SigmaAnalytics & Tech
H5PythonAdvanced

Customer Churn Prediction

Keeping a customer is far cheaper than winning a new one, so predicting churn — who's about to leave — is one of the highest-value tasks in applied data science. This capstone runs the full ML workflow on a telecom-style churn problem, but with a crucial business twist most tutorials skip: the default 0.5 threshold is almost never right.

Notebook

Enrol to read the notebook and download it to run.

What this module covers

Keeping a customer is far cheaper than winning a new one, so predicting churn — who's about to leave — is one of the highest-value tasks in applied data science. This capstone runs the full ML workflow on a telecom-style churn problem, but with a crucial business twist most tutorials skip: the default 0.5 threshold is almost never right.

By the end of it

  • Who churns, and what customer characteristics predict it?
  • Can we build a well-calibrated, honestly-evaluated churn model?
  • Given the economics, what probability threshold should trigger a retention offer?
  • Which levers (contract type, tenure, charges) should the business pull?