GA4 E-commerce Analytics

Turn the GA4 public sample dataset into business insights, Power BI dashboards, a trained ML model, and a FastAPI prediction service running in Docker.

---

Preview

Architecture

Entity Relationship Diagram (ERD)

Power BI Dashboards

1. Channel Performance & Conversion

2. E-Commerce Performance Trends

3. Funnel Performance

---

1) Business Case & Questions

Use GA4 e-commerce events to improve funnel conversion and channel ROI.

Key questions

• Which channels convert best from sessions → purchases?
• How do sessions, revenue, and conversion change by day/week/month?
• Where are the biggest drop-offs in the funnel?
• (ML) Can we predict purchase propensity from simple session features?

---

2) Data & ERD

Source (BigQuery public dataset)
bigquery-public-data.ga4_obfuscated_sample_ecommerce.events_*

Star schema

• dim_date (PK = date_pk)
• dim_channel (PK = channel_key)
• fact_funnel_by_date_channel
  • FK: date_pk → dim_date.date_pk
  • FK: channel_key → dim_channel.channel_key
  • Metrics: sessions, add_to_cart, purchases, revenue, conversion_rate, aov

SQL models (/sql)

• 10_dim_date.sql
• 20_dim_channel.sql
• 30_fact_funnel_by_date_channel.sql

Visual ERD
See images/erd.png (rendered above).

---

3) Project Architecture

Flow: GA4 → BigQuery → SQL models → Python notebook (exports + model) → Power BI → FastAPI (Docker)

• BigQuery builds the star schema.
• Notebook creates exports/*.csv for BI and trains a purchase propensity model.
• Power BI reads exports for dashboards.
• FastAPI serves a /predict endpoint using the trained model.
• Docker packages the API for consistent local runs.

Diagram: images/architecture.png (rendered above).

---

4) Analysis (Python)

Notebook: notebooks/01_eda_kpis.ipynb

Covers

• Channel performance (sessions, purchases, revenue, conversion rate)
• Time trends (weekly/monthly)
• Funnel analysis (Sessions → Add to Cart → Purchases)
• Trains a RandomForestClassifier for purchase propensity

Exports produced (used by BI)

• exports/channel_summary.csv
• exports/time_summary.csv
• exports/funnel_summary.csv

---

• PBIX (download): dashboard/powerbi/GA4 Dashboard.pbix
• Slides (overview): Google Slides

---

6) Machine Learning

Model: RandomForestClassifier for purchase propensity.

Example features

• Numeric: sessions, add_to_cart
• Categorical one-hots: channel_group, day_name, month

Artifacts (/models)

• purchase_rf.joblib
• expected_cols.json ← columns used at inference

---

7) Prediction API (FastAPI) + Docker

Code: src/api.py

Run locally (virtual env)

python -m venv .venv

# Windows (PowerShell)
.\.venv\Scripts\Activate.ps1

# macOS/Linux
source .venv/bin/activate

pip install -r requirements.txt
uvicorn src.api:app --host 127.0.0.1 --port 8000 --reload

Health check

curl http://127.0.0.1:8000/health

curl -X POST http://127.0.0.1:8000/predict \
  -H "Content-Type: application/json" \
  -d '{"sessions":3,"add_to_cart":1,"channel_group":"Referral","day_name":"Saturday","month":12}'

Docker (local)

docker build -t ga4-api .
docker run -p 8000:8000 ga4-api
curl http://127.0.0.1:8000/health

Endpoints

• GET /health → service status
• POST /predict → { "prediction": 0|1, "probability_purchase": float }

8) Results & Business Impact

• Identified top-performing channels by conversion rate and revenue.
• Highlighted the largest funnel drop-offs (Sessions → Add to Cart, Cart → Purchase).
• Machine Learning flagged high-propensity sessions for remarketing and CRO (conversion rate optimization).
• Delivered a full end-to-end pipeline (SQL → Python → BI → ML → API → Docker) showing both technical depth and business value.

9) Reproduce

Clone & setup

git clone https://github.com/Egbe34/ga4-ecommerce-analytics.git
cd ga4-ecommerce-analytics
python -m venv .venv

# Windows (PowerShell)
.\.venv\Scripts\Activate.ps1

# macOS/Linux
source .venv/bin/activate

pip install -r requirements.txt

Run notebook → generate exports & models

Open notebooks/01_eda_kpis.ipynb

Run all cells to create exports/* and models/*

Start API

10) Folder Structure

ga4-ecommerce-analytics/ ├─ sql/

├─ notebooks/

├─ exports/

├─ models/

├─ src/

├─ dashboard/

│ └─ powerbi/

├─ docs/

├─ images/

├─ Dockerfile

└─ README.md

**License:** MIT