Case Studies

Predictive Analytics Scorecard: Telecom Collections


Objective

Identify the Risk behavior of customers who are likely to default within the next six month period of coming into Collections. Analyze using a segmented approach by forming sub populations within the customer universe.

Extremely pleased with the numbers. Thanks for the hard work, my client is super happy because of you.
- Director, Telecommunications Consulting Company, USA.

Background and Challenges

One of the top telecom services provider of U.S. facing challenges in collection of payments with high default rates. Customers are defaulting leading to high delinquent payments that needs to be taken care by the service provider.  Many even opt for disconnection while raking huge bills that will never be paid. Bills that intentionally never meant to be paid. The company decides to get control over the bankrupt customers and at the same time disrupting the losses accumulating month on month.

The company had a large portfolio with following portfolio:

· 16 million accounts,

· 450+ Attributes

· Bad Rate in population = 7.5%

· Consumer & Business Accounts

Collection Scorecards, Recovery Scorecards, Early Warning Scorecards

 

Our Approach

 

Project Scope: Data Dictionary, Data Audit - Bivariate Analysis, Features Creation, Sub-Population Analysis, Feature Selection, Model Results and Validation, Final Implementation Reports. 50% of the data was used to build the model while remaining was preserved for model validation.

Method: Logistic Regression, Cramer’s V.

Analyses Performed: Univariate, Bivariate, Binning.

Methodology: The project begins with building a clear understanding of the customer attributes. This is comprehensively done by building tools for Univariate and Bivariate Analyses.

 

Collection Scorecards, Recovery Scorecards, Early Warning Scorecards

Collection Scorecards, Recovery Scorecards, Early Warning Scorecards

The variables were cleaned, transformed, and binned to derive the maximum value out of the logistic regression - which was the main technique to model bad rate or delinquency rate.

Collection Scorecards, Recovery Scorecards, Early Warning Scorecards

 

Collection Scorecards, Recovery Scorecards, Early Warning Scorecards

Features Creation & Selection: Creating statistically significant variables, we are able to get meaningful variables that are also hugely correlating to the bad rate.

Collection Scorecards, Recovery Scorecards, Early Warning Scorecards

Results: Excellent results were obtained for all four sub-populations. Top Deciles were able to capture almost 80% of the bad customers and validation data nicely with the training data.

Collection Scorecards, Recovery Scorecards, Early Warning Scorecards


You may save a copy of this case study:

Contact Us

• Collection Scorecards
• Recovery Scorecards
• Cross Sell Scorecards

Feel free to give us a call for more information on our services:

White Capers

(+91) 11-41010146
info@whitecapers.com