Topics: Coronavirus | AI | Data Science | Deep Learning | Machine Learning | Python | R | Statistics

KDnuggets Home » News » 2020 » May » Tutorials, Overviews » Customer Churn Prediction: A Global Performance Study ( 20:n20 )

Customer Churn Prediction: A Global Performance Study


This article details an automated machine-learned approach to predict customer churn and its results across selected communication service providers around the globe.



By Dr. Rahul Vyas, Head of Data Sciences Practices at Flytxt

Customer retention is one of the major challenges in any industry. Many firms realize that their existing customers are their most valuable asset, and it is more beneficial to keep and satisfy existing customers than to acquire new customers. Several studies such as the research done by Frederick Reichheld of Bain & Company (the inventor of the net promoter score) suggests that increasing customer retention rates by 5% increases profits by 25% to 95%.

Due to saturated markets and stiff competition, it is very essential for Communication Service Providers (CSPs) to identify customers who are prone to switch to other networks and to develop an effective and accurate customer churn model to efficiently manage the customer relationships.

 

Earlier efforts on analytics to identify potential churners

 
With the early advancement of analytical capabilities, CSPs tried to group their customers into different segments based on a handful of parameters. The identified ‘potential churners’ were often pampered with the lucrative package and discounted rate in an effort to retain them.

However, these initial approaches were not without their negatives;

  • Lack of capacity to process all possible data sources
  • Analytics at individual customer level was not available
  • Lack of flexibility; the analytics models were often not adaptable or easily changeable to behavioral, regional or organization level trends

 

An automated approach to predict customer churn

 
Flytxt’s Automated Machine Learning (auto-ML) is the process of automating the time consuming, iterative tasks of model development and deployment. It allows data scientists, data analysts, and business users to build inactive churn model with high scale, efficiency, and productivity all while sustaining model quality. Flytxt’s unique Feature Engineering and Feature Selection technique enables the extraction of key variables. It is now possible to identify the potential inactive customers that are likely to churn and take measurable steps to retain them quickly with the advancement of data-driven auto-ML framework. This auto-ML framework has been validated by major telecom operators across the globe.

In general, a supervised classification Machine Learning problem required labeled data that requires each instance to be labeled as churn or not churn. The data received from major telecom operators are generally not labeled. Hence Flytxt has developed a unique logical way to determine and label the consumers as churn or not churn depending on their inactivity over the previous two months’ of usage history.

A typical data set available with any CSP will have huge volume and veracity, and a typical dataset would be of the order of:

Figure

 

The data in the telecom industry is difficult to process using conventional computing systems as it requires a lot of actions and transformations to be done on a dataset in order to complete the task. As a result, this kind of Big Data is processed using memory-efficient distributed computing techniques. This enhances the robustness of the churn model deployed in multiple locations of African, Asian and American continents

 

Results of the auto-ML churn prediction models across selected CSPs around the globe

 
Flytxt has been enabling CSPs across the globe by addressing their customer’s concerns and creating value. Listing below some impressive numbers from Flytxt’s predictive churn models deployed across a handful of CSPs across the globe.

Figure

 

The model is widely accepted by various telecom operators across the globe because of its ability to deal with the variation in data and proving to be highly effective with accurate and precise churn prediction. However, identifying churn consumers alone is not sufficient. Flytxt also offers telecom operators to retain these likely to churn consumers by running the Winback campaign allowing telecom operators to retain the revenues. Evidently, the telecom operators have benefited by Flytxt’s hassle-free churn prediction solution.

 
Flytxt is a market leading supplier of analytics and AI-based customer value management solutions. Dr. Rahul Vyas heads the Data Sciences Practices at Flytxt.

Related:


Sign Up

By subscribing you accept KDnuggets Privacy Policy