Researching behavioral patterns of elderly federal employees to fix accessibility challenges on BENEFEDS.gov
Research & Design at FedPoint · Resolving WCAG 2.2 violations for task success rate + error rate
In this Project
Confidentiality notice
To comply with strict non-disclosure agreements, all proprietary data, live interfaces, and specific workflows have been omitted. As the primary contributor and owner of this project at FedPoint, I outline what I did, how I did it and the exact steps I followed to successfully solve the challenges in a set time frame. To know more about the research data, design wireframes and the resulting business impact, please connect.
Overview
What is BENEFEDS.gov?
BENEFEDS.gov is the official secure online portal used by federal employees, retirees, and uniformed services members to enroll in and manage FEDVIP dental and vision plans, as well as pay premiums for Federal Long Term Care Insurance (FLTCIP). It simplifies benefits administration by automating premium payments, often through payroll deductions.
The Problem
BENEFEDS.gov's enrollment tool contained critical accessibility violations and UX errors, that prevented a significant portion of federal employees, many of them elderly or assistive-technology-dependent from completing benefits enrollment independently.
My Role
I led the UX investigation to identify the violations/UX errors, and provided accessibility fixes that turned performance around. Performed research, interviewed users, identified the accessibility violations and eventually, re-designed the tool with better accessibility and information clarity.
Team
Ella Manzelli (Research), Andrew Zarbo (Design), James Roth (Team Lead), Ali Shah (Engineering), Harshveen (UX Lead)
My Work Process
ISO 9241-210 compliant HCD, widely used in enterprise &government SDLC
A simple yet powerful framework which gave me me clear direction based on what I already knew, what needed to be uncovered, and what came next. Over the years, my work process has evolved to be a mix of agile and double diamond but today it adapts to the context and whatever best fits the problem at hand. For this one, Human-centered Design made more sense.
Human-centered Design
- Research
- Define
- Design
- Deliver
Proxy Interviews
Persona Identification
Moderated Interviews
Heuristic Evaluation
User-specific analytics
Feature Breakdowns
Impact Effort Matrix
Sprint Planning
Dev-handoff Plan
Prototyping
Wires
Responsive Adaptations
Affinity Mapping
Journey Map
Problem Statement
fig, Activities performed at each stage of the work process.
Research
1. Understanding the target persona.
Internal survey data from 2021 revealed that the average age of BENEFEDS user is mid-40s - elderly civilians.So my first step was to understand how users in this age group think, feel, and act on the web, and simultaneously add those relevant attributes to the persona. Eventually I understand and solve challenges for this persona, I solve challenges for the primary users of BENEFEDS.gov.
On the side, I juggled articles on NNGroup.com which spoke about UX for older adults. Two articles that helped me out the most were,NNGroup.com/Usability for Older AdultsNNGroup.com/Middle-Aged Users’ Declining Web Performance
3 key persona attributes based on research articles,
Decline in cognitive resources
Research indicates that usability performance decreases by approximately 0.8% per year between ages 25 and 60, reflecting slower information processing and reduced working memory among older adults.
Require clearer navigation patterns and interfaces
Middle‑aged users spend 0.5% more time per page and visit 0.3% more pages per task, highlighting the need for streamlined, intuitive navigation that reduces cognitive effort and supports efficient information discovery.
Accessible Design Gains
Optimizing usability & accessibility can improve task success rates by up to 25% and reduce error rates by 30%, creating a digital environment that’s not only reliable and intuitive but also instills confidence, comfort, and trust.
Confidentiality notice: Connect to see how I realized the target persona, and the full research behind it.
Research
2. Empathizing with the persona, via Qualitative Interviews.
To understand where the tool was actually going off track when users were using it extensively, I conducted moderated interview sessions with 6 users using structured questions and task-based activities on the BENEFEDS.gov portal to uncover friction points and observe where users got stuck. Synthesizing Ella’s proxy interview data with my moderated interview findings, I identified recurring pain points and surfaced the issues users flagged most consistently. I prioritized resonance over echo.
Confidentiality notice: Connect to see who I interviewed and how their insights guided the project.
Research
3. Performing heuristics evaluation.
I also conducted a full heuristic walkthrough of the tool to understand what worked and what failed according to the users. Walked the enrollment flow 20+ times across different entry points, different use cases, and every possible scenario, stress-testing navigation and surfacing edge cases that broke the experience. I mapped my observations against Nielsen's 10 principles.
Heuristic 1
Visibility of System Status
Fail
Heuristic 2
Match between System and Real World
Fail
Heuristic 3
User Control and Freedom
Fail
Heuristic 5
Error Prevention
Fail
Heuristic 9
Help Users Recognize, Diagnose, and Recover from Errors
Fail
Confidentiality notice: Connect to see the full heuristic evaluation data
Research
3. Digging up the platform analytics to get Quantitative Numbers.
I spent time analyzing the data on Looker Studio dashboard with platform analytics gathered from GA5 to check the “time” a user spent on the specific section, that interview users mentioned they have issue with.
A 4-step onboarding with simple questions shouldn't take more than 5 minutes. But when the data showed an average of 6–7 minutes on that section, it confirms that the users were spending too much time in that section. Too much cognitive resources being spent in the wrong place. Time & resources saved there, could be used in deciding the right plan in the plan section part of the user journey.
