Project details

• Platform: iOS, Android
• Design system management: Supernova.io
• Product type: B2B SaaS, SDK, White label app
• Design system: Material 3, UCDL
• Design tool: Figma

What is it?

A clinical-like health assessment consisting of multiple stages that is designed to assist clinicians and upstream providers with combined risk analysis of chronic diseases.

Assessment stages

1. Cardiovascular Health: A 30 second face scan that measures heart rate, blood pressure, heart attack, stroke.
2. Health at Rest: Lying down for 2 minutes and then conducting a 30 second finger scan to measure resting heart rate.
3. Body Analysis: A 3-5 minute body scan to measure body circumference, body fat, obesity, central obesity.
4. Fitness Evaluation: A 3-5 minute physical step test and then a 30 second finger scan to measure maximum heart rate, and heart rate recovery.

What do you need?

• Face & Body Scan: Desk, chair, table.
• Health at Rest: A bed or couch.
• Fitness Evaluation: Some stairs or a riser step.

How long does it take?

• First-time: 30 minutes, split over a couple of days.
• Ongoing: About 15 minutes.

Use cases

• Health insurance underwriting, dispensary of pharmaceutical drugs (e.g. statins, semaglutide), telehealth, population risk, corporate wellness.

What did I do?

Hands on:

• Initial MVD (Minimum Viable Demo), and MVP (Minimum Viable Product)
• My responsibility varied over time, but I was responsible for the ‘entire app’.
• This included research, stakeholder mapping, persona mapping, user journeys, user testing, prototypes, developer handoff, the scans (each stage), scan guides, results, product docs, marketing assets, and video guides/promos.

Minds on:

• Worked with the MLCV team and engineering to validate model outputs at each stage of the assessment.
• Defined what “good enough” looked like for each model, based on the clinical, commercial, and reputational cost of getting it wrong.
• Defined safety guardrails for the system.
• Evaluated model outputs against medical reference data and identified where accuracy dropped around across body types, scan quality, questionnaire inputs, and exemption paths.
• Tested the full assessment pipeline end-to-end, confirming each stage produced reliable inputs for the next and that final risk classifications held up.
• Validated edge cases. Cancellations, incomplete assessments, exemptions, etc, to make sure the system behaved safely outside the expected flow.
• Used validation and UX findings to adjust thresholds, decision rules, and workflow logic, separating model problems from design, data quality, and integration issues.
• Translated validation results into product and engineering priorities, so the team could focus model improvements on what mattered most to users and the business.

Major challenges

• Product: High drop off due to usability oversight, notification & message fatigue, regulation and app store guidelines, customisation and theming, no localisation.
• Product complexity: High complexity and barriers to change due to emerging technology innovation cycles. i.e. Working with data science and training models, then testing is expensive.
• Organisational: Design & features by leadership (top down hierarchy), misaligned reward systems, managerial self-interest, team silos, misaligned commercial messaging.