Telco Customer Segmentation

Problem Background

The telco company experiences a significant amount of churn and is interested in knowing more about the churn customers, to better understand their profiles.

Data

In this project, we derive insights from data on telecom customers from profiling the churn customers. The data for this project was sourced from kaggle and basic preprocessing done to ensure data is fit for purpose. 

Approach

The K-Means algorithm was used to cluster the features. 

The elbow method identified two clusters.

Insights & Recommendations

Two main profiles on churn customers, which I termed “Free agents” and “Somewhat free agents“. 

 

The free agents have majority unmarried people with no dependents. With the somewhat free agents, more than half married with a third having dependents. 

Internet use: significant use of internet for both groups with higher use of fibre among the somewhat free agents.

Contract tenure: both groups are big on short-term contracts with the longest contract being one year, for the somewhat free. 

Streaming movies: a significant number of the somewhat free stream movies, unlike for the other.

Recommendation

Provide more value added services in the areas of international plans, internet service, online backup, movie streaming support.