Segmed: Improving the Website Experience for AI Researchers and Healthcare Partners


Timeline: 2 months

Tools: Figma, Google Analytics, Squarespace

Skills: User Interviews, User Research, Branding, Prototyping, A/B Testing

At Segmed, a Series A startup (YC20) dedicated to democratizing healthcare data with AI, I embarked on a critical project as the team’s first design intern to spearhead our website’s first signficant redesign. Under the guidance of my mentor, Akemi Leung, I led the project from audit to launch, leveraging A/B testing to optimize our outcomes.

About SegmedSegmed’s mission is to revolutionalize healthcare by providing secure medical imaging data. Founded by physicians and researchers from Stanford, Segmed empowers AI development and global health initiatives.

However, Segmed had trouble attracting potential developers sign up their beta platform. While website traffic saw a slight year-over-year increase, the bounce rate after the home page was a staggering 85%, well above industry average of 40%. This indicated a significant portion of visitors were leaving the site immediately without exploring Segmed’s offerings.

ObjectiveOur goal was to design and implement a user-friendly site that resonated with potential AI developers. To improve acquisition for potential developers at this early stage, we aimed to reduce the bounce rate by 20% and increase user signups for the beta platform by 15%, while enhancing the brand image.

This all needed to come together in 2 months for the 2021 HIMSS Conference, a global health conference with 19,300+ attendees, offering a prime opportunity for Segmed to showcase their new beta platform.


Discovery

NeedfindingTo understand our users, I took a crash course in radiology and AI fundamentals, complemented by in-depth interviews with Segmed’s partners and founders. These discussions revealed several challenges that Segmed’s two primary stakeholders face:
  • Data Scalability & Diversity: Regulatory constraints and fragmented healthcare data lead to significant research delays and data inconsistencies for AI developers.
  • Operational Barriers: Hospitals and medical centers, the primary data providers, required multiple meetings to ensure data compliance and security.


Initial Audit
Armed with these insights, I conducted an initial audit with our design lead, employing Google Analytics to identify underperforming pages and a heuristic review to assess usability and pain points



Key Insights 
Based on the audit, I discovered several key pain points for our data partners:
  • Inefficient copy: Current features mentioned on the home page are vague; it was difficult to know how Segmed is unique compared to its competitors.
  • Scattered resources: Due to lack of organization and dead ends, some of the main features / best selling points of the product were not mentioned on the home page.
  • Unclear CTA: The process of reaching out or contacting Segmed team members is unclear.


Design

User Personas
Based on previous research and documentation by Segmed, I created user personas. Our target audience included AI engineers, academic researchers, medical device manufacturers, and more.



User Journey
I created a user journey for one of our user personas, Rachel, an AI engineer at a healthtech company. I came up with 5 opportunity areas that could be applied to the website.


User Flow Diagram

Data-Driven Design RecommendationsConsidering researcher & developer needs, I created top 3 priorities for my designs:
  • Diversity of datasets: Diverse datasets in specific domains are hard to come by. Segmed offers data from over 40+ US states, 15+ countries, and 5 continents, which makes it stand out from competitors, and is continuously growing.
  • Standardization of data: Healthcare data is massively fragmented, which means there are high error rates and inconsistent patient identities. Segmed offers custom support from in-house engineers.
  • Filtering Capibilities: Most AI engineers looking for medical datasets already have established partnerships with healthcare partners. However, the speed and precision of data filtering remain key challenges.


Prototyping
I created 23 medium-fidelity wireframes to test with users. Focusing on the copy, I generated 3-4 options for our essential pages based on interal team feedback: 
1. Philosophy & Security
2. FAQ
3. Connect with Us

User TestingI recruited 10 participants: 5 healthcare administrators, 2 researchers, and 3 internal team members. We focused on tasks related to understanding Segmed’s data pipeline and the process and opportunities of partnering with Segmed after scaling to a high-fidelity Figma prototype. Following up with post-interview surveys helped us refine the website to effectively address their needs.

A/B Testing at HIMSS 2021
To determine the most effective hero card headline for attracting AI developers, we ran a small A/B test comparing two options at the HIMSS Conference:

Option A: Generate Revenue While Reducing Bias in Healthcare

Option B
: Generate Secure, Diverse, & De-Identified Data for AI Research 


The A/B test revealed that Option A outperformed Option B by 12%,  and published that headline on the website.


    Result

    The website redesign delivered significant results:
    - Bounce Rate Reduction: The bounce rate after the homepage dropped by 24%.
    - Increased Data Partner Signups: Data partner signups soared by 20%, surpassing our initial 15% target.



    Reflection

    Thankful for feedback from the C-Suite team at Segmed, and guidance from Akemi! In the future, I would have loved to test stakeholders who weren’t aware of Segmed but might be interested in the platform, as well as test on different devices across mobile, tablet, and desktop to address any device-specific issues. 

    The best way to complain is to design.

    Coffee chat?   LinkedIn     Email