Customer Retention Use Case Updated
I’m following up from an old document Burn that Churn? Checking Retention of Customers in Predictive Analysis and doing the same analysis but this time using SAP Predictive Analytics 2.2 Expert Analytics – the same data, the same model. We’re looking at the attributes of customers retained and comparing them to those who have left (banking sample customers)
Since we are trying to predict churn, that is the target variable, using the amount of deposits, investment amount, income and number of checking transactions.
After running the model, you can see the execution status above.
Above is the model summary
What is different from when I originally posted? The above shows a confusion matrix, with a false positive 6% of the time, true positive 94% and a false negative 1% of the time after outputting the results to Excel:
| Predicted | Predicted | ||||||
| Actual | No | Yes | Total | Actual | No | Yes | |
| No | 1458 | 85 | 1543 | No | 94% | 6% | |
| Yes | 14 | 986 | 1000 | Yes | 1% | 99% | |
| Total | 1472 | 1071 | 2543 | ||||
I think you also need a big screen to better see/view the decision tree:
The predicted values are the same as before, with younger people likely to “churn” and not be retained by the bank:
Reference
Learn more at Join ASUG at #SABOUC in Austin – Featuring SAP Predictive Analytics
|
Can’t make this conference? Consider joining us at SAP TechEd:
|
|
Join Us for These Upcoming ASUG Webcasts |
|
We hope to see you in person in Austin or Las Vegas
Nice example Tammy - I enjoyed reading it!
Based on the analysis and due to my soon-to-be-40 age, I am not very likely to "churn" my bank.