Helping a young start-up grow

Vive retention model : Insight Data Science Project

Transcript

1. VIVE and Nevena Francetić HELPING A YOUNG START-UP GROW
2. Introduction Problem: customer retention and churn prevention Deliverable: prediction model for churn rates
3. Pipeline Feature engineering 10-fold cross-validation, auc score Random forest classifier
4. Interesting story 90% confidence: auc score between 0.91 and 0.94
5. Interesting story 90% confidence: auc score between 0.91 and 0.94 Scheduling algorithms
6. Interesting story 90% confidence: auc score between 0.91 and 0.94 Scheduling algorithms Coincides with: • Change in the business model • Followed by introduction of a new algorithm
7. Feature importance 90% confidence: auc score between 0.85 and 0.90
8. Feature importance 90% confidence: auc score between 0.85 and 0.90 Target new users in a promotional campaign
9. Feature importance 90% confidence: auc score between 0.85 and 0.90 Investigate the common reasons for users to call the support center
10. Feature importance 90% confidence: auc score between 0.85 and 0.90 Encourage users to leave salon ratings
11. Deliverables • Modified schema to allow easier tracking of customers history • Predictive model with high accuracy Actionable insights • Promotional campaigns targeted to users in their 2nd or 3rd month of membership • Encourage users to leave salon ratings • Investigate the main reasons for users to call the support center
12. About me • PhD in Mathematics from University of Toronto (Canada) • Combinatorial design theory (networks, hypergraphs, test suits) nfrancetic@gmail.com • Hobbies: yoga, hiking, travelling, reading