about access.
RESEARCH
RESEARCH
The work behind the solution
The work behind the solution
15
15
15
flows mapped and built,
wireframe to hi-fi
flows mapped and built,
wireframe to hi-fi
flows mapped and built,
wireframe to hi-fi
50+
50+
50+
screens wired across
mobile and desktop
screens wired across
mobile and desktop
screens wired across
mobile and desktop
13
13
13
logo concepts across
four documented rounds
logo concepts across
four documented rounds
logo concepts across
four documented rounds
12
12
12
names through trademark
knockout screens
names through trademark
knockout screens
names through trademark
knockout screens
The research was secondary: market data, community forums, a competitive teardown, and a STEEP analysis.
What I was looking for was where existing products stop short and why.
What I was looking for was where existing products stop short and why.
The research was secondary: market data, community forums, a competitive teardown, and a STEEP analysis. What I was looking for was where existing products stop short and why.
The research was secondary: market data, community forums, a competitive teardown, and a STEEP analysis. What I was looking for was where existing products stop short and why.
Personas
Personas
M
M
Maya, 34
Maya, 34
Deep · Mobility
Deep · Mobility
Marketing manager, Chicago. Power wheelchair user since 19. Travels 6+ times a year. Plans with a spreadsheet.
Marketing manager, Chicago. Power wheelchair user since 19. Travels 6+ times a year. Plans with a spreadsheet.
"I don't need inspiration. I need to know the bed height before I fly 2,000 miles."
"I don't need inspiration. I need to know the bed height before I fly 2,000 miles."
GOALS
GOALS
Book a trip in one evening, not one week. Arrive knowing the room will work. Keep her $30k chair safe in transit.
Book a trip in one evening, not one week. Arrive knowing the room will work. Keep her $30k chair safe in transit.
FRUSTRATIONS
FRUSTRATIONS
"Accessible" means nothing specific. Photos never show the bathroom. Calls to verify leave no proof, and she still gets the wrong room.
"Accessible" means nothing specific. Photos never show the bathroom. Calls to verify leave no proof, and she still gets the wrong room.
NEEDS FROM FARBOUND
NEEDS FROM FARBOUND
A profile with her exact measurements and equipment. Verified fit signals on every result. Written confirmation from properties. Damage-claim support.
A profile with her exact measurements and equipment. Verified fit signals on every result. Written confirmation from properties. Damage-claim support.
D
D
Devon, 27
Devon, 27
Deep · Cognitive + sensory
Deep · Cognitive + sensory
Data analyst, Austin. Autistic. Loves travel, dreads unpredictability. Researches airport layouts on YouTube for hours.
Data analyst, Austin. Autistic. Loves travel, dreads unpredictability. Researches airport layouts on YouTube for hours.
"Nobody can tell me if the lobby is loud. That one unknown can end the whole trip."
"Nobody can tell me if the lobby is loud. That one unknown can end the whole trip."
GOALS
GOALS
Travel independently with a plan that holds. Know the sensory environment before arriving. Have a backup when plans break.
Travel independently with a plan that holds. Know the sensory environment before arriving. Have a backup when plans break.
FRUSTRATIONS
FRUSTRATIONS
Sensory information is nowhere: noise, lighting, crowds. Filters only cover wheelchairs. Surprises compound until the day collapses.
Sensory information is nowhere: noise, lighting, crowds. Filters only cover wheelchairs. Surprises compound until the day collapses.
NEEDS FROM FARBOUND
NEEDS FROM FARBOUND
Sensory detail per place: noise, light, crowd levels. Step-by-step day-of plans. Predictability signals and pre-planned alternatives.
Sensory detail per place: noise, light, crowd levels. Step-by-step day-of plans. Predictability signals and pre-planned alternatives.
R
R
Ruth, 68
Ruth, 68
Secondary · Aging + low vision
Secondary · Aging + low vision
Retired teacher, Sacramento. Arthritis and low vision. Would never call herself disabled, so she never taps an accessibility filter.
Retired teacher, Sacramento. Arthritis and low vision. Would never call herself disabled, so she never taps an accessibility filter.
"I just want a ground floor room and menus I can actually read."
"I just want a ground floor room and menus I can actually read."