Another exciting insight,
While observing page views on GA4 and analyzing the pattern, we found that the users were resuming to the first page of the onboarding page again and again.Extremely high # of engagedSessions on that page and average session time being very less. It again matched with the interviews where 8/10 participants expressed “Back Navigation” to be an issue. Click back and it takes you out of the entire flow. No matter how many steps you completed to get there, no matter how many filters applied to reach the right plan, one accidental back click and you need to repeat all the clicks again. A real pain point for not just for the younger audience but for the older adults for sure.
Confidentiality notice: Connect to discuss analytics and see how I mapped numbers to my design decisions.
Challenge 1
Back navigation was broken.
Accidental back-navigation triggered full restarts. There was no recovery path. A user who clicked the wrong thing didn't go back one step, they went back to the beginning. For someone with limited time and limited technical confidence, that's not a minor inconvenience. It's a reason to give up entirely.
Challenge 2
Visibility of system status and jargon overload
"PPO." "FEDVIP." "FEHB." Acronyms with no explanation, no context, no plain language alternative. Users averaging mid-40s were encountering terminology they'd never seen before and being asked to make decisions based on it. There were no progress indicators. No sense of how far they'd come or how much was left. Just a sequence of unfamiliar questions with no visible end.
Challenge 3
Cognitive load was front-loaded, not progressive.
The plan cards were packed with information presented at equal visual weight, making it impossible to quickly identify what mattered. Users didn't realize clicking a link would reveal plan details. The cognitive load was front-loaded, not progressive.
Challenge 4
Filters were hidden. Sort was nearly invisible.
The tools that should have helped users narrow down their options were buried. Filters were sticky and lacked critical options, no way to filter by carrier or annual premium. The sort feature was so poorly placed that users scrolled past it entirely. The interface had the functionality, but the users just couldn't find it.
These weren't separate problems. They were the same problem, a system that violated Nielsen's heuristics 1, 2, 3, 5, and 8 simultaneously, across the exact touch-points that mattered most.
The Problem Statement
HMW improve the benefits enrollment experience so that middle-aged federal civilians can navigate, compare, and select a benefit plan without cognitive overload, navigational dead-ends, or missing system feedback and complete enrollment with confidence?
Confidentiality notice: Connect to discuss how I prioritized what needed the maximum attention.
The Solution
First, simplified the onboarding from the root.
I started with flowcharts to map every step a user took before they even reached the plan selection tool. Then I started reverse engineering. Combined steps. Replaced acronyms with plain language, and introduced tool tips at the right places. Added progress indicator component in the design system first, in this project second and then, every other part of the portal wherever required. So our users always knew where they were and what came next. The goal was to surface system status to the user at all times, and minimize interaction cost before the user even gets to the decision they came to make.
Confidentiality notice: Connect to see the refined wires.
Second, fixed navigation so mistakes weren't permanent.
Fixing back navigation was a debate between UX and engineering, where we couldn’t help but stick to the existing navigation due to standard protocols for .gov websites. A quick fix? Let me introduce an alert modal every time a user wants to, or accidentally moves away from the page. I identified the links, blind spots and spaces where that occurred and introduced a modal with the right UX copy for every instance. My goal here was to absorb accidental user errors, and guide them back to safety, not amplify them.
Third, restructured plan cards for scanning, not reading.
A live brainstorming session with Zarbo and Lucy (UX Designers) intended for 30 mins that crossed hours. What’s the top 3 priority information a user wants to see when they see scan a plan card? What should pull their eye first, second and third? So many design options. So many different card designs, Different typographical hierarchy. Layouts. But the one that showed the right information, the right way, was selected. Visual clutter eliminated. The most important signals price, coverage, and carrier, all readable at a glance. Click the card to see all the details, available on demand, not front-loaded onto a card that was already asking too much.
Confidentiality notice: Connect to see the refined wires.
Fourth, surfaced filters and sort where users actually look.
Filters came out of hiding. Sort placed where the eye naturally goes. New filter options added by carrier, by annual premium. Page layout readjusted on a golden ratio base. My goal here was to help users make a confident decision without the interface or a layer of UI getting in the way. Every configuration, every option, every next step available exactly when they needed it. Not something they had to hunt for after frustration had already set in.
What I eventually fixed and delivered
5+ WCAG 2.2 violations resolved. Navigation broken for 8/10 interview users, fixed. Onboarding jargon replaced with plain language. Plan card information hierarchy redesigned for scalability. Filters and sort surfaced and functional. Enrollment flow restructured to eliminate the drop-off points identified in testing. Tool layout improved for eye movement and natural web usage patterns.
How I delivered the solution
The full feature roadmap sitting on IME matrix
While the tool itself is was big change requiring a good set of resources for development. I strategized a feature lifecycle plan based on discussions with engineering, business and other team members to best utilize the amount of resources available every sprint, and push the most important part of the project first.To me, understanding how a solution gets implemented within existing development resources, limitations, and bandwidth is just as critical as translating the written requirements into a functional prototype. The Impact Effort Matrix helps evaluate, break down, and prioritize solutions into smaller sprint-digestible projects, and assign them to Big Bets, Quick Wins, Fill-Ins, and Time Wasters.This helped me create a dev-go-to-plan which I pitched to the stakeholders to eliminate the overwhelming feeling, while successfully getting a buy in.
Confidentiality notice: Connect to see how I break the full project into sprint-digestible features for faster deployment cycles
Thank you for reading.
If you want to brainstorm on your product to move the needle on a meaningful metric, I am one call away from the conversation that matters.
Let’s Talk
© 2026 harshveenkalsi.com