Churn and Loyalty Analytics
Customer Churn or Customer Attrition analysis is one of important business activities for Banks, Telecom companies , Retails firms, financial services institutions and Insurance companies for single most important reason that cost of acquiring a new customer is far greater than cost of retaining an existing customer.
Background and Challenges
Our client, a leading telecom company, wanted to use analytics to arrest the increasing churn in their prepaid telecom portfolio. The Customer Relationship Management team (CRM) wanted to be able to use their available resources optimally and reduce customer churn at the minimal cost. The challenges were – to identify subscribers
who were most likely to abandon the service as well as to tag subscribers as “one time subscriber” or ” subscribers who have reached the point of no return” so that customer service could be focused on those subscribers who were most likely to continue the services.
The objective of Churn and Loyalty Analysis Model was identified as follows:
- Early identification of subscribers who are more likely to Churn
- Identification of important factors driving the churn
- Optimization of operational costs
The key challenge was to define the “churn” and to understand key drivers of Customer churn. The data sources that were identified to use for Churn and Loyalty Analysis were – subscriber Demographics, Subscriber Spend, Call Center Records, Service Request Records, Vintage of customers, activity/Non-Activity of the customers, Subscribers’ usage history and product/packages type characteristics. We tried to identify the optimal days after lapse in which churn can be reactivated (Win-back Opportunity) and optimal days after lapse in which there is minimal chance of reactivation (Point of no return).
We observed that about 94% of customers who are not using the prepaid services from last 7 days are proceeding to service termination. Based on this observation, 7 days was defined as opportunity to win-back and the subscribers with more than 7 inactive days were tagged as gone into no return zone. The target was defined to identify subscribers who are most likely not to use their pre-paid service in next 7 days.
The important attributes/ Key drivers of churn behavior were selected using various statistical techniques like clustering, principal component analysis, correlation etc. Based on Dichotomous nature of the dependent variable, Logistic Regression technique was selected as the most suitable model technique.
The Predictive model thus developed was used to make their Churn and Retention campaign more scientific in approach and system driven. After the model was implemented in the CRM system, the scorecard is used to score the active portfolio every fortnight (1st and 15th of the month). The scores are used to identify the segments with highest propensity to churn in the next 15 days (between 15th to 30th and 1st to 15th respectively). The churn management and retention teams have 15 days window to run segment specific churn prevention campaigns and reduce voluntary churn.