Case Study: Epsilon boosts marketing ROI — +15K relevant customers per campaign and $9M incremental revenue — with H2O.ai

Epsilon Increases its Customers’ Marketing ROI with H2O.ai

Epsilon, through its Abacus data coop, faced a scalability and accuracy challenge: building more than 100,000 custom marketing models a year on terabyte-scale data while maintaining model interpretability and bias controls. To modernize its long-running SAS workflow and improve list accuracy, Epsilon partnered with H2O.ai and adopted H2O.ai’s Accelerate machine learning platform to support large-scale, explainable model building.

Working with H2O.ai, Epsilon implemented a pipeline of proprietary feature engineering, feature reduction in PySpark, and automated job provisioning to H2O.ai’s ML infrastructure, producing multiple models per campaign and combining them into optimized audience lists. The H2O.ai solution drove a 3–5% lift in direct mail response rates (adding about 15,000 highly relevant customers per campaign), produced a 1.10% uplift that generated $9.0M incremental revenue for a single campaign, and scales across roughly 8,000 campaigns a year.

Epsilon

Andrea Thornton

VP of Analytics