Predicting Customer Churn for a Leading European Elevator Company

Client

The client is a Swiss multinational company that manufactures escalators, moving walkways, and elevators worldwide. The company produces, installs, maintains, and modernizes lifts and escalators in residential, commercial, and high-rise buildings.

Challenges

• Understanding customer behavior towards elevator maintenance services is a time-consuming and cost-efficient process.
• Develop a solution to help overcome this challenge by providing a list of customers more likely to end the relationship (churn).

Solution

• Developed a Proof of Concept (PoC) solution that analyzes historical data and uses machine learning/deep learning techniques to identify customers likely to churn due to elevator maintenance services.
• Analyzed historical data of maintenance service contracts related to indicators that might effectively identify churn.
• Derived specific features from the indicator data and used this in the machine learning/deep learning model.
• Finalized the best suitable algorithm based on business-specified metrics.
• Classified the customers as churn or non-churn using the best suitable algorithm.
• Implemented the solution in Spain and it's being expanded to other regions. 

Business Benefits

• Helped the sales and marketing leaders to focus on the subsets of customers likely to end the relationship, improving cost efficiency.
• Ensured 2% reduction in customer churn rate.

Tools, Technology, ML Techniques

• Azure Machine learning | Python | Sci-kit learn | Keras 
• Classification | Tree-Based Algorithms | Neural Networks