The objective of this project was to analyze the results of an A/B test for an e-commerce website to determine if a new checkout page UI (Treatment) drove a higher conversion rate than the existing UI (Control).
This project encompasses end-to-end data pipeline execution: extracting and cleaning raw web traffic logs in PostgreSQL, conducting statistical hypothesis testing in Python, and delivering a final business recommendation via a Power BI executive dashboard.
- Database & Data Cleaning: PostgreSQL (CTEs, Window Functions)
- Statistical Analysis: Python (Pandas, Statsmodels, SciPy)
- Data Visualization: Power BI (DAX, Data Modeling)
-
Data Cleaning (SQL): Handled a raw dataset of ~300,000 web sessions. Engineered a SQL script utilizing
ROW_NUMBER()window functions to remove duplicate refresh events and isolate 290,584 clean, unique user sessions. -
A/B Testing (Python): Formulated the null hypothesis that the new page would not increase conversions. Executed a Two-Proportion Z-Test to calculate statistical significance (
$\alpha = 0.05$ ). - Business Intelligence (Power BI): Developed an executive-facing dashboard to visualize the conversion funnel, daily performance trends, and final KPIs.
- Control Group (Old Page): 12.04% Conversion Rate
- Treatment Group (New Page): 11.88% Conversion Rate
- P-Value: 0.90
Conclusion: The test yielded a p-value of 0.90, failing to reject the null hypothesis. The Treatment group underperformed the Control group consistently over the testing period.
Recommendation: Do not deploy the new checkout page. Retaining the current UI prevents a projected drop in overall purchase conversions, protecting company revenue.