During the Boston Consulting Group (BCG) job simulation on The Forage platform, I worked on a project for a fictional energy services provider seeking recommendations to improve customer retention. To support their efforts, I analyzed customer data and built predictive models using the provided information. The dataset included historical energy pricing and customer subscription details.
The steps I followed were:
- Analyzing the data to identify patterns.
- Creating additional features that could enhance the model’s predictive power.
- Building a predictive model using a random forest algorithm.
- Determining which customer characteristics were most influential in predicting retention challenges.
The resulting model had a low F1 score (0.09) and requires further refinement. However, it provided valuable insights into the key attributes associated with retention risks. Based on the analysis, we recommended offering discounts to customers with high energy consumption to increase retention within this segment.
The tools and libraries used in this project include Python, Pandas, Seaborn, Matplotlib, and scikit-learn. Visualizations from the data analysis and feature importance are shown below. Links to the code are: part 1, part 2, part 3. The certificate of completion can be found here.
Cover photo by Brett Sayles