Dynamic, Multi-faceted Relationship Health Score

Insight Data Science consulting project using machine learning to assess business-to-business relationships.

Transcript

1. Measuring Business-to-Business Relationship Health
 Michael Levy
2. Managing SaaS Relationships Start-up
3. Platform Managing SaaS Relationships Start-up
4. Platform Manage Subscriptions Managing SaaS Relationships Start-up
5. Platform Manage Subscriptions Managing SaaS Relationships Start-up
6. Want data-driven score Old way
7. Want data-driven score Old way
8. Want data-driven score Old way New way
9. Want data-driven score Old way New way
10. Desired score features 1. Update in realtime
11. Desired score features 1. Update in realtime 2. Customize to companies and learn with time
12. Desired score features 1. Update in realtime 2. Customize to companies and learn with time 3. Implementable in Java/Clojure
13. Score responds to events
14. Score responds to events
15. Score responds to events
16. Score responds to events
17. Score responds to events
18. Score responds to events
19. Machine learning to assess relationship health Data • 6GB MySQL database dump • 100+ tables • Platform-use, CSM assessments, 
 communications, etc.
20. Machine learning to assess relationship health Relationship Health Score ?
21. Machine learning to assess relationship health Probability subscription renewal Relationship Health Score Transform ?
22. Machine learning to assess relationship health Random Forest Probability subscription renewal Predict Relationship Health Score Transform ?
23. Machine learning to assess relationship health Random Forest Probability subscription renewal Build Features Predict Relationship Health Score Transform ?
24. Feature Engineering Feature Categories
25. • Subscription attributes Feature Engineering Feature Categories
26. • Subscription attributes • Relationship details Feature Engineering Feature Categories
27. • Subscription attributes • Relationship details • CSM scores & notes Feature Engineering Feature Categories
28. • Subscription attributes • Relationship details • CSM scores & notes • Customer pipelines Feature Engineering Recruit On-board Adopt Feature Categories
29. • Subscription attributes • Relationship details • CSM scores & notes • Customer pipelines • Email, call & meeting records Feature Engineering Feature Categories
30. • Subscription attributes • Relationship details • CSM scores & notes • Customer pipelines • Email, call & meeting records • Subscription attributes • Relationship details • CSM scores & notes • Customer pipelines • Email, call & meeting records • Subscription attributes • Relationship details • CSM scores & notes • Customer pipelines • Email, call & meeting records Feature Engineering Feature Categories
31. Dramatic Increase in Accuracy
32. Michael Levy PhD in Computational Social Science, UC Davis
33. From Probability to Score Score = 100 x √Probability
34. Variable Importance
35. 82% Decrease in Log Loss