Client Experience Analytics: PAF & UEM

Product Operations Anomaly Detection Healthcare SaaS

A statistical signaling framework (PAF) and User Experience Model (UEM) designed to detect platform friction and "bad experience minutes" in real-time.

The Challenge

Platform reliability issues were being handled reactively. Alerts were primarily triggered by client-reported defects rather than automated internal signals.

Site Reliability Engineers (SREs) had to manually dig into logs and metrics, making Root Cause Analysis (RCA) slow and inconsistent. Consequently, issues often escalated before detection, increasing defect counts and client dissatisfaction.

The Solution

I designed the Performance Abnormality Flag (PAF) and User Experience Model (UEM) frameworks to detect friction and "bad-experience minutes" proactively.

Statistical Modeling: Engineered SQL-based detection models using stepped data structures and automated baselines to flag anomalies with high signal accuracy.
Real-Time Signaling: Built a near-real-time system to identify application stability issues (e.g., AppHang, Freeze) before clients reported them.
Dashboarding: Created the "UEM SLI PAF" dashboard to aggregate signals into a single operational view for the Service Resiliency Program.
Collaboration: Worked with SRE and Product Ops to tune model thresholds using real-world scenarios, ensuring reliability across live client environments.

The Impact

RCA Velocity: Reduced Root Cause Analysis (RCA) turnaround time by 63%.
Detection Accuracy: Improved incident detection accuracy by 30%.
Client Satisfaction: Reduced client-reported defects and improved internal "Go-Green" Jira performance.