Prediction of customer
probability to churn

INITIAL SITUATION

Our client faced with the scenario that there was a high turnover and many of their customers wanted to terminate their contract after the first year.

The challenge was to predict what reasons a customer might have to terminate and when.​

SOLUTION

A classification model has been tuned based on specific client data.

After several iterations and through continuous business and technical validation the model was able to report the most important drivers of churn as well as providing daily predictions to the business that can now take actions on specific customers.

BUSINESS VALUE

The increased business understanding of churn drivers led to improved and targeted marketing campaign design.

Given a specific target, they are now 20% more likely to identify a true churner.

​

FRAMEWORK & TOOLS

Open source and Big Data technologies (Azure, Python) have enabled the development of the pipeline, which stores the predicted results in an additional database.

Companies can query the results there. The pipeline was developed using Azure Data Factory and the project was based on the Scrum methodology.