Key results
improvement in churn prediction accuracy using ML models trained on historical engagement behavior
reduction in patient churn through targeted outreach based on churn scoring
drop in no-show rates with predictive scheduling optimization
faster campaign performance analysis using centralized data in Redshift and Tableau
About the Client
A leading U.S.-based digital health platform enabling smarter patient engagement across 10M+ users through AI-driven communication, referral automation, and intake workflows. The client needed a predictive analytics system to reduce churn, improve no-show rates, and unify engagement insights across disparate systems.
Key Challenges
Unidentified patient churn patterns - leading to revenue leakage and disengagement
Manual, delayed campaign reporting – slowing down optimization decisions
High no-show rates – hurting provider utilization and operational planning
Fragmented data across platforms – limiting a 360° view of patient journeys
Our Approach
Assessment & Strategy
- Audited existing EHR, chatbot, and communication data flows
- Identified key churn and performance KPIs
- Defined analytics roadmap aligned with retention goals
Data Architecture Design
- Designed a HIPAA-compliant data model on AWS Redshift
- Defined ETL pipelines for engagement, referral, and scheduling data
- Planned snapshot tables for churn, NPS, and campaign analysis
ML Model Development
- Built churn prediction models using XGBoost and scikit-learn
- Engineered features like time since last visit, NPS score deltas
- Validated model with historical cohort data
BI & Reporting Framework
- Designed Tableau dashboards with Redshift extracts
- Enabled user segmentation, campaign performance, churn risk tracking
- Established role-based access across product, operations, and leadership
Implemented Solution
Central Data Warehouse
- AWS Redshift used as the source of truth
- Data staged in S3, transformed via Python scripts
- Integrated with all core engagement and scheduling systems
Churn & Risk Scoring
- Trained ML models deployed on scheduled batches
- Risk scores used to drive retention workflows
- Triggered alerts for high-risk patient cohorts
Performance Dashboards
- Campaign, referral, and patient engagement dashboards in Tableau
- Automated daily refreshes via Redshift extracts
- Enabled real-time drilldowns and export-ready views
Governance & Enablement
- Created centralized KPI dictionary and definitions
- Implemented row-level security and user filters in Tableau
- Enabled ad hoc query access via Redash
Technologies Used
