Case Study: Merkle achieves faster ML model runtimes, expanded modeling capacity, and lower cloud costs with Qubole

Merkle’s Data Science Group Increases Capabilities While Reducing Model Runtimes And Cloud Costs With Qubole

Merkle, a global data-driven marketing agency, faced growing pressure to advance its machine learning for audience creation as competition commoditized basic targeting. Their on-premises environment struggled with massive, multi-million-record datasets, producing model runtimes of about a day and stretching end-to-end projects to as long as three weeks—too slow for clients that need near real-time insights.

By adopting Qubole’s cloud data lake platform, Merkle gained autoscaling compute, faster runtimes (models reduced to roughly 5–6 hours), and the ability to run and compare many more models quickly. The move cut project durations to about four days, lowered compute costs, enabled large-scale reporting (up to 400 million records), sped onboarding, and delivered client impacts such as ~25% lift in sales, ~26% revenue increase, and ~25% reduction in analysis costs.

Merkle

Luke Berszakiewicz

Senior Manager of Data Science