Visuals and interface screenshots are under NDA and cannot be shared. All findings described below are based on research conducted during the internship.
context
Context
A self-service hub that members rely on — often when confused or stressed
The Ambetter Knowledge Center is where members learn about plans, benefits, and how to use their coverage. At the time of this project, the page consisted of a simple grid of categories, each with a list of hyperlinks to articles.
My team was tasked with bringing the Knowledge Center up to the visual and accessibility standards of the broader Ambetter website. My role was to lead the research stream: understand how well members could currently navigate the space, and surface insights that would guide the design team's decisions.
Three questions shaped the work:
- Can members efficiently find information without excessive clicks or back-and-forth?
- Do current labels and categories match how members naturally think about health insurance?
- Where do mobile web and app navigation patterns diverge, and which is more intuitive?
methods
Methods
A mixed-methods approach, sequenced for signal clarity
I chose a four-method sequence that moved from broad to specific — each phase designed to answer different questions while building on what came before. The goal was to isolate distinct types of navigation problems: structural issues, labeling issues, and interaction-pattern issues.
01 — Competitive Analysis
Reviewed how UnitedHealth and other insurance portals structure health information architecture. Established a benchmark for industry-standard patterns before evaluating Ambetter's approach.
02 — Heuristic Evaluation
Team evaluated the Knowledge Center against Nielsen's 10 usability heuristics. Low-cost way to surface obvious gaps before investing in participant recruitment.
03 — Tree Testing
40+ participants completed 8–10 navigation tasks using a text-only site structure in UserZoom. Isolates IA issues from visual design — tests whether the structure works before adding interface noise.
04 — Unmoderated Usability Testing
Mobile web (n=12) and app (n=8), with realistic task scenarios and think-aloud protocol. Tasks reflected real member situations, such as receiving an unexpected bill.
findings
Findings
Four patterns, each with a clear design implication
Three numbers framed the findings: 35% task success rate on "Plan Details" — the most-visited label — 4–5 taps required to reach commonly needed information on mobile, and 30% faster task completion in the app versus mobile web.
Labels used internal terminology, not member language
In tree testing, 12 of 40 participants struggled to find "benefits covered under your plan" when it was labeled "Plan Benefits." They searched for "What's Covered?" or "My Coverage" — the language they used naturally. Recommended a content audit to rewrite labels in plain language aligned with how members speak about their insurance. Worked with the content designer to prioritize the most-visited pages.
Navigation depth required too many clicks on mobile
Usability testing revealed members taking 4–5 taps to reach commonly needed information. One participant described it as "feeling like a maze." On mobile, small touch targets and pagination made navigation feel slow and error-prone. Recommended flattening the IA and surfacing top tasks at the first level, reducing depth by 1–2 levels. The design team implemented this in a later sprint.
Members couldn't confirm they were in the right place
8 participants asked "Is this the right page?" or spent time re-reading to confirm they were in the correct section — a sign that pages lacked enough context to validate that members were on-task. Recommended adding inline help text and clarifying subheads on key pages so users could self-confirm without backtracking.
outcomes
Outcomes
Findings translated into shipped changes
Rather than delivering a report and walking away, I worked directly alongside UX and content designers to translate each finding into a concrete, prioritized action. I created deliverables in multiple formats — a stakeholder-facing PowerPoint, a simplified one-pager for the design team, and annotated Figma screenshots that surfaced friction points directly on the interface.
Changes implemented or scheduled:
- Top 20 navigation labels rewritten in plain language (e.g., "My Costs" replacing "Cost Sharing Information")
- IA flattened by 1–2 levels for the highest-traffic tasks on mobile
- Persistent tab-bar navigation prototyped and scheduled for development on mobile web
- Inline contextual help text added to key pages to reduce navigation uncertainty
Recommendations for the next phase:
- Conduct moderated discovery sessions to understand member intent and emotional state when visiting the hub
- Pair qualitative insights with analytics (drop-off rates, search queries) to quantify the business impact of navigation friction
- A/B test the new label set and navigation structure to validate improvements against a baseline
reflections
Reflections
What this project reinforced
Research is a bridge, not a silo. The impact came not from the volume of data collected, but from the ability to distill patterns and translate them into decisions the team could act on. Collecting usability videos is one thing; turning them into a clear recommendation like "change the label from X to Y because Z participants struggled" is where research becomes useful.
Navigating ambiguity is a skill. The mid-project scope shift initially felt disruptive. Reframing it as an opportunity to focus on higher-impact areas made the findings more actionable. The mobile-first pivot meant testing what mattered most, not everything.
Stakeholder alignment prevents rework. The scope shift that happened midway could have surfaced in week one with early clarifying questions — "Who is the primary user? What platform matters most?" — rather than working from assumptions and adjusting later.