Transitioning to Advanced ML-Powered Audience Segmentation
PUMA’s journey into advanced analytics began in August 2022 when they decided to harness Google Cloud’s machine-learning capabilities. While Google Analytics offered predictive audiences, PUMA chose to develop a custom ML model to address their specific requirements and gain deeper control over their data analysis.
The implementation leveraged several key Google Cloud products:
- Cloud Shell for framework setup
- Instant BQML for audience configuration
- CRMint for orchestration
- BigQuery for advanced analytics
The process involved BigQuery handling the modeling and machine-learning aspects, while CRMint managed data integration and audience creation within Google Analytics. The seamless integration with Google Ads enabled automatic sharing of audience segments for strategic activation.
Implementation and Support Experience
The deployment process was notably efficient, thanks to close collaboration with Google Cloud and gTech Ads teams. PUMA received comprehensive support and guidance throughout the implementation, along with extensive documentation that simplified the development process.
Impressive Performance Metrics
The results within the first six months were remarkable:
- 149.8% increase in click-through rate among top three audiences compared to traditional audiences
- 4.6% improvement in conversion rate
- 6% increase in average order value (AOV)
The solution also provided transparent, well-structured pricing, enabling better cost optimization and prediction.
Future Roadmap and Expansion Plans
PUMA’s future initiatives include:
- Developing advanced audiences using internal data and offline purchase information
- Scaling the solution to 20+ international entities
- Exploring server-side tagging using Tag Manager
- Implementing real-time reporting based on server-side data collection
Infrastructure Evolution and AI Integration
PUMA’s strategic plans involve:
- Migrating e-commerce infrastructure to Google Cloud
- Implementing event-driven architecture
- Restructuring data-management processes for AI operationalization
The success of this ML-powered audience segmentation project has demonstrated the potential of data-driven automation and strengthened PUMA’s confidence in machine learning implementations.
Learn more about PUMA’s journey with Google Cloud and BigQuery