Wes Gahagan and Abilash Gatti, Verizon
The problem isn’t that companies are lacking data these days. The problem is capturing the right data and sifting through all of the normal data to find the outliers. Wireless phone companies face the same challenges that most large corporations experience. A multitude of Full and Self-service channels exist for each business event. The number of unique Transaction/Channel Types each day is in the thousands and tens of millions of transactions are processed daily. The challenge facing many companies today is that slight deviations in tracking metrics go undetected and can result in missed business opportunities or unidentified problems. Companies need to gain insight into outlier trends to properly prioritize fixes and enhancements.
Implementing a Business Transaction Outlier and Anomaly Detection System helps companies optimize and prioritize fixes and enhancements. In this session, we’ll review our method of tracking business events by channel and then ingesting these events into the RapidMiner Outlier/Anomaly detection process. We’ll discuss the benefits of tracking Outliers across tens of thousands of potential combinations and storing and aggregating this information. We’ll also show how we display the Outlier information in dashboards that are easily understood by technical and business teams. We know that spotting outliers simply isn’t enough. Verizon’s Outlier detection system, uncovers anomalies, and then allows us to deep-dive into actionable insights instead of just looking at seemingly meaningless data points. We’ll also recap the challenges and lessons learned during this process, as well as, two other Verizon Wireless RapidMiner projects (Predictive Churn and Promotional Impacts on churn).