📈 Cohort Retention Analysis

An end-to-end cohort retention analysis pipeline for a B2B SaaS subscription product. Analyzes 10,000 users over 2 years using both weekly and monthly cohorts, measuring user retention and Net Revenue Retention (NRR) — the two most critical metrics for SaaS product health and investor reporting.

---

🎯 Business Questions

• What percentage of users are still active after 1, 3, and 6 months?
• Is cohort quality improving over time (is the product getting better)?
• What is our Net Revenue Retention — are existing customers growing or shrinking in value?
• When is monthly vs weekly cohort analysis the right tool?

---

📁 Project Structure

cohort_retention_analysis/
├── outputs/                     # Charts, cohort tables (CSV), executive summary
├── src/
│   ├── data_prep.py             # 2 years of synthetic SaaS subscription events
│   ├── cohorts.py               # Monthly + weekly cohort table construction
│   ├── retention.py             # Retention curves, churn rates, NRR
│   └── visualizations.py       # Heatmaps, curves, dashboard
├── main.py                      # Run the full pipeline
├── requirements.txt
└── README.md

---

📊 Data

Synthetic SaaS subscription data simulating realistic behavior:

Parameter	Value
Users	10,000
Period	Jan 2023 – Dec 2024
Plans	Starter ($29), Professional ($49), Enterprise ($149)
Month 1 churn	~25% (industry benchmark)
Steady-state churn	~3%/month
Expansion probability	15% of retained users upgrade
Cohort quality	Improves 0.5%/month (product improving)

---

📐 Methodology

Cohort Construction

Frequency	Best For
Weekly	Evaluating product releases, short-term A/B tests, operational decisions
Monthly	Strategic planning, investor reporting, long-term trend analysis

Retention Metrics

User Retention Rate

Retention(cohort, period) = Active users in period N / Users at signup

Net Revenue Retention (NRR)

NRR = (Starting MRR - Churned MRR + Expansion MRR) / Starting MRR

• NRR > 110%: World-class (Snowflake, Twilio tier)
• NRR > 100%: Healthy — revenue grows even without new customers
• NRR < 100%: Revenue leaking — must be addressed

Churn Analysis

• Monthly churn rate with 3-month rolling average
• Annualized churn = 1 - (1 - monthly_churn)^12
• Cohort quality comparison at Month 1, 3, 6, and 12

---

📈 Output Files

File	Description
01_monthly_retention_heatmap.png	Classic cohort heatmap (monthly)
02_weekly_retention_heatmap.png	Weekly granularity heatmap
03_retention_curves.png	Avg retention curve + individual cohorts
04_nrr_over_time.png	Net Revenue Retention trend
05_churn_rate.png	Monthly and annualized churn over time
06_cohort_quality.png	Retention improvement across cohorts
07_summary_dashboard.png	Full summary dashboard
monthly_user_retention.csv	Monthly cohort table (exportable)
monthly_mrr_retention.csv	Monthly MRR retention table
weekly_user_retention.csv	Weekly cohort table
retention_executive_summary.txt	Business-ready summary

---

🚀 Getting Started

git clone https://github.com/yourusername/cohort_retention_analysis.git
cd cohort_retention_analysis
pip install -r requirements.txt
python main.py

No data download required — data is generated synthetically with realistic SaaS benchmarks.

---

🛠 Tech Stack

• Python 3.14+
• pandas — cohort table construction and time-series manipulation
• numpy — simulation and statistical computation
• matplotlib — heatmaps, curves, and dashboards
• scipy — statistical analysis

---

💡 Key Concepts Demonstrated

• Weekly vs monthly cohort construction from event data
• User retention and revenue retention (NRR) analysis
• Cohort heatmap visualization (most-recognized product analytics chart)
• Churn rate calculation (monthly and annualized)
• Cohort quality trend analysis
• Net Revenue Retention with expansion modeling
• SaaS subscription event simulation

---

👤 Author

Built as part of a data science / product analytics portfolio project.
Background: 15+ years in Marketing Analytics | SQL | Python | Statistical Modeling