Sentiment Analysis

Problem Statement


Client wanted to develop a early attrition indicator for their customer base so that targeted campaigns can be designed with customer retention standpoint.




Exhaustive Brainstorming was performed to chalk likely behaviors which customers would showcase before attrition which were validated using bivariate testing.


History of Customer’s Relationship with the Client like loyalty behavior, purchasing patterns, Interaction with customer care, store visits, Payment Patterns were found to be influencing customer’s decision to attrite.


Logistic Regression Model was built for customer base snapshot as of 2 years back keeping the attrition flag as the dependent and the earlier mentioned factors as Independent factors.


The accuracy of the model was assessed by checking the model results against the 30% customer base (Control Group)




~13%(~1.5M Customers) of the existing customer base was found to be likely attrition candidates Upon deep dive analysis, the effect of attrition was found to be prominent in the urban areas especially New York, California, San Francisco etc. Marketing Team was given the responsibility to design targeted campaigns to retain these customers