Improved User Experience Through Real-Time Monitoring: Telefónica Relies on GCP and AI
Intro
e-dialog implemented a solution that analyzes website data in near real-time in Google Cloud and automatically notifies O2 Telefónica when an issue occurs. The approach consists of two pipelines developed in GCP: a training pipeline and a real-time prediction pipeline.
Challenge
Timely detection of outages or issues and subsequent resolution of these problems are critical for smooth website operations and an excellent user experience. The earlier an issue is detected, the faster it can be resolved. Traditional manual methods are too slow to respond to sudden changes in user behavior, which impacts response times and user satisfaction.
Our Google Cloud-based anomaly detection solution has significantly improved our ability to quickly identify and resolve issues, minimizing disruptions and ensuring a smooth, uninterrupted experience for our website users. This proactive approach helps us maintain a high level of reliability and user satisfaction.
Gerald Maier
Senior Analytics Engineer, Telefónica Germany
Measures
The developed training pipeline transforms historical Google Analytics 4 data in BigQuery and trains an ML model to determine expected behavior for various metrics, such as traffic & product page views. The model then continuously analyzes data in near real-time to estimate the probability of anomalies. The system alerts stakeholders and provides additional statistics via the Looker Studio dashboard.
Results
The implementation of the Google Cloud monitoring and anomaly detection solution has improved website uptime and reliability by detecting changes in near real-time and immediately notifying the app and web teams. Faster response times have enhanced our ability to proactively address website issues, resulting in higher user satisfaction and reduced potential revenue loss.
6-12x
faster response time
Engagement
> 62,000 clicks
Efficiency
Relevance
1.68% click-through rate
- x faster response time
- clicks + engagement
- % click-through rate + relevance