DESIGN IMPLICATION
DESIGN IMPLICATION
Needs-based language, not identity-based. Large readable UI. Comfort framing: fewer stairs, shorter walks, good lighting. She is the demographic wave.
Needs-based language, not identity-based. Large readable UI. Comfort framing: fewer stairs, shorter walks, good lighting. She is the demographic wave.
S
S
Sam, 41
Sam, 41
Secondary · Caregiver
Secondary · Caregiver
Operations lead, Denver. Plans every trip for his son Eli, 12, who has a developmental disability, while coordinating a family of four.
Operations lead, Denver. Plans every trip for his son Eli, 12, who has a developmental disability, while coordinating a family of four.
"I'm booking for someone who isn't me. Every site assumes I'm booking for myself."
"I'm booking for someone who isn't me. Every site assumes I'm booking for myself."
DESIGN IMPLICATION
DESIGN IMPLICATION
Shared profiles and booking on behalf of someone else as first-class flows. Family trip-fit: results must work for Eli and for everyone.
Shared profiles and booking on behalf of someone else as first-class flows. Family trip-fit: results must work for Eli and for everyone.
Two deep personas, two secondary: Maya carries the verification burden: the 40 phone calls, the wrong room at check-in. Devon's barrier is unpredictability and not travel itself. Ruth never calls herself disabled, which rewrote the product's language. Sam books for his son, which rewrote its architecture.
Two deep personas, two secondary: Maya carries the verification burden: the 40 phone calls, the wrong room at check-in. Devon's barrier is unpredictability and not travel itself. Ruth never calls herself disabled, which rewrote the product's language. Sam books for his son, which rewrote its architecture.
Journey map · Maya
Journey map · Maya
Power wheelchair user. Worst moments: booking blind and the check-in lottery.
Power wheelchair user. Worst moments: booking blind and the check-in lottery.
01 Plan
01 Plan
DOING
DOING
Researches blogs and forums, builds a spreadsheet, cross-checks photos
Researches blogs and forums, builds a spreadsheet, cross-checks photos
PAIN
PAIN
"Accessible" is undefined. Photos never show bathrooms. Verification takes hours
"Accessible" is undefined. Photos never show bathrooms. Verification takes hours
OPPORTUNITY
OPPORTUNITY
Profile set once. Results pre-filtered to her measurements with verified photos
Profile set once. Results pre-filtered to her measurements with verified photos
02 Book
02 Book
DOING
DOING
Books hotel and flight, then calls to confirm the roll-in shower
Books hotel and flight, then calls to confirm the roll-in shower
PAIN
PAIN
OTA filters are self-reported. Calls leave no record. Guarantee sits with the hotel
OTA filters are self-reported. Calls leave no record. Guarantee sits with the hotel
OPPORTUNITY
OPPORTUNITY
AI agent confirms details in writing. Fit badge on every result
AI agent confirms details in writing. Fit badge on every result
03 Fly
03 Fly
DOING
DOING
Gate-checks her $30k power chair, requests an aisle chair
Gate-checks her $30k power chair, requests an aisle chair
PAIN
PAIN
4 in 10 chairs damaged or lost. Long assistance waits. High anxiety
4 in 10 chairs damaged or lost. Long assistance waits. High anxiety
OPPORTUNITY
OPPORTUNITY
Chair specs in profile. DOT rights card, day-of checklist, claim support
Chair specs in profile. DOT rights card, day-of checklist, claim support
04 Stay
04 Stay
DOING
DOING
Checks in, inspects the room before unpacking
Checks in, inspects the room before unpacking
PAIN
PAIN
54% get a room that does not match the booking. Beds too high
54% get a room that does not match the booking. Beds too high
OPPORTUNITY
OPPORTUNITY
Reconfirmation 48h out. In-app mismatch escalation and rebooking
Reconfirmation 48h out. In-app mismatch escalation and rebooking
05 Explore
05 Explore
DOING
DOING
Books tours and rideshares, maps curb cuts near the hotel
Books tours and rideshares, maps curb cuts near the hotel
PAIN
PAIN
79% hit in-market transport problems. Attraction info stale
79% hit in-market transport problems. Attraction info stale
OPPORTUNITY
OPPORTUNITY
Verified activities and accessible transit, kept fresh by community
Verified activities and accessible transit, kept fresh by community
Maya's journey, mapped across five stages. Every pain cell traces back to a statistic.
