Churn Modeling and Acquisition Fraud Modeling for a Telecom Client
C H U R N
M O D E L I N G &
A C Q U I S I T I O N
F R A U D
M O D E L I N G
T E C H N O L O G Y
ETL – Hadoop
Data warehouse – Amazon
Reporting – Tableau
Models – Spark ML,
Regression model, K-mean
O B J E C T I V E
To predict subscribers who have a high propensity to churn
(voluntary/rotation), so that relevant marketing and
retentions strategies can be executed.
Identify retailers who are acquiring ‘Born Dead’ subscribers
who have zero or near zero usage post activation.
To cluster subscribers based on usage pattern to identify
groups of similar subscribers.
Circle wise cross-sell model to identify subscribers with a
higher probability to subscribe to VAS services.
A P P R O A C H
Built separate logistic regression models to predict churn and
cross sell for the 13 telecom circles of the client.
Analyzed 6 months of daily customer CDR data to build 250+
customer metrics for each of the circles.
Used K-Mean clustering to group customers based on 6 month
usage patterns as well as rolling weekly usage behavior post
O U T C O M E
Targeting based on regression model helped a more focused
campaign and increased retention and cross sell rates.
Identification of retailers selling ‘fraudulent ‘SIMs resulted in
huge cost savings on retailer commission.