Healthcare & Wellness SaaS
Lead UX/UI Designer & Design Engineer (solo owner from lo-fi to hi-fi).
Defined IA, user/doctor/coach flows, and AI Insights spec (inputs, drivers, confidence, provenance).
Built design system (tokens, components, chart patterns) and interactive prototypes (onboarding with one-tap sync, Today/AI Dashboard, doctor summary, coach triage).
Ran usability testing (plan, scripts, analysis) and drove iterations (today-first dashboard, conversational onboarding, right-rail AI for clinicians).
Set success metrics (onboarding ≤2 min; sync ≥90%; doctor decision <60s; coach review <90s/client) and instrumentation guidance.
Rapid digitization across health/medical/fitness; mobile + connected devices now core.
Personalization is mainstream: 54% of U.S. consumers tracked ≥1 health metric (2024).
Providers/gyms are scaling AI: 78% used AI in ≥1 function (2024).
Fitness/health app usage (U.S.): 27.9% track via mobile apps.
Wearable adoption (U.S.): 53% own ≥1 wearable/connected device (2024).
User vs. Organization needs
What they need most
👤 Users
Frictionless capture
Personalized goals & nudges
Easy, transparent booking
Privacy controls
🏥 Providers / Orgs
Standardized longitudinal data
AI summaries & risk flags
Integrated scheduling/capacity
Adherence & outcomes analytics
Compliance & security
We fielded a short role-based survey (Users, Doctors, Coaches) to validate interview themes around capture friction, metric overload, workflow gaps, and scheduling pain. Questions map directly to product decisions (capture method, guidance format, integration/EHR, coaching triage), enabling fast, evidence-backed prioritization.
Distribution/collection: Sent to 87 people; 64 valid responses collected over 2 weeks (online form).
Sample split: Users 31, Coaches 17, Doctors 16 (convenience sample via gym mailing list + clinic partners; anonymous).
Survey Results
Logging preference (users): Auto-sync 40%, Voice 30%, Photo 25%, Typing 5%, Would not log ~2%.
Finding trends (users): 60% rated it difficult (4–5), 20% neutral, 20% not difficult.
Coach view value: ~85% top-2 box valuable; 10% neutral; ~5% not valuable.
Doctor agreement (AI summary helpful): ~68% agree, 20% neutral, ~12% disagree.
Affinity Diagram
Inputs: 7 semi-structured interviews (3 users, 1 clinicians, 2 coaches) + open-text from 64 survey responses.
Team: 5 people — Lead UX (me), PM, clinical advisor (MD), coach lead, data analyst.
Method: two 60-min remote clustering workshops in FigJam. We de-duplicated and tagged ~100 notes, ran bottom-up grouping → named clusters → merged into themes and a final “evidence → design lever” mapping.
Outputs: 5 themes, each tied to sample quotes, metrics, and a concrete design response.
Capture Friction → {manual entry fatigue, multi-app juggling} → One-tap voice/photo + background auto-sync
Insight Overload → {too many metrics, unclear priorities} → “What matters today” digest; goal-linked insights
Workflow Integration → {EHR gap, coach data scattered} → Standardized longitudinal cards; export/share to EHR & coach hub
Adherence & Context → {drop-offs, unknown reasons} → AI-detected context (sleep/travel/illness) + gentle nudges
Trust & Privacy → {accuracy doubts, sharing anxiety} → Confidence badges, data provenance, granular consent
Users overwhelmingly prefer auto-sync/voice/photo (not typing) and report frequent metric confusion, confirming the need for low-friction capture and a “What matters today” summary. All roles cite fragmentation, with doctors favoring a 90-day trend view yet blocked by EHR integration—so summaries must export cleanly to charts/notes. Coaches are time-limited and prize weak-spot detection (sleep/travel/illness) for triage, and users face booking friction—pointing to a unified hub, clinician-ready summaries, coach attention queues, and smarter scheduling.
Personas summarize patterns from our interviews/survey so the team can make consistent trade-offs. Each one represents a cluster of motivations, behaviors, and constraints that shape what “good” looks like for them—and which UX decisions we should prioritize.
A journey map shows the sequence of stages, moments of truth, pain points, and opportunities.
Rows = what we observe: triggers, actions/touchpoints, pains, ⚡ AI interventions, and success metrics.
Columns = stages users go through.
⚡ icon marks where intelligence enters the flow (e.g., summarization, prediction, automation).
Metrics anchor design to outcomes, not outputs.
Lo-fi prototypes explore structure and flow before visuals. I built two User versions: one that leads with data cards and one that leads with the AI “Today” panel, both sharing the same basic navigation and appointment steps.
The Doctor and Coach versions follow the same card-based layout with a summary view and a focused insights area to keep patterns consistent across roles.
These wireframes aim to confirm hierarchy, scalability, and task clarity—can users act quickly, can doctors grasp status fast, and can coaches triage efficiently—before we invest in high-fidelity design.
The hi-fi explores a cohesive, production-ready system across User, Doctor, and Coach, unifying a card-based layout, 12-column grid, and tokenized styles for speed, clarity, and consistency. AI is explicit and trustworthy: insights live in cards; actions happen in the assistant thread with provenance and confidence.
System highlights
Consistent card anatomy, spacing, and iconography for fast scanning
Clear hierarchy: greeting/intent → trends → insights → action
Accessible charts and copy; status is always color + text
Assistant thread converts guidance into scheduled actions (no context switching)
Key pages
User Dashboard: Goals-first layout with three progress cards anchors intent; compact KPI chips and a trend chart show trajectory; The right-side AI Assistant converts guidance into actions (confirm run, reminders, meal saves) without leaving the page.
