Reduce Customer Churn by 40%

Reduce Customer Churn by

0
%

Enhanced churn models helped the client to reduce churn in their most valuable customer segments by more than 40%.
For data monetization the accuracy in prediction of the subscribers’ profile was over 96%.

The challenge

Our client a large communications service provider with USD 5 billion revenue was facing a subscriber churn problem when mobile number portability was introduced. In addition, it was observed that this was higher in the high value segment. Efforts to proactively identify a potential churner were not yielding good results and the revenue was declining.

Client was also running data monetization initiatives and due to lack of demographic profile information for the prepaid subscribers, the campaigns associated to data monetization were not delivering good uptake for their partners.

Our solution

CoreCompete deployed a Customer Analytics Solution built on the SAS analytics platform. We developed powerful analytical machine learning model utilizing SAS Enterprise Miner to predict customer churn. Subscribers’ usage data and trend in change of service affinity information were used to predict likelihood of churn. The robustness and performance of the model developed was validated by scoring subscribers’ probability to churn using multiple out of time sample data across two years. The model helps in running retention campaigns with lesser costs targeting only potential churners.

We also developed multinomial logistic regression to predict pre-paid subscribers’ demographic profile such as Nationality, Gender and Age to design campaigns with relevant offers. This resulted in increased campaign responses enhancing the attractiveness of the client’s data monetization platforms.

We used Greenplum native programming ability to process the data and to automate periodical scoring of all models saving the time taken in data movements.

Our site uses cookies to improve your experience. You can control cookies by adjusting your browser or device settings. By using our site and closing this box, you consent to our use of cookies.