- Context: A SaaS platform with 1.2 million login events showed steady sign-ups, but the "Daily Habit" (retention) of users was fluctuating.
- The Task: Build a diagnostic tool to determine if users were leaving due to behavioral shifts or technical bottlenecks.
- The Goal: Connect technical performance (Error Rates) to user engagement (Stickiness) and identify the specific segments facing friction.
To handle the high volume of logs without lag, I focused on a clean data model.
- Action: Generated a 1.2M row synthetic dataset using Python to simulate real-world system logs.
- The Model: Implemented a Star Schema in Power BI to allow fast filtering across Regions, Devices, and Subscription types.
- Metric Logic: Built a "Golden Triangle" of KPIs: Reach (MAU), Habit (DAU), and Quality (Error Rate).
I developed custom measures to calculate "User Health" and isolate errors across 1.2 million rows.
- Stickiness Check: Used the ratio of Daily to Monthly users to measure habit formation, resulting in a 19.69% score.
- The Heatmap: Built an Error Troubleshooting Matrix to cross-reference every device and subscription tier.
- Impact Tracking: Created a measure to count Impacted Users to quantify the human cost of technical errors.
- Root Cause Analysis: Discovered a global 12.06% Error Rate, which acts as a ceiling preventing higher user stickiness.
- Finding: Pinpointed North America as the primary hotspot, representing 39.89% (6K) of all impacted users.
- Segment Isolation: Identified that Enterprise users on Tablets were facing the highest error share at 11.57%.
- The Result: Delivered a real-time troubleshooting template that reduced "Time-to-Discovery" from manual log review to interactive filtering dashboard.