Analyzing Customer Retention

Insight data science project

Transcript

1. Joy Wedding App: Analyzing Customer Retention Tyler Nelson
2. Customers sign up for Joy, but stop using it. *Actual percentages not represented to preserve company privacy
3. • Identify key components driving user retention from clickstream data • Predict & visualize customer retention based on 2 weeks usage The Deliverable *Actual counts not represented to preserve company privacy
4. Model Comparison Difference Precision 10% Recall 11% Accuracy 9%
5. Model Design Visualized data and Designed 20+ features Selected tables that contained relevant info Performed k-fold cross validation
6. Feature Engineering Users/NonUsers retention for App features Eight binary significant features used
7. An Unexpected Deliverable Previous Threshold Implemented Threshold Threshold for user/non-user distinction for completed weddings *Actual percentages not represented to preserve company privacy
8. Deliverable ➢ Accurate model to detect a user/non-user in the first two weeks of signing up ➢ iPython notebook joining all of the tables needed to assess user retention ➢ iPython notebook to automate and streamline user retention projections ◦ Plots of projections ◦ A list of events at risk ➢ A threshold for a successful user
9. Tyler Nelson PhD, Statistics