Customer Churn Analysis

Purpose

Find out what are the factors that affect the customer churn more, and what we can do to decrese the churn.

Analysis

I found that some columns have dependencies on the other column. For example, a couple of columns below are applicable only if the Internet Service is enabled.

I excluded those columns and ran the variable importance analysis.

Next, I ran the decision tree to see more detail. I picked only the columns that have higher importance. I excluded tenure and charge related columns because those are the results and not controllable. The result is following.

The biggest factor is the contract. So the first thing we have to do is to run a campaign to covert customers from Month-to-month contract to One/Two-year contract.

The second biggest factor is the Internet Service. Somehow customers with the faster (Optical Fiber) internet connection churned more. I quickly checked the numbers and found that those customers are the majority so we have to do something to stop the churn.

Numbers:

Ratio:

I ran the logistic regression with the Internet Service related service columns that I exlucded at the beginning.

Customers with Internet Service churned more if they used TV/Movie Streaming services. On the other hand, customers with Online Security, Tech Support, Online Backup and Device Protection churned less. So, we need to focus more on selling those services instead of TV/Movie Streaming services.

Conclusion

• We can run a campaign to covert customers from Month-to-month contract to One/Two-year contract.
• We can focus on selling Online Security, Tech Support, Online Backup and Device Protection services more for the customers using the Internet Service. We should tone down the TV/Movie Streaming services.