An end-to-end analytics project focused on understanding e-commerce conversion behavior, funnel drop-offs, revenue efficiency, and purchase likelihood.
E-commerce businesses lose revenue at multiple points of the customer journey — from initial visit to final purchase. Without clear funnel visibility, it's hard to identify where drop-offs happen, which channels and campaigns underperform, and how to prioritize revenue optimization efforts.
The dataset contains 120,000 sessions with 17 columns, covering July 2025 to December 2025 at session-level granularity.
Models tested during development:
The final Streamlit application uses a Random Forest classifier with a 92% decision threshold.
Note: The live Streamlit application URL has not yet been deployed publicly. The GitHub repository and reports are linked below.
Combines SQL-based funnel analysis, business intelligence dashboards, and a deployed prediction model to help stakeholders understand where customers drop off, why revenue leaks, and how to prioritize optimization. The Streamlit app supports what-if purchase likelihood prediction for new sessions.
Note: Stage-wise conversion percentages and revenue leakage amounts were not part of the verified brief and are not included here.