Journey map · Devon
Journey map · Devon
Autistic traveler. Worst moments: planning blind on sensory data and unpredictable travel days.
Autistic traveler. Worst moments: planning blind on sensory data and unpredictable travel days.
01 Plan
01 Plan
DOING
DOING
Watches YouTube walkthroughs of airports and hotels for hours
Watches YouTube walkthroughs of airports and hotels for hours
PAIN
PAIN
Sensory info is nowhere: noise, lighting, crowds unknowable in advance
Sensory info is nowhere: noise, lighting, crowds unknowable in advance
OPPORTUNITY
OPPORTUNITY
Sensory profile per place: noise, light, crowd ratings, photo walkthroughs
Sensory profile per place: noise, light, crowd ratings, photo walkthroughs
02 Book
02 Book
DOING
DOING
Picks flights by time of day to avoid crowds
Picks flights by time of day to avoid crowds
PAIN
PAIN
Filters cover wheelchairs, nothing for sensory or cognitive needs
Filters cover wheelchairs, nothing for sensory or cognitive needs
OPPORTUNITY
OPPORTUNITY
Sensory needs in the profile. Quiet hours and low-stimulation options surfaced
Sensory needs in the profile. Quiet hours and low-stimulation options surfaced
03 Fly
03 Fly
DOING
DOING
Arrives very early, noise-canceling headphones on
Arrives very early, noise-canceling headphones on
PAIN
PAIN
Security unpredictable. Audio-only announcements. Overload compounds
Security unpredictable. Audio-only announcements. Overload compounds
OPPORTUNITY
OPPORTUNITY
Step-by-step day-of plan: TSA Cares, gate walkthrough, visual updates
Step-by-step day-of plan: TSA Cares, gate walkthrough, visual updates
04 Stay
04 Stay
DOING
DOING
Settles in, builds a routine around the room
Settles in, builds a routine around the room
PAIN
PAIN
Alarms, hallway noise, fluorescent lobbies discovered on arrival
Alarms, hallway noise, fluorescent lobbies discovered on arrival
OPPORTUNITY
OPPORTUNITY
Room-level sensory detail verified. AI agent arranges quiet-floor requests
Room-level sensory detail verified. AI agent arranges quiet-floor requests
05 Explore
05 Explore
DOING
DOING
Plans one anchor activity per day
Plans one anchor activity per day
PAIN
PAIN
Crowd levels swing. One surprise collapses the day
Crowd levels swing. One surprise collapses the day
OPPORTUNITY
OPPORTUNITY
Crowd forecasts, buffer time, pre-planned backup for every anchor
Crowd forecasts, buffer time, pre-planned backup for every anchor
Devon's journey as an autistic traveler. His pain isn't the wrong room; it's not knowing what to expect at any stage. Mapping this is where the scope changed: the product needed to keep working after checkout, so I created an in-trip mode Farbound.
Beneficiary map · who Farbound serves
Beneficiary map · who Farbound serves
Designed for the center; useful outward. The curb-cut effect: what is essential for some is better for everyone. The center ring is the main reason for creating Farbound.
Designed for the center; useful outward. The curb-cut effect: what is essential for some is better for everyone. The center ring is the main reason for creating Farbound.
Disabled
travelers
Disabled
travelers
designed for, first
designed for, first
Temporary + situational
injury, recovery, pregnancy, chronic illness
injury, recovery, pregnancy, chronic illness
Caregivers, companions + aging travelers
planning for others, or for changing bodies
planning for others, or for changing bodies
Every traveler
verified truth, predictability, calmer trips
verified truth, predictability, calmer trips
User flow · end to end
Solid: primary path. Amber: mismatch branch. Dashed: the community data flywheel feeding back into search.
Solid: primary path. Amber: mismatch branch. Dashed: the community data flywheel feeding back into search.
Solid: primary path. Amber: mismatch branch. Dashed: the community data flywheel feeding back into search.
