Predicting Customer Churn

Student Results:

http://predictive.analyticsight.com/predicting-churn-with-decision-trees/

Introduction.

Endurance International Group (EIG) serves millions of customers with a variety of web hosting services. Given the nature of the business, customers come and customers go. Regarding the latter, in an ideal world, EIG would know which customers were considering leaving, and when. With that information, EIG would be able to reach out to these customers to prevent unprofitable churn. We will build a predictive model to hopefully help EIG get closer to this ideal world.

Assignment Details.

Each group must perform the below tasks. Upon completion, the group must write a short blog post describing their work. The intended audience of the blog post is a manager at EIG.

Decision Tree:

Task: Build a decision tree model that predicts customer churn. Explore a variety of variables and tree settings, on both “training” and “validation” data sources. Your final model should balance accuracy and complexity. Be sure to include a screen capture of your final tree (or build your own display).

Benefit: The management at EIG can utilize such a model for predicting customer churn and/or identify potential predictors of churn.

Profit Benefit/Cost Analysis Sheet:

Task: Create a google spreadsheet that demonstrates the value of your tree model. The sheet should contain:

  1. a clearly labeled probability matrix (rows = predicted, columns = actual)
  2. equations that calculate the expected profits from a 3 business strategies (interventionist, selective interventionist, non-interventionist)
  3. inputs for revenue/cost associated with customers staying/leaving and interventionist actions
  4. input for impact of intervention on the likelihood of customers staying
  5. Instructions on how to use the sheet

Benefit: The site owner will be able to identify the usefulness of such a tree with a variety of assumptions concerning interventionist actions.