From outdated SaaS to efficient care delivery
74% of in-app feedback was negative. A card sorting study with 35 doctors across 12 specialties revealed no universal record exists. I used that evidence to redesign the system from the ground up — and validate it with 113 psychologists.
👩🏻💻 Role
UX Research · UI Design · Design System · Product Strategy · Lead Design
🤝 CLIENT
Conexa · Telehealth Platform · Medical Team
👥 Team
Design · Product · Engineering · Medical Ops
· Context
Conexa is one of Brazil's leading telehealth platforms, connecting patients with doctors, psychologists, and other healthcare professionals through digital consultations. At the core of the professional experience sits the medical record — the tool specialists use during every single appointment. When feedback data showed that 74% of in-app reviews were negative and complaints pointed consistently to the same problems, it became clear the record wasn't working for the people who depended on it most.
This project was about understanding why — and redesigning the experience around how clinicians actually think, not how a system assumes they do.
01 — The Problem
74% of feedback was negative.
Doctors described the system as confusing, polluted, and not built for them
🗂️ Visual overload
The top complaint in feedback analysis: too much information per screen made it impossible to find critical data during live consultations.
🔀 Broken navigation
Since the app had no analytics or tracking tags — only Excel files listing user sTabs without clear visual separation, unintuitive flows, and zero hierarchy. Doctors had to memorize where everything lived — every single time.uggestions and support tickets — we analyzed these files to identify key insights and areas to focus our efforts.
📐 Wrong language, wrong model
Psychologists were forced into a medically-framed record — including mandatory CID codes that have no place in psychological practice.
📋 No universal record exists
Card sorting across 12 specialties proved it: only 3 fields had 80%+ consensus. A one-size-fits-all record was never going to work.
Regarding the old medical record, note that it uses multiple different fonts; on this screen, it uses two different design systems and lacks a clear logic for data visualization for medical professionals.
The layout feels like it wasn't made for the people who actually use the system. The more fields, the more it gets in the way.
— Healthcare professional, in-app feedback
02 — Research Approach
35 doctors. 113 psychologists. +700 feedbacks.
Every method possible.
01. In-App Feedback Analysis
From over 700 responses collected from the in-app survey, 74% of all in-app record feedback was negative. I categorized it by theme, usability, complexity, performance, and incompleteness, to build an evidence-based brief.
02. Card Sorting
35 doctors across 12 specialties categorized 27 clinical fields as essential, optional, or unnecessary. The finding was decisive: no universal record is possible.
03. Unmoderated Usability Testing
113 psychologists tested a functional prototype via Maze — completing real clinical tasks and rating the new layout on structure, field relevance, and flow.
04. Internal Data Analysis
Analytics revealed that only 2–8% of consultations resulted in referrals — exposing a critical adoption problem in a strategically important feature.
05. Stakeholder Interviews
Conversations with internal healthcare teams surfaced friction points invisible in data: confusing flows, lack of feedback, and zero visibility on referral status.
A little bit of the figma file with the screens created to the unmoderated test and perception test with specialized doctors and psychologists. Additionally, with the informations I get with the perception test, its was possible to do upgrades in the new EHR at the same time of our conversations.
03 — Design Process
How I turned research into decisions
PHASE 01
🧏🏻♀️ Discover & Listen
→ Recruited 30+ specialists
→ Mapped pain points & mental models
→ Analyzed usage data patterns
→ Benchmarked competitor systems
PHASE 02
🖼️ Define & Frame
→ Synthesized research into themes
→ Prioritized critical user journeys
→ Defined success KPIs
→ Aligned with product & engineering
PHASE 03
📝 Design & Evolve
→ Rebuilt information architecture
→ Updated design system tokens
→ Iterated on components & flows
→ Validated each cycle with users
PHASE 04
Measure & Roadmap
→ Usability testing post-launch
→ Error rate & satisfaction tracking
→ KPIs embedded into roadmap
→ Continuous feedback loops
04 — Design system
Evolving the system
without breaking it
Rather than a full redesign, the approach chosen was evolutionary — upgrading what existed instead of discarding it. This reduced engineering risk while delivering immediate visual coherence.
One of the key drivers behind the component updates was a company-wide rebranding. The new brand direction required not only visual alignment, but also the creation of entirely new components and variations that had never been mapped before — gaps that the old system simply couldn't cover. And just as importantly, the design system's construction and updates were designed with accessibility in mind.
Color, typography, and spacing tokens were updated first, establishing a reliable foundation. Inconsistent components were then systematically rebuilt, the missing pieces the rebranding demanded were introduced, and every decision was documented so the team could move fast without diverging.
The result: a design system the engineering team could actually use — and a visual language doctors could finally trust.
The color documentation image shows a portion of how the before and after results were presented to the engineering team.
a glimpse of what was done in revitalizing the design system
05 — Outcomes
What changed — and how we know
Results from the first usability test with 113 psychologists — unmoderated, via Maze, Q3/2025.
Field Relevance · 77%
Fields finally matched how clinicians actually think
77% of participants rated the new fields as more useful than the existing system — validating the research-led curation and moving away from a generic medical template.
Clinical Flow · 87%
Information order followed clinical reasoning
87% confirmed the new field order matched their thought process during consultations — reducing cognitive load and unnecessary navigation mid-session.
Structural Design · 88%
Separating first session from recurring ones was the right call
88% validated the two-template model. Psychologists described it as more practical, less repetitive — and a meaningful improvement to continuity of care.
History Navigation · 79%
Patient history became easier to read and navigate
79% said reviewing patient history was clearer in the new layout. Professionals could pick up where they left off — without hunting through dense, undifferentiated records.
07 — What I Learned
The things that actually mattered
🧩 Personalisation by specialty isn't enough
The card sorting showed that 100% of specialties had high internal disagreement — not just between fields, but within the same specialty. Individual customisation has 3× more impact than specialty templates.
✅ The right structure changes everything
Separating first-session from recurring records wasn't a minor UX improvement — 88% of psychologists said it changed how they work. Sometimes the biggest win is a structural decision, not a visual one.
🟢 Low adoption numbers are a design signal
The referral feature had just 2–8% adoption — not because doctors didn't need it, but because the flow was backwards. Data analysis led directly to a redesigned interaction model and a 400% adoption target.
you’ve reachead the end!