Printable Handouts

PDF

Navigable Slide Index

1. Introduction
2. What to expect from this talk
3. Speaker's background
4. What does data mining really do?
5. Data mining is about customers
6. Traditional approach to data mining
7. Survival data mining
8. Example results
9. Survival analysis has a long history
10. Original statistics
11. Survival for marketing has some differences
12. Focus on the end of the relationship
13. How long will a customer survive? (1)
14. Calculation of hazard probabilities
15. Bathtub hazard (risk of dying by age)
16. Hazards are like an X-ray into customers
17. Hazards show features of the customer lifecycle
18. How long will a customer survive? (2)
19. Survival is similar to retention
20. Why survival is useful
21. Median lifetime: the tenure where survival = 50%
22. Average tenure is the area under the curves
23. Survival to quantify marketing efforts
24. We can use area to quantify results
25. Results: loyalty-responsive customers vs. others
26. Customers may leave for many reasons
27. Why do customers stop?
28. Overall survival for a group of customers
29. Competing risks: voluntary and involuntary stops
30. Involuntary churn and non-credit card payers
31. A better way to look at this
32. Customer-centric forecasting with survival analysis
33. Using survival for customer-centric forecasting
34. The forecasting solution is a bit complicated
35. Survival data mining - summary

Topics Covered

• Background on survival analysis from a data miner's perspective
• Introduction to key ideas in survival analysis: hazards, survival, competing risks
• Understanding survival and hazard charts
• Quantifying loyalty marketing program
• Voluntary and involuntary churn
• Forecasting

Links

Series:

• Data Mining

Categories:

• Technology & Operations

Talk Citation

Publication History

• Published on June 17, 2008
• Archived on March 17, 2022