Growing a clinical app into an open research platform
As Principal Designer at Sage Bionetworks, I led the redesign of mPower, a digital study tool for Parkinson's research, then evolved it into an open-source platform reused globally across phone, wearable, and voice.
Sage Bionetworks & Apple · Principal Designer · 12 months
Increase in study completion
Daily ask adherence for study duration
Studies using the new design system
The problem
A viral launch that collapsed within a month
mPower launched at Apple's WWDC to massive hype, with tens of thousands of downloads in days. But within a month daily retention dropped to 5 percent, starving the research. Without sustained use, scientists could not collect enough data to validate digital biomarkers like tremor or voice change.
Retention drop-off and the original mPower interface
The original experience prioritized scientific completeness over usability. Consent was legalistic, tasks felt like chores, and the only payoff was a dashboard of charts people did not understand. Despite strong scientific interest and the Apple partnership, Sage needed an experience people would actually return to.
The opportunity
Reframe it as a participant-first platform
Instead of patching flaws, I reimagined the experience from the ground up, putting participants first to build trust, reduce burden, and deliver high-quality data. The goal was a platform that engaged people with clear, meaningful tasks, worked across phone, wearable, and voice, and gave scientists modular, reusable tools for future studies.
Understanding the problem
Listening to 24 people living with Parkinson's
We spoke with 24 individuals living with Parkinson's, and the themes were clear. The app was confusing, the process was frustrating, and the experience felt one-sided. Consent happened only after download, an immediate barrier, and people got thrown into long explanations without context, often stuck in loops they could not complete. One usability participant got caught in a verification loop.
"I have 30 years of data points in my head, but I can't really do anything with them."Participant 3
The tone and interactions lacked warmth and clarity. Email verification confused many users, instructions were missed, and even minor friction like not knowing where to tap compounded the emotional fatigue of participating in a study about decline.
This validated a core principle. Scientific rigor and empathy are not mutually exclusive. To succeed the product needed to feel as supportive as it was structured.
Original mPower consent flow with multiple friction points
My approach
Four pillars for long-term engagement
Designing for long-term engagement meant rethinking every layer, from participant trust to platform flexibility. I anchored the redesign around four pillars.
"You're asking me to give up my data and time, and I'm not sure what I get out of it."
1. Participant reality as a foundation
Participants voiced emotional and cognitive hesitation, so I reframed onboarding and consent as reciprocal rather than extractive, adding clearer messaging, progress cues, and tone testing across age and identity groups. The result was a 50 percent reduction in onboarding drop-off.
2. Rebuild consent and onboarding
The original flow overwhelmed people with dense legal text upfront. We broke it into plain-language summaries with progressive disclosure and inline illustrations, making expectations transparent and manageable, which lifted comprehension and completion and smoothed the handoff into the first task.
3. Modularize the measurement experience
Task flow was rigid, mixing surveys, voice prompts, and motion tests into one unbroken timeline that left people lost. We broke tasks into named, reusable modules with clear instructions and feedback, shaped by co-design sessions, so each module stands alone and is easy to pause and return to. Study completion rose 7 times.
4. Extend the platform beyond phones
To support Apple Watch and voice-based tasks, I partnered with engineering on a device-aware design system with theming, illustration templates, fallbacks, and platform-integrated UX for sensors. Forty percent of participants stayed active for 30 or more days, and sensor tasks drove the majority of completions in later phases.
The solution
Consent as a relationship
Many users went through multiple steps, including downloading the app, only to discover they were not eligible, which left them frustrated. We redesigned the consent flow and built a dedicated study website that communicated the study's value upfront. People could review requirements, understand what to expect, and consent directly on the site, with a link to download the app only after confirming eligibility.
Original study website, overwhelming consent flow
Redesigned site, streamlined and participant-friendly
Key enhancements included:
- Introducing the study team up front to create a human connection
- Large type, simplified language, and visual PDFs to improve comprehension
- Chat, email, and phone support to reduce participant isolation
- Eligibility and comprehension checks built in to ensure informed enrollment
This consent model became a reusable pattern across later studies and set a precedent for scalable ethical UX in mobile research.
Consent overview with detailed study information
Digital signature and consent confirmation
Step-by-step onboarding process
Registration confirmation and app download link
Study requirements overview
Data privacy and security details
Phone number registration
SMS confirmation and app download
A scalable, modular design system
Rather than designing individual screens per study, I created a scalable design system of themed, reusable modules that kept consistency across clinical research apps while staying flexible for diverse populations and study types. Designed with tremor-friendly touch targets and passive interaction modes, it became the foundation for products like BP Lab and Heart Snapshot, and was eventually integrated into Apple's ResearchKit.
Core measurement modules
Modules ranged from physical-activity trackers and cognitive assessments to self-report surveys, each built as a self-contained UX unit and co-developed with researchers to ensure scientific validity.
Tremor tracking, for example hold-phone tests
The tremor test shows the modular approach: clear step-by-step guidance, accessible instructions with visual aids, and a streamlined experience that reduces cognitive burden while collecting clinically valid data.
Test introduction
Instructions
Sit down
Hold phone
Test in progress (30s)
Test progress (15s)
Test complete
Gait and balance, for example step cadence via motion sensors
The gait and balance test shows another modular approach: clear setup, audio guidance preparation, and seamless phone-based motion tracking that captures walking patterns for clinical analysis.
Test introduction
Walking instructions
Volume setup
Pants with pockets
Phone placement
Test complete
Design system foundations
To ensure long-term reusability and inclusivity, the system later expanded to include assessment modules such as a VO2 max chair-stand task and passive heart-rate variability monitoring through wearables, plus cognitive activities and self-report surveys. I embedded accessibility-first patterns including high-contrast support, motion-sensitivity defaults, and voice narration hooks.
Cross-platform UI component library
A single cohesive set of UI elements, buttons, forms, navigation patterns, and feedback indicators, that adapt across mobile apps, watch faces, and voice interfaces, keeping consistency and accelerating cross-platform development.
Small sample of ResearchKit's design system
Brandable themes and adaptive, inclusive illustrations
A unified theming framework and illustration library that lets research organizations and pharmaceutical sponsors apply their own branding, and re-skin illustrations to represent diverse age groups, races, and abilities.
Theming and swapping illustrations by age and gender
Mobile, wearables, and hands-free
I tailored layouts, touch targets, and interaction patterns for a seamless experience across smartphones, smartwatches, and voice-driven hands-free interfaces.
Morning check-in
BP measurement
Check-in reminder
Stress tracking
Data history
Detailed metrics
Guided breathing
How it performed
More trust, less friction, scaled across studies
| Outcome | Before | After |
|---|---|---|
| Study completion | Baseline | 7x |
| Consent drop-off | 41% | 15% |
| Daily retention | 5% | Sustained |
| Studies adopting the system | 0 | 3+ |
| Design time per study | Baseline | Cut 50 to 70% |
The modular system accelerated research launches and broadened access. It powered BP Lab and Heart Snapshot, prompted researchers to adopt participant-first, scalable design, and its theming and accessibility patterns enabled broader reach across age, ability, and identity. Journey PRO earned an IXDA Interaction Awards Honorable Mention and BP Lab was a Fast Company Most Innovative in Health honoree.
Why it matters
This work turned a failed launch into the foundation of a global research platform. By centering human needs, burden, clarity, and dignity, we did not just improve usability, we made science possible again.
Today these tools power studies around the world, help pharmaceutical companies refine treatments, and drive product innovation in wearables. Adopted into Apple's ResearchKit, they reached a global audience across partner institutions. We did not just fix an app, we fixed the relationship between people and research.