Growth Funnel & Retention Analysis

E-commerce Product Analytics Case Study

Project Overview

This project analyzes user behavior in a large-scale e-commerce environment to understand how product engagement translates into revenue and long-term value.

The main goal is to identify where users drop off in the conversion funnel, how purchasing behavior evolves over time, and which segments truly drive business impact.

Business Context

The dataset contains event-level user interactions from a multi-category e-commerce platform, including product views, cart additions, and purchases across two months (October and November 2019).

From a business perspective, this analysis answers questions like:

• Is user engagement improving over time?
• Are increases in activity translating into real revenue?
• Which product segments drive growth versus just traffic?
• Are new customers returning and generating long-term value?

This mirrors real scenarios where growth teams need to balance acquisition scale with sustainable customer value.

Key Questions This Analysis Answers

• How efficient is the view -> cart -> purchase funnel?
• Did conversion performance improve month over month?
• Which price segments generate revenue versus just engagement?
• Which product categories are driving growth?
• How does customer retention behave over time?
• What is the observed lifetime value (LTV) of different cohorts?

Dataset and Tooling

Dataset

• Public Kaggle dataset: E-commerce Behavior Data from Multi-Category Store
• Around 110 million raw event records

Technology stack

• Python
• DuckDB (large-scale analytical processing)
• Pandas
• Matplotlib and Seaborn (visualization)
• GitHub Pages (docs/index.html) for portfolio presentation

To keep processing fast and scalable, raw CSV files were converted into Parquet and queried with DuckDB.

Data Modeling and Performance Decisions

• Conversion of raw CSV files to Parquet for efficient storage and querying
• DuckDB queries directly over Parquet files
• Enriched event views with:
  • event_date
  • event_month
  • event_week
• Separation of raw and processed data to keep the repository organized

These are practical data engineering choices aligned with day-to-day analytics work.

Funnel Analysis

Core journey analyzed: View -> Cart -> Purchase

Key Findings

• View-to-cart conversion doubled from October to November, indicating stronger purchase intent.
• Cart-to-purchase conversion dropped, suggesting checkout friction became more relevant as top-of-funnel activity expanded.
• Overall view-to-purchase improved only slightly, so funnel gains did not convert proportionally into final purchases.

Revenue and Price Segment Insights

• Revenue increased month over month
• High-priced products represent most of total revenue
• Low-priced products add engagement but contribute little to total revenue

This indicates November growth was mainly premium-led rather than broad-based across all price tiers.

Retention and Cohort Analysis

Retention was analyzed through weekly purchase cohorts.

Key Findings

• Strong early repeat behavior in the first week
• Sharp drop after week 2
• More recent cohorts start weaker, which may indicate:
  • lower acquisition quality
  • lower short-term repeat intent
  • growth driven by broader traffic mix

These patterns point to opportunities in lifecycle marketing and post-purchase engagement.

Observed LTV (Lifetime Value)

An observed LTV proxy was calculated from total revenue per user in each cohort.

Main pattern:

• Older cohorts show higher observed LTV (longer monetization window)
• Newer cohorts show lower early observed value

This reinforces the need to balance scale with customer quality.

Portfolio Visuals (Current State)

• Visual assets are generated by generate_portfolio_assets.py
• Output files:
  • assets/funnel.png
  • assets/retention.png
  • assets/revenue.png
• The portfolio page is rendered through docs/index.html

Project Structure

growth-funnel-analytics/
├── assets/
│   ├── funnel.png
│   ├── retention.png
│   └── revenue.png
├── data/
│   ├── raw/                # raw dataset (ignored)
│   └── processed/          # parquet and duckdb files
├── docs/
│   └── index.html          # GitHub Pages portfolio
├── notebooks/
│   ├── 01_exploration_and_funnel.ipynb
│   ├── 02_growth_funnel_analysis.ipynb
│   ├── 03_retention_and_cohorts.ipynb
│   └── 04_revenue_ltv_insights.ipynb
├── generate_portfolio_assets.py
└── README.md

Next Steps

Possible extensions:

• Attribution modeling by traffic source
• Predictive LTV modeling
• Experimentation design (A/B test simulation)
• Integration with richer product usage signals

---

This project shows end-to-end product analytics thinking: from user behavior and conversion to retention and monetization, with a focus on clear business decisions.