Splash
Splash
Splash
Onboarding
profile built
Onboarding
profile built
Onboarding
profile built
Home
Home
Home
Search results
fit-ranked
Search results
fit-ranked
Search results
fit-ranked
Listing
fit breakdown
Listing
fit breakdown
Listing
fit breakdown
AI agent
verifies
AI agent
verifies
AI agent
verifies
Written reply
attached
Written reply
attached
Written reply
attached
Book exact
room
Book exact
room
Book exact
room
Confirmed
48h recheck
Confirmed
48h recheck
Confirmed
48h recheck
Trip overview
Trip overview
Trip overview
Today plan
in-trip mode
Today plan
in-trip mode
Today plan
in-trip mode
Arrival check
Arrival check
Arrival check
Post-trip
confirmations
Post-trip
confirmations
Post-trip
confirmations
Data stays fresh
Data stays fresh
Data stays fresh
Report mismatch
evidence attached
Report mismatch
evidence attached
Report mismatch
evidence attached
Resolution
fix or rebook
Resolution
fix or rebook
Resolution
fix or rebook
End-to-end user flow across fifteen mapped paths. The branching path in amber is the arrival mismatch — the scenario where the room does not match the booking. The product needed an answer to that moment, not just the booking experience leading up to it.
Zoom in to see the end-to-end user flow across fifteen mapped paths. The branching path in amber is the arrival mismatch — the scenario where the room does not match the booking. The product needed an answer to that moment, not just the booking experience leading up to it.
Storyboard · The arrival that didn't go wrong
Storyboard of the arrival mismatch flow.
Panel four is the one that matters: the traveler has the written confirmation already, so they do not have to argue at the counter.
Panel four is the one that matters: the traveler has the written confirmation already, so they do not have to argue at the counter.
Storyboard of the arrival mismatch flow. Panel four is the one that matters: the traveler has the written confirmation already, so they do not have to argue at the counter.
Storyboard of the arrival mismatch flow.
Panel four is the one that matters: the traveler has the written confirmation already, so they do not have to argue at the counter.
Panel four is the one that matters: the traveler has the written confirmation already, so they do not have to argue at the counter.
1
1
1
Maya arrives at Hotel Aurora after a long flight. Room 204 is guaranteed: roll-in shower, confirmed in writing on May 12.
Maya arrives at Hotel Aurora after a long flight. Room 204 is guaranteed: roll-in shower, confirmed in writing on May 12.
Maya arrives at Hotel Aurora after a long flight. Room 204 is guaranteed: roll-in shower, confirmed in writing on May 12.
2
2
2
The bathroom has a tub. The 54% problem, live. The old script says: argue at the front desk, and probably lose.
The bathroom has a tub. The 54% problem, live. The old script says: argue at the front desk, and probably lose.
The bathroom has a tub. The 54% problem, live. The old script says: argue at the front desk, and probably lose.
3
3
3
Instead, she taps "Something doesn't match." The May 12 written confirmation attaches itself as evidence.
Instead, she taps "Something doesn't match." The May 12 written confirmation attaches itself as evidence.
Instead, she taps "Something doesn't match." The May 12 written confirmation attaches itself as evidence.
4
4
4
The phone does the arguing. The front desk and Farbound support read the same written promise at the same moment.
The phone does the arguing. The front desk and Farbound support read the same written promise at the same moment.
The phone does the arguing. The front desk and Farbound support read the same written promise at the same moment.
5
5
5
Room 312, with the verified shower, is ready by 4 pm. A nearby backup room was on hold the whole time, just in case.
Room 312, with the verified shower, is ready by 4 pm. A nearby backup room was on hold the whole time, just in case.
Room 312, with the verified shower, is ready by 4 pm. A nearby backup room was on hold the whole time, just in case.
6
6
6
Settled by sunset. Her report has already updated the listing for the next traveler.
Settled by sunset. Her report has already updated the listing for the next traveler.
Settled by sunset. Her report has already updated the listing for the next traveler.
Honesty note
Honesty note
This is design-stage architecture: services are real and chosen deliberately, but nothing is integrated. The AI answers only from the verified dataset; where data is unverified, the product says so instead of guessing. That rule is the difference between an AI feature and a safety hazard in this domain.
This is design-stage architecture: services are real and chosen deliberately, but nothing is integrated. The AI answers only from the verified dataset; where data is unverified, the product says so instead of guessing. That rule is the difference between an AI feature and a safety hazard in this domain.
System map · what Farbound runs on
The system map with real, named services.
The verified dataset and the written-confirmation store are what separate this from a booking tool with better filters.
