This project simulates product metric analysis and causal modeling for Streamly (a fictional subscription-based streaming platform). It demonstrates how product data scientists quantify user retention and estimate causal uplift using Bayesian methods.

• Simulated transactional and usage data including churn events and plan types
• Computed ARPU and retention by cohort, with behavioral segmentation
• Applied Bayesian logistic regression to estimate churn reduction from annual plans
• Visualized posterior distribution and interpreted business impact
• Posterior results indicate a ~30 percentage point reduction in churn with high certainty

For full code and methodology, see the GitHub repository.