Customer churn analysis and save-desk playbook for a telecom operator

Borealis Telecom sells mobile, home internet and TV to 1.2 million customers. Monthly churn has risen from 1.6% to 2.1% in a year, mostly among month-to-month internet customers. The retention team, called the save desk, handles about 9,000 cancellation calls a month and offers whatever discount the agent thinks will work. It saves about a third of callers, but the finance team suspects many of them would have stayed anyway, and that the discounts cost more than the revenue they keep.

The customer director has asked you, as a customer analytics associate, to answer: - which customers are most likely to leave, and what is associated with it; - what each saved customer is worth, and which offers pay for themselves; - how the save desk should decide what to offer, to whom.

You will use the Telco Customer Churn data set published on Kaggle (an IBM sample of about 7,000 customers with contract type, tenure, services, charges and churn). Treat it as a representative sample of Borealis customers and say so in your report.

You will audit and explore the data, build and evaluate a churn model, calculate customer lifetime value and the economics of save offers, write a save-desk playbook with offers by segment and guardrails, and design a pilot that proves whether the playbook beats today's approach.

Done means the customer director can see which segments drive churn, what a saved customer is worth, a playbook agents can follow on a call, and a pilot with a control group that will show, within eight weeks, whether the save rate and net revenue improve.