Challenge:

A global corporate trade company offers advertisers cost-effective national coverage at aggregated local rates. The company needed to improve their media buy process. Their buying team runs multiple scenarios to meet customers’ specific criteria. Speed of execution, turn-around time, and operational costs are paramount to client’s business. As their business grew, they needed a better way to plan and make their media buys.

Solution:

Converge migrated several Machine Learning algorithms to the Google Cloud Platform (GCP) and deployed them across clusters of GKE instances, Google’s Kubernetes cloud services. Converge data scientists optimized several machine learning models, deployed on GCP using a genetic algorithm configuration for maximum insight. This migration directly addressed the client’s challenges.

Results:

With the custom-built Machine Learning programs, the client increased what-if scenario computation from two per day to over 200, ultimately increasing productivity and profits. Their switch to GCP resulted in significant operation cost savings of more than five times what they would have seen with their previous platform. Cost per prediction were reduced by 5x as a result.