Value from Predicting Customer Churn

Predicting customer churn allows businesses to leverage predictive analytics to classify customers based on how likely they are to churn.  You can divide your customers into segments.  Then get very granular such as calculating each customers probability of churn.  By using historical data from your operational systems you can mine the data to create predictive classifications.  Then using this information you can modify your operational processes to take advantage of this new insight.  And determine how best to handle your customer base customized to each customer.  While data quality is important.  When you first get started building predictive modeling you should start with the data you have.  It is more important to build a model, filter out bad data, and learn about your data and how it can improve your operations.

Using data from a fictitious telecommunication company.  You will see how data mining works to determine factors that impact churn based on each customers unique features.  The dataset contains 7000 rows with the following columns.

Predicting Customer Churn Data

After running the data through a Decision Tree algorithm the data is split to show the impact of the data on churn.  By reviewing the tree you will see customers on the Month-to-Month plan with Fiber Optics have the highest probability of churn.  While customer on the Month-to-Month plan with no Internet Service has the least probability of churn.

Predicting Customer Churn Decision Tree

With just level of insight you can begin to further investigate the difference in customers that use fiber optic and no internet services.  Are there factors in the products or customers that are leading to the higher attrition in fiber optic customers?  How do fiber optic customers on a Month-to-Month contract compare to other customers not on a different contract?