Optimizing E-com Conversions: A/B Testing Analysis 📈

📌 Project Overview

This project evaluates the impact of a new e-commerce feature on user conversion rates. The primary focus was on ensuring a high-integrity experiment by identifying and removing data contamination (where users accidentally saw both versions), ensuring the final results were statistically sound.

🎯 Business Challenge

The client needed to know if a UI change actually increased sales. My role was to clean the raw experiment data, handle overlapping user segments, and perform a formal hypothesis test to determine if the "Lift" was statistically significant.

🛠️ Tech Stack

• Database Querying: SQL (Data auditing & cleaning contamination)
• Statistical Analysis: Hypothesis Testing (Z-test / P-value calculation)
• Visuals: Tableau / Python (Select your tool) for conversion funnels

📊 Methodology & Logic

1. Data Auditing (SQL): Identified "contaminated" users who appeared in both Control and Variant groups due to tracking glitches.
2. Data Cleaning: Excluded contaminated records to ensure a "Clean Test" environment.
3. Hypothesis Testing:
   • Null Hypothesis ($H_0$): The new feature has no effect on conversion.
   • Alternative Hypothesis ($H_a$): The new feature significantly increases conversion.
4. Result: Calculated a P-value of [Insert P-Value, e.g., 0.03], rejecting the null hypothesis with 95% confidence.

📁 Project Contents

• data: Raw experiment logs and cleaned data samples.
• analysis: SQL scripts used to isolate contamination and calculate metrics.
• visuals: Conversion funnel charts and statistical distribution plots.
• report: Final summary report with actionable insights for content creators.
• report: Final report on experiment validity and business recommendations.