User Activity: Route + workout stats paired with trend context; persistent assistant for quick reminders and nutrition follow-ups.
User Appointment: Provider details, pre-visit tasks, and history in one view; timeline reduces hunt time; assistant finalizes preperation.
Doctor Patients List: Triage at a glance with badges (AI Suggest/Upcoming/Urgent); calendar + searchable cards minimize navigation.
Doctor Patient Details: Centered patient sheet modal summarizing identity, last-visit reason, diagnoses, pharmacy, booking info, and a timeline. Right rail keeps the Assistant thread available; a dedicated clinical insights panel is not yet present.
Outcome
The hi-fi validates that users grasp “what to do now” in seconds, doctors get a safe, defensible summary, and the assistant closes the loop—turning insight into booked behavior or documented care within the same screen.
Study goals
Validate if the lo-fi → hi-fi flows enable:
New user onboarding and sync health data from outside source.
Doctors to find 90-day health trends quickly and act.
Coaches to understand AI suggestions and prioritize clients.
Participants & method
n=6 (3 users, 2 clinicians, 1 coaches); 30–40 min remote and in-person moderated sessions; think-aloud + clickstream.
Success = task completion, time-to-signal, error count; post-task confidence (1–5).
Core tasks
User: complete new user onboarding, provide health data, and set a starter plan.
Doctor: open a patient, understand last 90 days, decide next step.
Coach: scan roster, identify who needs attention, draft a weekly plan.
Issues → Design decisions
Users hesitated on the goals-first home (scanning overhead; unclear next action).
Change: elevate “Today / AI Insights” at the top; keep progress cards secondary; keep assistant persistent for immediate confirm/schedule.
The form-heavy intake produced high perceived effort, and drop-off/hesitation at the health data fill in step.
Change: redesign onboarding into a conversational, AI-assisted flow that auto-fills basics, smart-detects devices for one-tap sync, and ends with an auto-generated starter plan
Clinicians lacked a concise, defensible summary (trend hunting, provenance unknown).
Change: add right-rail AI Insights panel with 90-day one-liner, risk flags, explainable drivers, ranked next steps, and confidence & provenance.
Coaches needed triage and comparison.
Change: introduce attention queue and side-by-side compare in Training Trends (coach portal).
Conclusion:
Findings directly drove the shift from goals-first to today/insight-first on User, and from a static patient sheet to a defensible AI summary on Doctor—reducing time-to-action and increasing confidence across roles. The onboarding pain observed specifically informed the AI-assisted redesign to lower cognitive load, clarify consent, and raise device-sync completion.
Onboarding
Replaced long forms with a brief conversational flow. Feels quick and seamless, aligning with users’ desire to start immediately.
Combined permissions + connection into one clear step; de-emphasized manual setup. Clarifies what’s shared and why, building trust while saving time.
Previous Onboarding: 9 Steps with manual input
Current Onboarding: 4 Steps with auto-fill
Previous Dashboard: hesitates on goals
Current Dashboard: AI insights guides direction
Doctor Patient Details
Added right-rail AI Insights. Gives a defensible summary at a glance, matching clinicians’ need to decide quickly.
Displayed confidence & provenance; streamlined path: context → insights → orders/note. Makes evidence traceable and the next step obvious, which aligns with clinical workflow expectations.
Previous Patient Details: Lack information
Current Patient Details: AI insights guides direction
Why we iterated
Users said the goals-first home “looked nice” but didn’t always answer “What should I do now?”
During testing, the previous onboarding created intake friction and consent ambiguity, threatening activation and sync completion; the redesign focuses on conversational guidance, one-tap device sync, grouped consents, and immediate value.
Clinicians could read the patient sheet but still asked for a safe, concise trend summary with defensible evidence before acting.
Result
Faster user action from insight to scheduled behavior.
Onboarding now shorter, clearer, and more likely to complete device sync and consent correctly.
Clinicians reach confident decisions in <1 minute with traceable evidence.
Hi-Fi UI: User Dashboard (Today/AI, quick log, appointment reminder), Doctor Summary (90-day AI insights), Coach Trends (attention queue, recs).
Interactive prototypes: onboarding with one-tap sync; doctor review → note; coach triage → plan.
Design system: tokens (color/type/spacing), core cards & chips, chart styles, states (empty/error/low-data).
Research pack: test scripts, key findings, before/after screenshots showing reduced steps/time.
Accessibility & privacy: WCAG contrast/focus/ARIA; clear consents + revoke controls.
Method & workflow
UCD spine from problem framing → journey maps → IA/flows → lo-fi → hi-fi → usability testing → iteration; evidence from interviews + short survey + task data; stage KPIs (time-to-action/insight, triage).
What we learned
Consistency > novelty. Shared card anatomy and assistant pattern lowered cognitive load across roles.
AI as a bridge. Best used to translate streams into ranked, explainable actions and close the loop in-flow.
Impact of AI on the process
Speed to insight: generated copy/variants quickly, enabling more iteration on decision friction instead of UI polish.
Personalization at scale: produced “tasks” and clinician one-liners with rationale/confidence, improving relevance.
Ops leverage: automated starter plans and note stubs reduced manual prep for users, coaches, and clinicians.
Limitations & Next Steps
Need longer-run, noisy data to validate robustness and confidence ranges.
EHR/FHIR integration and accessibility testing require deeper pilots.
Finalize flows, hi-fi specs, and component tokens/cards for developer handoffs.
Define data contracts for device sync, consent scopes, and audit.