Case Study UX Research Social Impact Accessibility Design

Designing Digital Equity for New York's 60,000 Homeless Individuals

As UX Researcher and Designer for EmPact, I led 50+ interviews with people experiencing homelessness, documented my own biases before starting research, watched each one get contradicted by actual user data, and delivered an iOS and web application that extended empathy through technology. Five weeks. Three design iterations. One product homeless New Yorkers could actually use.

50+

User interviews with homeless individuals across NYC

Design iterations, each driven by user feedback not assumptions

5 wks

Research to deployed product across three channels

Problem Statement

New York had the resources. 60,000 people experiencing homelessness could not find them. The technology existed. Nobody had built it for the actual user.

In 2018, NYC ranked first nationally with 14.23% of the U.S. homeless population and 60,000 people placed in shelters every night. The city had extensive resources: shelter networks, food banks, healthcare clinics, legal aid, job placement services. The information infrastructure was a fragmented landscape of siloed single-purpose applications, each designed for a different vertical, most requiring email addresses or desktop access, none considering mobile-first users without stable connectivity. The problem was not scarcity of services. It was that nobody had ever designed access to them for the people who needed access most.

Cognitive Overload by Fragmentation

311 for city services. DHS HELP NYC for shelter. NYC Well for mental health. Benefits.gov for federal programmes. Findhelp.org for local services. Each had its own interface, login requirements, and data format. For someone managing the daily stress of homelessness, navigating five incompatible systems was not an inconvenience. It was a barrier to access.

Designed for the Wrong User

Every existing solution assumed desktop access, reliable internet, an email address, and English literacy. None of those assumptions held for the actual user population. The fundamental design flaw was building for homeless people instead of with them, and never testing whether solutions worked in real-world conditions for the people they were supposed to serve.

Accessibility as an Afterthought

12px fonts on screens with cracked glass. No colour-blindness accommodation. No distance-by-foot metrics for users without transit funds. No Spanish-language support in a city where Spanish is the second most common language. Accessibility requirements were not an enhancement for this user group. They were the baseline minimum for equitable access.

Research and Design Process

Document your assumptions. Let research prove them wrong. Then build for what you found.

I documented six assumptions before starting research. Research contradicted every single one. Some homeless users earned up to $4,000 monthly. They paid for iPhone subscriptions. They researched shelter safety before visiting. They had savings accounts. They needed job placement and legal services.

WHAT 50+ USER INTERVIEWS REVEALED - EMPACT RESEARCH PHASE

🪞

Phase 1: Remove Yourself From the Process

Documented designer assumptions before any research: homeless people do not use mobile phones, they walk into any available shelter, they do not need legal or job placement services. Written down before a single interview. This created space to recognise when findings contradicted assumptions rather than searching for data that confirmed them.

🎤

Phase 2: 50+ Interviews Across New York City

Research covered daily routines, resource access patterns, mobile capabilities and data limitations, visual impairments, language preferences, literacy levels, mental health conditions affecting technology use, and trust factors with service providers. Every design requirement originated from what users described, not what designers assumed they needed.

🔁

Phase 3: Three Iterations, All User-Driven

First iteration: category block-frame. Users said something was off. Second iteration: simplified, larger tap areas. Users identified colour problems and remaining complexity. Third iteration: 16px fonts, accessibility palette for 7 colour blindness types, icon plus text labels, distance-by-foot geolocation, single-tap navigation. Each change responded to observed user behaviour, not reported preference.

♿

Phase 4: Accessibility as the Foundation

Colour-blind safe palette covering deuteranopia, protanopia, tritanopia, and four additional variants. 16px minimum font for poor eyesight without corrective lenses. Icon plus text label system for users with mental illness affecting complex interface processing. Spanish support throughout. Distance-by-foot metrics using geolocation for users who could not afford transportation.

Implementation

Five weeks. Three channels. Eight people.

One UX Researcher and Designer. Five web developers. Two iOS developers. Research defined exact requirements before development began, eliminating rework cycles. All three deployment channels delivered simultaneously in five weeks from research start to production.

▶ Version 1 Demo ▶ Version 2 Demo

Week 1

Assumptions Documentation and Research Protocol

Documented all designer assumptions in writing. Designed the 50+ interview protocol. Identified research locations across New York City. Established ethical framework for research with a vulnerable population before any fieldwork began.

Weeks 2 to 3

Field Research and Insight Synthesis

Conducted 50+ interviews across New York City. Documented where research contradicted assumptions. Synthesised findings into specific design requirements: font size, colour palette constraints, navigation complexity limits, language requirements, distance display format. Every requirement had a source interview.

Weeks 3 to 4

Three Design Iterations with Real Users

Rapid iteration cycles tested in real conditions: outdoor lighting, limited data plans, varying phone models. Each iteration addressed specific friction points observed in the previous round. Third iteration achieved positive feedback across all tested accessibility dimensions including colour blindness and font size.

Week 5

Three-Channel Deployment

Native iOS app, responsive web application, and e-kiosk interface deployed simultaneously. E-kiosks placed at strategic locations where homeless individuals congregated. All three channels validated against accessibility requirements with users who had visual impairments and colour blindness.

Final Outcomes

Measured by whether users could accomplish their goals in real-world conditions.

Qualitative validation through moderated testing with actual homeless users in real conditions. Quantitative delivery metrics across five weeks from research start to production deployment.

"You can always get around on foot. Good shelter is not always guaranteed. I love that I can know the distance by foot from my present location."

EMAPACT USER - NYC

"Grateful that you considered people like me who are color blind."

EMAPACT USER - NYC

"Overall, easy to follow and not too overwhelming."

EMAPACT USER - NYC

"The design is more inviting and relaxing."

EMAPACT USER - NYC

3 channels

iOS app, responsive web, and e-kiosk all delivered in 5 weeks from research start to production

7 types

Colour blindness variants accommodated in the accessibility palette across all three delivery channels

Bilingual

English and Spanish throughout. A legal and civil rights necessity in NYC, not a product enhancement.

Lessons Learned

Principles from social impact design that transfer to every product context.

Assumptions Are Design Inputs, Not Truths

Every designer carries assumptions about their users. The discipline is making them explicit before research begins so you can recognise when findings contradict them. I assumed homeless people did not use smartphones. Research showed some earned enough to pay for iPhone subscriptions. That contradiction reshaped the entire product direction.

Accessibility Benefits Everyone

Features designed for users with disabilities: larger fonts, icon plus text labels, simplified navigation, colour-blind palettes improved usability for all users, not just those with impairments. Designing for the hardest constraint produces the most durable solution. Accessibility-first is not a constraint. It is a quality standard.

Context Matters More Than Features

Distance-by-foot metrics mattered more than rich multimedia or social features. Real-world constraints: limited data plans, outdoor use, visual impairments, shaped what was genuinely useful. The simplest, most direct information architecture won every user test. Restraint is a design skill.

Technology Access Is Not Technology Literacy

Some homeless users had iPhones but limited exposure to complex applications. Intuitive design is not about simplifying content. It is about removing unnecessary friction so people can accomplish tasks efficiently. Fewer steps, tested in real conditions, with real users, in the environment where they will actually use the product.

Designing Digital Equity for New York's 60,000 Homeless Individuals

New York had the resources. 60,000 people experiencing homelessness could not find them. The technology existed. Nobody had built it for the actual user.

Document your assumptions. Let research prove them wrong. Then build for what you found.

Five weeks. Three channels. Eight people.

Measured by whether users could accomplish their goals in real-world conditions.

Principles from social impact design that transfer to every product context.

If you are building something that cannot afford to fail, let's talk.