Guide to Customer Retention Analytics- Part 2: Telecom Case Study

This post is a continuation of an earlier post on Customer Retention Analytics by our guest bloggers Sandhya Kuruganti and Hindol Basu, authors of a book on Business Analytics titled “Business Analytics: Applications to Consumer Marketing”. The first part of the post Guide to Customer Retention Analytics- Part 1 focused on understanding retention, attrition and the various retention strategies used by marketers. Part two will explain retention analytics, along with a Telecom Case Study that showcases the Proactive Approach to Retention Management using Churn Model.

Telecom Case Study: Proactive Approach to Retention Management using Churn Model

Business Scenario

• The telecom industry has become very competitive with price cutting and service enhancement. 
• Subsequently annual churn rate have doubled from 18% to 36%. This implies that average tenure of a telecom subscriber has halved and the telecom subscriber is losing more than 1/3^rd of its subscriber base between the beginning and end of the financial year.
• The telecom subscriber is aggressively looking to arrest churn proactively

Objective

• Predictive analytics is required to enable proactive retention efforts to reduce telecom churn. A churn model outputs a score which measures the likelihood to churn in the next one month
• The churn model will prioritize subscribers  in order of their propensity to churn that would be used for focussed targeting of “high risk of churn” subscribers
• Exploratory business analysis and segmentation is done prior to deciding on the sample for model development

Segmentation:

• Pre Paid and Post Paid: Pre Paid has significantly higher churn rates than Post Paid. However, ARPU for Post Paid is 5x that of Pre Paid. ARPU is the average revenue per user.
• Pre Paid Segment: This segment comprises 70% of the portfolio and contributes to 80% of total churners. Annual churn rate is 41%.
• Post Paid Segment: This segment comprises 30% of portfolio and contributes to 20% of total churners. Annual churn rate is 24%.
• Separate models may need to be developed for Pre Paid and Post Paid separately as these segments behave differently (churn, revenue and usage). Performance of a single model may be sub-optimal.
• The model is first developed for the Post Paid segment as the telecom subscriber’s priority is to curtain potential value loss followed by potential subscriber loss. 
• Subsequently, the Pre Paid segment model would be developed.

Data Preparation for Churn Model (Post Paid Segment)

Telecom Churn Model:  The following table shows the potential predictors of Churn and sign of relationship. Logistic Regression is a popular statistical method that is used.

Implementation:  Based on the churn mode l, a cut-off for the score can be decided. Subscribers exceeding the cut-off should be considered for contact. A suggested strategy could be:

• Segment 1 has a churn rate  which is about 2x the current  sample churn rate
• Campaign budget permits contacting 40,000 subscribers  in a month for a proactive anti-attrition strategy
• Significant marketing saves, to the tune of 50%. The telecom service provider would have had to contact 80% of the subscribers to reach out to 80% of the total potential churners. Now they have to contact only 30% using the churn model.

Sandhya Kuruganti and Hindol Basu are authors of the book “Business Analytics: Applications to Consumer Marketing”, recently published by McGraw Hill and available on Flipkart and Amazon India/UK/Canada.  Jigsaw students can avail of a discount of 20% with a coupon code at McGraw Hill website (valid until Nov 30, 2015).  . The authors are seasoned analytics professionals with a collective industry experience of more than 30 years.

Suggested Read:

Using Big Data to Improve Customer Experience

Jigsaw Academy Announces the Launch of its Online Certification Course in Retail Analytics