A Customer Journey Mapping Approach to Improve CPFL Energia Fraud Detection Predictive Models

Non-Technical Losses have a profound economical impact in distribution utilities; hence, reducing them and pursuing revenue recovery makes up for an essential means to secure utilities’ financial health. In that regard, this paper proposes a compound method that combines a Logistic Regression predictive algorithm with the concept of Customer Journeys, in order to enhance the accuracy of the fraud detection method traditionally applied in CPFL Energia (a major Brazilian utility). Every interaction of each CPFL’s customer is taken in consideration to track fraudulent customers record and identify suspicious patterns, with the purpose of adding new data to the present model and better refining its outcome, thus reducing false-positives and achieving both greater accuracy and recovered revenue amounts, besides better operational efficiency. The proposed methodology was field tested, resulting in a 2.2 times greater fraud inspection success rate, and the model was effectively implemented on the internal processes of the company’s department responsible for fraud detection and revenue recovery.