Churn No More: Mastering Customer Retention in the Telecom Landscape
For my third project at Moringa, I analysed data for a telecommunications company, SyriaTel. The dataset was retrieved from Kaggle, and you can access it using this link.
The main objective for the project was to develop a binary classification model to predict whether a customer of SyriaTel is likely to stop doing business in the near future.
My goal as the data scientist was to therefore, identify predictable patterns in customer behavior, in order to help the company reduce financial losses associated with customer churn.
Understanding Customer Churn
In the highly competitive telecom industry, retaining existing customers has become crucial due to customers having numerous options and increasing expectations. Churn not only leads to immediate revenue loss but also incurs significant costs for acquiring new customers. It is essential for telecom businesses to understand the factors contributing to churn and accurately predict it to develop effective retention strategies.
To effectively achieve the project objectives, I conducted a comprehensive business understanding and found that the current churn rate for SyriaTel was 15%, which is twice as high as the acceptable rate of at least 7%.
This led me to develop the following objectives that guided my analysis:
- Develop a highly accurate binary classification model that predicts customer churn for SyriaTel.
- Identify predictable patterns and insights in customer behavior to proactively identify customers at high risk of churning.
- Enable SyriaTel to optimize retention strategies, allocate resources effectively, and minimize financial losses associated with customer churn.
Building the Churn Prediction Model
The model evaluations will involve comparing their performances based on the Recall (Sensitivity) metric.
In the context of customer churn prediction, recall is preferred because it focuses on minimizing false negatives, ensuring that high-risk churn cases are not missed.
I tested the following models in order to maximise recall:
- Logistic regression model
- Decision Tree Model.
- Random Forest Model
- Support Vector Machine
- XGBoost
The best performing model was the Random Forest (GridSearchCV model) with a recall score of 74.77%. It outperformed the other models in all other metrics as well.
The features that hold importance are illustrated in the graph below:
Identifying Predictive Patterns
Some of the the key features that were shown to influence whether a customer would churn or not included:
- Total Expenditure
- Customer Service calls
- Total day charge
- Total international charge
These features provide valuable insights for SyriaTel to investigate further and improve their customer retention rate.
Proactive Retention Strategies
Based on the analysis I undertook, I would suggest that SyriaTel takes into consideration the following recommendations.
- Affordable Service Costs: The company should consider adjusting some of their service charges or introducing alternative payment plans to accommodate customers who don’t use their phones as much. This can help retain customers who may be looking for more cost-effective options.
- Invest in Quality Customer Care: Ensuring customers receive excellent customer service can play a significant role in maintaining or increasing the brand’s reputation, which may ultimately reduce the churn rate. Investing in training for customer care representatives can lead to improved customer satisfaction and loyalty.
Next Steps & Conclusion
In conclusion, SyriaTel should focus on the areas outlined in this analysis to reduce their churn rate. Additionally, there are further steps that can be taken to identify why some customers are unsatisfied with the service, such as:
- Competitive analysis: Conducting a thorough analysis of the company’s competitors can provide insights into where they may have an advantage over SyriaTel. This can help identify areas for improvement and enable the company to stay competitive in the market.
- Sentiment analysis: Performing sentiment analysis can help SyriaTel understand the different sentiments customers have towards their services. By analyzing customer feedback, reviews, and social media data, the company can gain valuable insights into customer satisfaction levels and areas that need attention.
By implementing these strategies and continually monitoring customer behavior and feedback, SyriaTel can enhance their retention efforts and improve customer satisfaction, ultimately reducing churn and increasing their overall success in the telecom industry.