The verified dataset and the written-confirmation store are what separate this from a booking tool with better filters.
The system map with real, named services. The verified dataset and the written-confirmation store are what separate this from a booking tool with better filters.
The system map with real, named services.
The verified dataset and the written-confirmation store are what separate this from a booking tool with better filters.
The verified dataset and the written-confirmation store are what separate this from a booking tool with better filters.
Accessibility data · the moat
Accessibility data · the moat
Farbound verified dataset (proprietary)
Farbound verified dataset (proprietary)
Measured room data: door widths, bed heights, roll-in showers, visual alarms, sensory ratings. Seeded by partner audits, kept fresh by community confirmations and mismatch reports.
Measured room data: door widths, bed heights, roll-in showers, visual alarms, sensory ratings. Seeded by partner audits, kept fresh by community confirmations and mismatch reports.
Written-confirmation store
Written-confirmation store
Every property reply saved verbatim with timestamp and reply ID. The evidence layer behind the guarantee.
Every property reply saved verbatim with timestamp and reply ID. The evidence layer behind the guarantee.
Google Places + OpenStreetMap / Wheelmap
Google Places + OpenStreetMap / Wheelmap
Baseline POI accessibility flags (step-free entrances) as cold-start data, never presented as verified.
Baseline POI accessibility flags (step-free entrances) as cold-start data, never presented as verified.
Transit accessibility feeds (GTFS + agency elevator status)
Transit accessibility feeds (GTFS + agency elevator status)
Real-time elevator and escalator outages, e.g. the MAAT alert in flow TR-03.
Real-time elevator and escalator outages, e.g. the MAAT alert in flow TR-03.
Inventory + booking
Inventory + booking
Amadeus / Sabre APIs
Amadeus / Sabre APIs
Hotel and flight inventory, rates, availability. Room-level booking where supported: the exact-room guarantee depends on it.
Hotel and flight inventory, rates, availability. Room-level booking where supported: the exact-room guarantee depends on it.
Duffel
Duffel
Modern flight booking API with seat and special-service requests (aisle chair, mobility aid in hold).
Modern flight booking API with seat and special-service requests (aisle chair, mobility aid in hold).
Stripe
Stripe
Payments, refunds, and the rebook-on-us flow.
Payments, refunds, and the rebook-on-us flow.
Twilio + SendGrid
Twilio + SendGrid
The agent's outreach channel: email and SMS to properties, and text-first communication for users like Amara.
The agent's outreach channel: email and SMS to properties, and text-first communication for users like Amara.
AI layer
AI layer
LLM API (agent)
LLM API (agent)
Drafts property outreach, parses replies into structured fields, interprets free-text profile input, and answers only from verified data with sources shown.
Drafts property outreach, parses replies into structured fields, interprets free-text profile input, and answers only from verified data with sources shown.
Speech + translation APIs
Speech + translation APIs
Text-to-speech and captioning for access; translation for international property replies.
Text-to-speech and captioning for access; translation for international property replies.
Crowd-level data (BestTime / Popular Times)
Crowd-level data (BestTime / Popular Times)
Crowd forecasts behind Devon's low-stimulation planning.
Crowd forecasts behind Devon's low-stimulation planning.
Vision API (photo intake)
Vision API (photo intake)
Auto face-blur on community photos and tape-measure reading assistance.
Auto face-blur on community photos and tape-measure reading assistance.
Platform
Platform
Passkeys / WebAuthn
Passkeys / WebAuthn
Passwordless-first auth: less typing, fewer barriers.
Passwordless-first auth: less typing, fewer barriers.
Mapbox
Mapbox
Maps with step-free routing overlays.
Maps with step-free routing overlays.
Push + email notifications
Push + email notifications
Verification replies, 48h reconfirmations, live trip alerts, always in the user's chosen channel.
Verification replies, 48h reconfirmations, live trip alerts, always in the user's chosen channel.
WCAG 2.1 AA web target
WCAG 2.1 AA web target
The EAA compliance baseline for the desktop product.
The EAA compliance baseline for the desktop product.
NAMING
NAMING
Twelve name candidates, eleven removed before subjective evaluation
Twelve name candidates, eleven removed before subjective evaluation
I ran every name through a trademark and domain knockout screen before subjective evaluation. Most of the early candidates failed on existing travel trademarks or category conflicts. Running the screen first meant the shortlist only contained names that could actually be used.
