Code Yellow: How we outperformed Epic and saved a $25M client
Cedar's biggest client threatened to churn, leading to a performance bakeoff with Cedar's biggest competitor
I lead the Design response to this - 25+ experiments over the course of 5 months. The result was outperforming the competitor by 24%, saved the client, and established a new rapid experimentation model.
Client
Cedar
Type
Cedar
Year
2025
Process
Identifying the opportunity
Data Science flagged that insured patients with large bills had the highest drop-off rate at the top of the funnel. Rather than redesigning the full experience, we focused on low-cost interventions: communications we could ship and test within days.
Running minimum viable tests
I designed rapid experiments using a test-and-learn framework, starting with one-off emails to small patient segments. Each experiment had a clear hypothesis, a defined audience, and a metric we were optimizing for: collections rate. We shipped 25+ experiments over 5 months.
Synthesizing research into experiments
I connected our Data Science findings to existing UXR on ALICE patients, asset-limited, income-constrained individuals facing impossible financial choices. This reframe shifted how we designed messaging to be more supportive and options-focused.
Iterating and scaling what worked
Winning experiments were refined and scaled. The flexible payment options email went from a single-segment MVT to a dynamic, personalized communication that adapted messaging based on balance size and number of bills, resulting in a +15.7% payment rate lift in the $500–$2,500 segment.
Enabling the broader team
Beyond my own experiments, I created a designer's manual to comms personalization, a Claude skill for email copywriting, and a legal x data guidance doc so other designers could run experiments faster and with fewer blockers.
Outcome
Cedar outperformed competitor by 24% in collections rate, securing a renewal. Across 5 months, we shipped 25+ experiments that moved the needle on patient engagement and payment behavior, proving that thoughtful, low-code design interventions can have meaningful business impact.
