The Problem: Advertisers were spending less and leaving the platform (churning) at an increasing rate, but the company didn't know the specific reasons why.

1. Phase 1 (The "Who"): A quantitative study used Log Analysis and a Logistic Regression model to analyze user behavior. This successfully identified who was at-risk of churning with ~80% accuracy, finding that declining conversion rates and unresolved support tickets were the biggest predictors.

2. Phase 2 (The "Why"): The researcher then interviewed "high-risk" users (identified by the model) to find the root cause. The "why" was clear: users were frustrated with a confusing conversion-tracking UI, poor customer support, and unclear ad policies.

3. Action & Impact: The findings were used to create real-time risk scores and run A/B tests on solutions (like a new dashboard and a priority support queue). This new proactive system improved advertiser retention by 17% and increased CSAT by 22 points.

Link to presentation: (Contact me at dtbyrd@gmail.com for access )