🚀 Event-Driven SaaS Analytics Platform

A production-style Data Engineering project that simulates a real subscription-based SaaS product and builds a complete modern analytics pipeline — from large-scale event generation to executive business dashboard.

This project models realistic SaaS behavior and implements a layered warehouse architecture (Bronze → Silver → Gold) using industry-standard tools.

---

📌 Project Overview

This system simulates a subscription-based SaaS platform (Spotify / Swiggy-style behavioral modeling) and builds an end-to-end analytics stack including:

• Large-scale synthetic event generation (~900K+ events)
• Subscription lifecycle simulation (6 months)
• Recurring billing system
• Churn probability modeling
• Retention cohort modeling
• KPI data marts
• Executive analytics dashboard

The goal is to demonstrate real-world Data Engineering practices such as:

• Event-driven architecture
• Star schema modeling
• Warehouse layering
• Analytical KPI marts
• Production-style project organization

---

🏗 System Architecture

Python Event Generator
        ↓
PostgreSQL (Bronze Layer - Raw Events)
        ↓
dbt Transformations (Silver Layer - Star Schema)
        ↓
Gold Layer KPI Data Marts
        ↓
Streamlit Business Dashboard

---

📊 Project Scale & Complexity

• ~900,000+ simulated user events
• 6-month subscription lifecycle modeling
• Multi-cohort retention tracking
• Recurring monthly & yearly billing simulation
• Payment failure probability
• Reactivation behavior modeling
• Fully normalized star schema warehouse

---

🛠 Tech Stack

Layer	Tool	Purpose
Event Generation	Python	Synthetic event simulation
Data Warehouse	PostgreSQL	Central analytics warehouse
Transformations	dbt	Staging, star schema, KPI marts
Visualization	Streamlit	Business dashboard

---

🧠 Data Warehouse Design

🥉 Bronze Layer (Raw Events)

Append-only immutable event store.

Table:

• raw_user_events
  • UUID primary key
  • Event timestamp
  • Event type
  • JSONB event_properties
  • User identifier

Design Principles:

• No transformations
• Fully replayable
• Audit-friendly

---

🥈 Silver Layer (Star Schema)

Staging

• stg_raw_user_events

Dimension Tables

• dim_users
• dim_plan
• dim_date
• dim_event_type

Fact Tables

• fact_user_activity
• fact_payments
• fact_subscriptions

Design Principles:

• Surrogate keys
• Clear fact-dimension separation
• Optimized for analytical joins

---

🥇 Gold Layer (Business KPI Marts)

Pre-aggregated business-ready models:

• kpi_dau
• kpi_mau
• kpi_mrr
• kpi_churn
• kpi_ltv
• kpi_retention_cohort

Optimized for dashboard consumption.

---

📊 KPIs Implemented

👥 User Metrics

• Daily Active Users (DAU)
• Monthly Active Users (MAU)
• Stickiness Ratio (DAU / MAU)

💰 Revenue Metrics

• Monthly Recurring Revenue (MRR)
• Customer Lifetime Value (LTV)
• Subscription distribution

🔁 Retention & Churn

• Monthly churn rate
• Cohort retention matrix
• Reactivation analysis

---

📊 Dashboard Capabilities

The Streamlit dashboard includes:

• Executive KPI summary cards
• DAU & MAU growth trends
• Revenue trend (MRR over time)
• Churn breakdown (absolute + rate)
• Retention cohort heatmap
• LTV distribution histogram
• Architecture overview section

---

🚀 How To Run Locally

1️⃣ Generate Events

python generator/event_generator.py

2️⃣ Run dbt Transformations

cd dbt
dbt run

3️⃣ Launch Dashboard

streamlit run dashboard/app.py

---

📁 Project Structure

event-driven-saas-analytics/

├── generator/
│   |── behaviour_engine.py
│   ├── config.py
│   ├── enums.py
│   ├── main.py
│   ├── postgres_writer.py
│   ├── revenue_engine.py
│   └── models.py
|
├── dbt/
│   ├── models/
│   │   ├── staging/
│   │   ├── marts/
│   │   ├── dimensions/
│   │   ├── facts/
│   │   └── intermediate/
│
├── db/
│   ├── indexes.sql/
│   └── raw_schema.sql/
│
├── dashboard/
│   └── app.py
│   └── database.py
│   └── queries.py
│
└── README.md

---

🧪 Business Logic Simulation

The synthetic generator models:

• User acquisition growth curve
• Signup → feature usage funnel
• Upgrade probability modeling
• Monthly vs yearly subscription logic
• Recurring billing cycles
• Churn after minimum subscription tenure
• Reactivation probability
• Payment failure scenarios

Designed to mirror realistic SaaS growth dynamics.

---

🎯 Engineering Concepts Demonstrated

• Event-driven system modeling
• Star schema warehouse design
• Layered architecture (Bronze / Silver / Gold)
• Analytical SQL modeling
• KPI mart construction
• Large-scale synthetic data handling
• Production-style project structure

---

📈 Business Impact Simulation

This architecture enables:

• Real-time KPI visibility
• Subscription revenue tracking
• Cohort-based retention analysis
• Data-driven churn reduction strategy
• Executive-level reporting readiness

---

🔮 Future Enhancements

Planned upgrades:

• Apache Airflow orchestration
• Docker containerization
• Kafka streaming ingestion
• Data quality tests in dbt
• Incremental model optimization

---

📊 Dataset

• ~900K+ events
• 6-month lifecycle coverage
• Subscription events
• Payment transactions
• Feature usage activity
• Upgrade & churn transitions

Fully synthetic data generated using Python.

---

⭐ Project Goal

To design and implement a production-style, end-to-end analytics system that simulates real SaaS business operations and demonstrates modern Data Engineering practices at scale .

This project showcases:

• Event-driven system design
• Layered warehouse architecture (Bronze → Silver → Gold)
• Star schema data modeling
• KPI data mart construction
• Large-scale synthetic data handling (~900K+ events)
• Business-ready analytical reporting

---

👨‍💻 Author

Kavin Kishore
Delhi Technological University (DTU)

Built as a production-style Data Engineering project.