A mobile company with over 1.5 million users needed to manage it's retention and activation marketing budgets and need a robust prediction of customer activity in the next 7 days and also churn in the next 30 days.

I built a system that considers users in-active for 60 days as churned and then user activity for every 60-day cycles was used to train ML models to predict activity in the next 7 days.

Using a model built using gradient boosting algorithms - LightGBM we provided and production model that provided excellent results.

Improved Retention: The significantly reduced customer churn because users received activity notifications and offers.

Reduced Marketing Budgets: Money spent on marketing reduced by over 40%.