Popular searches
Not sure where to start?Talk to our team about running sovereign AI in your environment.
↑ ↓ Navigate↵ OpenEsc Close
Epsilon lifted direct mail response rates 3 to 5% and added 15,000 relevant customers per campaign with H2O-3
Epsilon's Abacus data cooperative builds over 100,000 machine learning models a year to find consumers likely to respond to catalog, product, and promotional offers. Its legacy SAS system hit throughput and scalability limits on terabyte-sized datasets, and tightening privacy rules made model interpretability essential to prove list selection logic and prevent bias.
The legacy SAS-based system, dating to the 1990s, ran into throughput and scalability constraints when processing terabyte-sized datasets. As privacy regulations tightened, model interpretability became just as critical, because Epsilon needed to prove its list selection logic and prevent bias in the audiences it created.
Epsilon partnered with H2O.ai to build its "Accelerate" machine learning system on H2O-3. Proprietary feature engineering feeds a feature reduction step in pySpark, and clusters provision automatically, scaling computing resources up and down based on each job's requirements.
H2O-3's algorithms were chosen specifically for their ability to handle large datasets while providing the deep model interpretability that transparent, defensible list selection requires.
The Accelerate system improved direct mail response rates by 3 to 5%, adding roughly 15,000 highly relevant customers to each marketing campaign. For one major gift and B2B retailer, a 1.10% response rate increase on a single campaign generated $9.0M in incremental revenue.
The platform now supports 8,000 campaigns annually while giving Epsilon the transparent list selection its privacy obligations demand.
“In the environment of increasing privacy regulations, the tools Epsilon employs need to provide transparent list selection. H2O.ai help us meet this challenging and critical need.”Andrea ThorntonVP of Analytics
Epsilon is a global advertising and marketing technology company, part of Publicis Groupe, known for its data intelligence and customer insights, loyalty and CRM platforms, and data-driven creative and activation. Its Abacus data cooperative and identity assets help brands deliver personalized experiences at scale, drawing on a database of more than 250 million U.S. consumers and sending over 71 billion personalized emails each year.
More on the company is at https://www.epsilon.com.