Dynamic Customer Churn Prediction Engine

A predictive analytics engine that flags customers likely to churn before they leave. I built it in Python with scikit-learn for the modelling and pandas for the data pipeline, wrapped in an interactive Streamlit dashboard so the team can see who's at risk and act proactively.

Most businesses only notice churn after it happens. The client wanted to get ahead of it — to identify at-risk customers early using their behavioural data, and to do it in a way non-technical staff could actually explore and trust, not a black-box model buried in a notebook.

A churn-prediction model in Python using scikit-learn, trained on behavioural and account data

A pandas data pipeline for cleaning, feature engineering, and preparation

An interactive Streamlit dashboard surfacing at-risk customers and key drivers

Clear, explainable outputs the retention team can act on

Python, pandas, scikit-learn, and Streamlit.

The client shifted from reactive to proactive retention — spotting at-risk customers early through a model and dashboard that turn raw behavioural data into clear, actionable churn signals.