Skip to the case study
BACK TO HOME
Isometric illustration of a participant seated on the ElfieResearch platform, surrounded by consent, data and study objects

[ 2025-2026 ]

ElfieResearch

A patient-centric B2B clinical-research platform

  • healthcare
  • web-console
  • design system
[ role ]
Product Designer Collaborated with Jang Trinh - Head of Design
[ time ]
Aug 2025
[ platform ]
Web console, Mobile app

[ target ]

Design an end-to-end trial platform across 6 distinct personas, reframing electronic consent from a legal formality into an interactive learning experience.

[ Constraints ]

Audit trail was a design constraint from day one.

[ reframe from insights ]

While most platforms optimize for sponsors, we realized that a great participant experience is what actually drives retention.

This reframed the entire brief.

Instead of starting with the admin console, we focused on where trials usually break down: onboarding. That made eConsent the natural starting point—not just as a legal signature to collect, but as an accessible first impression and a crucial comprehension test.

[ scope ]

[ Key Decisions ][ System ]

Draw the permission model before the screens

Mapping six roles against seven classes of data made the overlap visible: the roles differ in the depth of their access, not in the objects they touch. Per-role applications would duplicate one data model rather than specialise it.

  1. 01

    Problem

    Six roles with conflicting access rights inside one product.

  2. 02

    Options

    • A. Separate application per role
    • B. One console gated by role-based access control.
  3. 03

    Chose

    One console, RBAC-gated.

  4. 04

    Rationale

    Unified data model: Maintains a single source of truth across the entire platform.

    Lower maintenance: Eliminates the overhead of building, deploying, and managing multiple distinct applications.

    Streamlined compliance: Ensures one consistent, centralized audit trail across every role for regulatory and security needs.

  5. 05

    Target

    A single audit trail spanning every role. Never verified in production.

Roles and permissions matrix: roles listed against actions, with a permission count per group
Permissions are expressed as a grid of role against action, with the count per group visible. It is the same shape as the compliance document it implements, which is what makes it auditable rather than merely configurable.

[ Key Decisions ][ the builder ]

Design the builder, not the form

Protocol amendments force re-consent, and every version of every form has to stay retrievable for audit. So the deliverable was the tool that makes the form. One primitive carries four modules: participant identification, screening, eConsent and withdrawal.

  1. 01

    Problem

    Consent and screening content changes after a study launches.

  2. 02

    Options

    • A. Design each form per study and re-issue it on amendment.
    • B. Build a block-based builder with version history.
  3. 03

    Chose

    The builder, with edit / preview / view states.

  4. 04

    Rationale

    The regulatory constraint requires every version retrievable and every change attributable. A per-study form has no version spine to hang that on.

  5. 05

    Target

    Study teams amend a consent form without a design request.

eConsent module in the builder's edit mode, with legal statements as reorderable rows and a placed signature block
eConsent in edit mode. Signature is a placed block, not a fixed footer; legal statements are reorderable rows; the language tabs and the AI-translate toggle belong to the primitive, so every module inherits them.

[ Key Decisions ][ the builder at work • screening ]

Branch the screening, not the study

Eligibility is routed per answer, in sentences a coordinator can read: if greater than 18, go to Q2. Every unhandled case falls to a named ending screen, so no participant can leave the flow undecided.

  1. 01

    Question types

    14, across number, single-choice and multi-select types

  2. 02

    Fallthrough

    All other cases go to Ending screen · Not eligible

  3. 03

    Supported languages

    Over 10 languages, with British English as primary. Translated by streaming AI and reviewed by the study team

  4. 04

    Preview

    Live mobile and desktop rendering beside the configuration panel

Screening builder with per-answer branching rules beside live mobile and desktop previews

[ Key Decisions ][ econsent ]

Turn consent into comprehension

Consent was scoped as a legal formality. FDA guidance on decentralised trials treats the eConsent system as where understanding is established, not only recorded. We argued for building the participant surface before the console, starting with consent.

  1. 01

    Problem

    Participants sign consent documents they have not understood.

  2. 02

    Options

    • A. Scroll-and-sign PDF
    • B. Explainer video plus signature
    • C. Interactive modular flow with comprehension checks gating the signature.
  3. 03

    Chose

    The check gates the signature rather than trailing it.

  4. 04

    Rationale

    Regulatory alignment: FDA guidance explicitly favors comprehension-oriented consent models.

    Higher engagement: Interactive elements directly boost completion rates and retention.

    Reduced anxiety: Structuring consent like an onboarding flow makes complex trial information approachable and manageable for participants.

Participant eConsent flow on mobile, one idea per screen with comprehension checks before the signature
eConsent in mobile for participants
eConsent flow in edit mode, segmented into steps with a refusal route
eConsent in edit mode. The flow is segmented rather than continuous: a participant meets one idea, answers for it, and only then moves on. Refusal is a route with a survey behind it, so a non-consent produces a reason instead of silence.

[ Key Decisions ][ Retention ]

Design for the participant who is alone

A decentralised participant has no waiting room and no nurse. Isolation drives dropout, and the control arm receives least by design. Reminders, visible progress and streak mechanics are configurable per arm; recruitment is drawn as four separately tracked sources.

  1. 01

    Problem

    Remote trial participants experience isolation and drop out at significantly higher rates compared to site-based participants.

    The control arm has to stay enrolled while receiving very little.

  2. 02

    Chose

    Implement a gamified engagement layer directly within the app, fully customizable per treatment arm.

  3. 03

    Rationale

    Preserving statistical validity: Retention is the critical KPI that determines whether a trial maintains statistical validity.

    Core product function: In remote trials, active engagement is not an optional feature—it is the primary driver of participant retention.

Participant app home screen beside an ePRO question
Participant home and an ePRO question. Everything above the fold is a reason to return between assessments rather than a report of study status; what a participant owes the study appears after what the study offers them. Participant app, 2025.
Protocol configuration screen in edit mode
Protocol configuration in edit mode

[ Key Decisions ][ the states ]

Design the states nobody demos

A regulated product is judged on the paths a demo skips. Each of these exists as a designed screen in the file, not as a note for engineering.

[ outcome ]

Complete multi-platform flows (console, web, mobile) in just 1 month from the initial PRD publication.

eConsent (End-to-end flows for electronic consent); ePRO / eCOA (Electronic patient-reported outcomes) ; Patient Recruitment (Landing page + pre-screening / eligibility flow); Console (Role-based web console used by trial staff.)

Flow board of the end-to-end user flows across console, web and mobile
Detailed, step-by-step user flows utilizing real-data mockups, critical edge-case coverage, and interactive prototypes for stakeholder presentations.

[ reflection ]

Positioning eConsent as a critical retention lever did more than improve the user experience - It anchored our design strategy directly to business KPIs, ensuring design and product were finally speaking the same language.

[ next project ]