E-commerce Funnel & Conversion Analysis

I used R to analyze more than 500,000 e-commerce interactions from an online cosmetics dataset. I wanted to see how user activity moves from views to cart actions and purchases, and whether brands with more traffic also had better conversion rates.

Questions I Looked At

• Which user actions happen most often?
• Which high-traffic brands have the highest conversion rates?
• Does getting more views actually lead to better conversion?

Main Results

Customer Activity

Views were the most common event in the data. Purchases happened much less often than views and cart-related activity.

Highest-Converting Brands

I compared brands with at least 10,000 views so very small brands would not look unusually strong from only a few purchases. Among the brands that passed that cutoff, Milv had the highest conversion rate.

Views vs. Conversion

More traffic did not reliably mean a higher conversion rate.

I ran a linear regression using log-transformed views:

• R² = 0.0845
• p = 0.335

Views explained about 8.5% of the variation in conversion rates, and the relationship was not statistically significant in this analysis.

What I Did

I used tidyverse to clean and summarize the event data. The analysis includes:

• filtering missing brand values
• counting views, cart actions, removals, and purchases
• calculating brand-level conversion rates
• filtering out low-traffic brands for more stable comparisons
• creating funnel and conversion charts with ggplot2
• using linear regression to test the relationship between traffic and conversion

Files

Ecommerce-Funnel-Analysis/
├── ecommerce_funnel_analysis.R   # main analysis script
├── Cosmetic.Rmd                  # report source
├── Cosmetics.pdf                 # rendered report
├── ecommerce_funnel.png
├── top_brand_conversion.png
├── views_vs_conversion.png
└── README.md

The main analysis is in ecommerce_funnel_analysis.R. It was run in R using tidyverse, ggplot2, ggrepel, and scales.

Key Takeaways

• Higher traffic did not automatically lead to higher conversion.
• Conversion rates varied a lot even among high-traffic brands.
• View volume alone explained only a small part of the difference in conversion rates.
• The regression result was not statistically significant.

Tools Used

• R
• tidyverse
• ggplot2
• ggrepel
• scales
• linear regression

Notes

This is an exploratory analysis of one e-commerce dataset. The conversion rate is calculated at the brand level as purchases divided by views, so it should not be treated as a full user-level or session-level funnel.

A next step would be tracking individual users or sessions from view to cart to purchase and adding variables such as price and product category.