Telecom Churn Analysis
Customer Churn is the percentage of customer that unsubscribed or stopped using companies product or services during a certain time. Reason for customer churn can be anything like quality of product, price, customer experience and so on. Nowadays, companies are spending most of their resources to stop customer churn.
This case study is part of Udacity Data Scientist Nanodegree program. I will be using CRISP-DM technique and it involved below process.
- Business Understanding
- Data Understanding
- Data Preparation
- Modelling
- Evaluation
I am using Telecom open source data from Kaggle and data can be found from this link. By end of this study I am going to answer below business related questions using given Telecom dataset.
- What is the overall Churn Percentage ?
- Which gender category has more churn rate?
- Which Internet service users has more churn percentage?
- Is high monthly charge leads to high churn ?
- What is the survival curve/probability for the given customer?
Let’s explore the data
Dataset consist of customer information related to their demographic, internet services and account management. It has 7043 rows and 21 columns.
What is the overall Churn Percentage ?
It’s good to know overall percentage for any data analysis and below fig shows that dataset has 26.54% churn rate.
Which gender category has more churn rate?
Telecom industry has different set of products and services can be used by any gender category. There is no big difference in service distribution and churn percentage for Male and Female. Whereas, 26.16% customers churned were Male and 26.92% customers were from Female.
Which Internet service users has more churn percentage?
Customers can use single service or multiple services and they only use just use phone service or bundle with internet services. Fig 4 shows % of distribution by Internet service. 44% customers using Fiber optic internet service, 33.4% customers using DSL service and remaining 21.7% customers are not using either of them. Fig 5 shows that percentage of customer churn by broadband(internet service) users, practically Fiber is high speed internet but churn percentage is more Fiber optic service with nearly 42%.
Fig 6 shows that churn rate for additional services by thier types. Churn % is similar in customers with multiple or without multiple services.
Is high monthly charge leads to high churn ?
Yes, it looks like that higher % of customers churn when the monthly charges are high. Churn Percentage is more for the customers who paid monthly charges more than 60$.
What is the survival curve/probability for the given customer?
Survival rate is probability of no churn in a given length of time. It is calculated as the number of subscriber retaining divided by number of subscribers at risk of leaving. Fig 8: shows that survival probability for the given customers. Nearly 85% of customers survived first 12 months, more than 65% of customers survived up-to 6 years and probability is decreased thereafter.
We can see that some variables have a negative relation to Churn, while some have positive relation. Negative relation means that likeliness of churn decreases with that variable.
In conclusion, we can say that gender category is nothing to do with churn rate, high monthly charges and fibre optic internet services leads to high churn rates. This is interesting because although fibre optic services are faster, customers are likely to churn because of high price or competitive promotion from other service provider.
This analysis is not enough to decide what is right and what is wrong, we need to explore more to get clear understanding of churn.
To see more about this analysis, please use the link to Github page.