Farbound cleared the screen and worked on multiple levels: far as in the distance you want to travel, bound as in the direction you are heading, and bound also as in the barrier the product is designed to remove.
Farbound cleared the screen and worked on multiple levels: far as in the distance you want to travel, bound as in the direction you are heading, and bound also as in the barrier the product is designed to remove.
I ran every name through a trademark and domain knockout screen before subjective evaluation. Most of the early candidates failed on existing travel trademarks or category conflicts. Running the screen first meant the shortlist only contained names that could actually be used.
Farbound cleared the screen and worked on multiple levels: far as in the distance you want to travel, bound as in the direction you are heading, and bound also as in the barrier the product is designed to remove.
Farbound cleared the screen and worked on multiple levels: far as in the distance you want to travel, bound as in the direction you are heading, and bound also as in the barrier the product is designed to remove.
Atlas
Passage
Leeway
Everyway
Meander
Portico
Voyable
Wend
Aditus
Amparo
Farbound
Atlas
Passage
Leeway
Everyway
Meander
Portico
Voyable
Wend
Aditus
Amparo
Farbound
LOGO EXPLORATION
LOGO EXPLORATION
What got ruled out across four rounds
What got ruled out across four rounds
I screened thirteen concepts across four rounds. The first round produced a route-to-dot mark I had selected, and then a misreading audit caught a sun-over-horizon concept reading as a no-entry sign — which made misreading audits a required step for every concept after that. A category check then found that route-to-dot marks are generic across travel, so round two revised toward a letterform. Round three started fresh with six directions and a constraint: no pin, compass, globe, or airplane.
I screened thirteen concepts across four rounds. The first round produced a route-to-dot mark I had selected, and then a misreading audit caught a sun-over-horizon concept reading as a no-entry sign — which made misreading audits a required step for every concept after that. A category check then found that route-to-dot marks are generic across travel, so round two revised toward a letterform. Round three started fresh with six directions and a constraint: no pin, compass, globe, or airplane.
Logo Explorations
Logo Explorations
fb.
fb.
The route-to-dot marks in the first two rows were ruled out after a category audit found the gesture generic across travel. The arch concepts in the third row came from round three, which started fresh with a constraint against the four most saturated marks in travel.
The route-to-dot marks in the first two rows were ruled out after a category audit found the gesture generic across travel. The arch concepts in the third row came from round three, which started fresh with a constraint against the four most saturated marks in travel.
VALIDATION
VALIDATION
A test plan to run
A test plan to run
The next step is testing with the people this was designed for.
Twelve participants across five groups– mobility aid users, autistic adults, blind and low-vision users, Deaf and hard of hearing users, and caregivers– recruited through disability-led organizations and paid properly, not with gift cards. Sessions happen on participants' own devices with their own assistive technology, because testing on simulated setups tells you nothing. ASL interpretation and live captioning are booked in advance, not as an afterthought.
The core question is whether the trust the product promises actually lands. Does someone understand what verified means here, and would they book a trip without calling the hotel themselves first? When something in the product is confusing, that gets recorded as a design problem to iterate on and fix, not as the participant misunderstanding something.
REFLECTION
REFLECTION
What I'd carry forward
What I'd carry forward
Restraint as a design decision
Restraint as a design decision
The AI agent's constraint — answer only from verified data, say so when you cannot — ended up being more defensible than an agent that answers everything. Decisions about what the system will not do shaped the product as much as decisions about what it would.
The AI agent's constraint — answer only from verified data, say so when you cannot — ended up being more defensible than an agent that answers everything. Decisions about what the system will not do shaped the product as much as decisions about what it would.
Architecture reflects who the product is for
Architecture reflects who the product is for
Caregiver support, needs-based language, and high contrast all had to be in the architecture from the beginning. None of them could have been added later without changing the fundamental structure.
Caregiver support, needs-based language, and high contrast all had to be in the architecture from the beginning. None of them could have been added later without changing the fundamental structure.
Showing your work changes what the work is
Showing your work changes what the work is
The product and this case study follow the same principle: show the source, name the limits, and let the work carry the argument.
The product and this case study follow the same principle: show the source, name the limits, and let the work carry the argument.