Executive Summary
When a global corporate trade company outgrew a legacy media-buying platform that could run only two “what-if” scenarios per day, it partnered with Pellera Technologies to modernize its machine learning workloads on Google Cloud. The result: scenario capacity jumped from two to more than 200 per day, operational costs dropped more than 5x, and the media planning and buying team gained the speed and accuracy to deliver higher-performing plans at scale.
The Challenge
In the fast-moving world of media buying, speed, accuracy, and efficiency define competitive advantage. A global corporate trade company known for helping advertisers achieve national reach at scaled local rates faced substantial operational pressure as its business expanded. The organization’s media planning and buying team relied heavily on running multiple scenarios to meet each client engagement’s specific criteria.
However, the legacy system supporting this process could no longer keep pace with rising demand. As the volume of client campaigns grew and expectations for faster turnaround intensified, several challenges emerged: rigid computational limits capped the team at just two scenarios per day; scaling the existing platform proved cost-prohibitive; scenario runs were slow and resource-intensive; and outdated models and aging infrastructure made it difficult to adopt more advanced analytics.
To remain competitive and deliver the flexibility clients demanded, the company needed a strategic machine learning optimization initiative, one that would expand scenario-modeling capacity, boost performance, and reduce costs.
The Approach
To remove the constraints of an aging on-premises environment, the company partnered with Pellera to design a machine learning optimization and cloud migration strategy built on the scalability and advanced tooling of the Google Cloud Platform (GCP). The plan centered on migrating core ML workloads to GCP, modernizing the underlying models, and enabling rapid, high-volume scenario modeling.
The Solution
A three-step strategy migrated key workloads to the cloud, modernized the models, and unlocked fast, high-volume scenario modeling:
01
Migrate ML Algorithms to GCP
Moved core machine learning algorithms to Google Cloud, immediately improving processing power, availability, and scalability.
02
Deploy Across GKE Clusters
Ran the models on Google Kubernetes Engine clusters for faster execution and consistent performance under heavy loads.
03
Genetic Algorithm Optimization
Applied advanced tuning, including genetic algorithm configurations, to sharpen predictive accuracy and analytical clarity.
The Results
The initiative reshaped the company’s capabilities through successful ML optimization and cloud migration:
200+ scenarios per day
Up from just two per day, letting the team explore far more media-plan configurations and respond to clients faster.
5x lower operating costs
Cloud-native architecture cut infrastructure spend, cost per prediction, and total overhead by more than 5x.
Higher productivity
Faster insights and a robust ML foundation boosted the team’s output and decision-making speed.
Improved profitability
More efficient, accurate media plans drove stronger, more profitable client outcomes.
These gains highlight the impact strategic ML optimization and cloud migration can have for organizations that depend on complex, high-volume modeling.
Conclusion
If you’re looking to elevate analytics performance, reduce ML operating costs, or modernize outdated workloads, the team at Pellera is primed to deliver maximum results for your company. Learn more about our Data & AI services and schedule a consultation today.
