Churn Analysis for Healthcare IT Service. A Predictive Analytic Solution and Customer Retention Strategy.
Service
Digital Healthcare Resource Application for Providers
Objective
Develop effective process to predict churn before customers terminate service
Approach
1. Create Project Plan and Pilot for proof of concept
2. Present to executive leadership team for buy-in
3. Establish Analytic Architecture
4. Build Refined Machine Learning Model
o Try multiple Algorithms
•Random Forest (RF)
•Support Vector Machine (SVM)
•Generalized Linear Model (GLM)
o Train on historic data
o Test on unseen data
5. Implement production and maintenance plan
Results
• Random Forest (RF) Model consistently performed best on unseen data
• 91% accurate on predicting customers who churn
• 85.1% overall accuracy
• Proposed retention plan estimates $1.4M in savings resulting from Predictive Analytic Solution