SQL project analyzing churn behavior in a subscription-based fintech app

Churn Prediction Audit (SQL) – Fintech App

This project analyzes churn behavior using SQL across a simulated fintech user lifecycle dataset. This project reflects the patterns and logic I’ve encountered in real-world churn and retention work.

---

Business Case

Your team has been asked to investigate potential user churn risks in a subscription-based finance platform. The product team suspects that support ticket trends, failed payments, and early inactivity may be predictive of churn.

---

Dataset Overview

Note: This dataset was synthetically generated to simulate churn patterns in a fintech app, modeled to mirror typical user lifecycles.

Tables used:

• users: user_id, signup_date, country, plan_type, age
• transactions: payment activity and success/failure flags
• logins: login frequency over time
• support_tickets: complaints, refund requests, or downgrade issues

---

Project Goals

1. Define churn behavior
   • Example: no login for 30+ days, refund + downgrade, or repeated payment failures
2. Identify churn risk signals
   • Early login drop-off
   • Support escalation
   • Failed payment patterns
3. Surface insights for the product team
   • What patterns precede churn?
   • Are there demographic or regional risk concentrations?

---

Churn Definition

For this analysis, a user is considered churned if they meet one or more of the following:

1. Have not logged in for 30+ days since their last recorded activity
2. AND have fewer than 3 successful transactions
3. OR have submitted a support ticket for a 'Refund Request' or 'Downgrade Dispute'

---

Tools Used

• MySQL + SQL Workbench
• GitHub for version control
• Manual data exploration via SQL queries
• Dataset generated and cleaned using Python (Faker, Pandas)

---

File Structure

/churn-prediction-sql ├── README.md ← this file ├── data/ │ ├── data_users.csv │ ├── data_transactions.csv │ ├── data_logins.csv │ └── data_support_tickets.csv ├── sql/ │ └── churn_queries.sql ← main analysis queries ├── insights.md ← summary of findings

---

Questions or Feedback?

I welcome feedback from data leaders and hiring managers — connect with me on via LinkedIn or reach out directly.