Turning “We Think” into “We Know” through data-driven design and experimentation.
Website Optimization for CVS Health.
Mission
Enhance the digital experience for millions of CVS Health and Aetna customers by applying experimentation, analytics, and design to solve friction points and validate improvements.
My Role
As lead UX Designer on the Customer Experience Optimization team, I collaborated with product partners, developers, and analysts to translate insights into testable design solutions. My work spanned discovery, wireframing, high-fidelity design, and validation through A/B testing.
Results
Through hundreds of experiments, our team improved engagement, clarity, and conversion across key customer journeys — proving that thoughtful, data-driven design can deliver measurable business growth while simplifying the user experience.
FEATURED PROJECT
Reducing Confusion in Multi-Person Vaccine Booking
CUSTOMER PROBLEM
Patients were adding additional people to their vaccine appointment but failing to select vaccines for those added individuals. As a result, many users unintentionally progressed without completing key steps, which led to errors, confusion, and scheduling abandonment.
DISCOVERY
Finding the cause
A backend logic issue in the current production flow created confusion for users scheduling multiple patients. Users could add additional individuals to an appointment and continue without selecting vaccines for them.
This confusion was made worse by the error handling. The system did not display an error message for added patients when no vaccine was selected. Because there was no way to remove an added patient, users who changed their mind and no longer wanted to schedule a vaccine for that person would become stuck. This explains why an error dialogue was never implemented in production.
The issue became more visible on the confirmation page. Only the patient with a selected vaccine appeared, and there was no option to review or update the other patients or add missing vaccines.
IDEATION
Mockups
On the Optimization team, solutions must be delivered quickly and built with front-end changes only. This is a standard constraint for our work, since our experimentation platform operates entirely on the client side and cannot modify backend logic. As a result, any approach needed to introduce clarity without altering the core scheduling architecture.
Working within these boundaries, I explored several lightweight options that could guide users more effectively through the multi-patient flow. After ideating, I landed on four potential solutions, each with its own tradeoffs:
Disabled next button
Prevent users from advancing until all patients had a vaccine selected.
Concern: This would create the same problem as a hidden error message. If a user added a patient by mistake and could not remove them, the disabled button would trap them in the flow.
Add vaccine selection as part of the “Add Person” flow
Prompt users to choose a vaccine at the moment they add each patient.
Concern: This would introduce a higher level of engineering effort and would be inconsistent with the established behavior of the flow.
Clarifying copy on each patient card
Add instructional text to reinforce that each patient requires a vaccine selection.
Concern: Copy alone could easily be missed, and users would still be able to continue without completing all fields.
A confirmation modal when the user attempts to continue
Surface missing selections at the point of progression and give users a chance to correct them.
Concern: A modal is not strictly the most accessible solution, but it aligned with all other constraints and provided the clearest path for users to review and fix errors.
CHOSEN DESIGN
Confirmation modal
RESULTS
Success
After running for 19 days and reaching more than 430,000 visitors split evenly between test and control, the experiment showed a clear improvement in the multi-patient scheduling experience. Users in the test group were more likely to complete their appointments and less likely to abandon the flow. The confirmation modal gave them a simple way to review and fix missing information, which strengthened accuracy and improved overall confidence in the process.
+7.97%
increase in 2-patient appointments
+10.9%
increase in 3-patient appointments
+9.09%
increase in 4-patient appointments
60k-63k
estimated annual increase in group appointments
INSIGHTS & NEXT STEPS
What we learned
While the modal proved effective, it highlighted that the underlying experience still needs structural improvements. The test demonstrated that users require clearer guidance and more flexibility when navigating the multi-patient flow. It also exposed several weak points that can be addressed through future product enhancements.
Recommendations
1. Implement a proper error message
The modal acted as an indirect workaround for missing validation. A clear error message is the more appropriate and scalable solution to prevent users from skipping required vaccine selections.
2. Allow users to remove additional patients
Users need the ability to undo added patients so they do not become trapped if they change their mind or make an incorrect selection.
3. Enable editing on the confirmation page
Users should be able to modify vaccines and patient information before submitting. This would reduce abandonment and improve accuracy for those who notice issues later in the flow.
CONCLUSION
Ultimately, the modal was a meaningful step forward, but the larger impact came from what we learned. These insights now serve as a foundation for future iterations that will create a smoother and more intuitive multi-patient flow.