To measure the impact of a rare disease patient campaign without an A/B test, Good Apple leveraged Harvest, a proprietary probabilistic model that provides directional guidance for media optimizations. By evolving Harvest to assess the marginal impact of increased media spend, we identified search as the highest-performing channel, delivering a 31:1 marginal ROAS. These findings shaped a more efficient media strategy, prioritizing search and display to maximize patient engagement. The results were further validated by a third-party Media Mix Model, confirming the effectiveness of this data-driven approach in driving real-world impact.