|
Activity Number:
|
200
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Business and Economics Statistics Section
|
| Abstract - #308160 |
|
Title:
|
Scoring Customer Probability of Runoff for Retention
|
|
Author(s):
|
Huaiyu Ma*+ and Deniz Senturk-Doganaksoy and Kareem Aggour
|
|
Companies:
|
GE Global Research and GE Global Research and GE Global Research
|
|
Address:
|
One Research Circle, Niskayuna, NY, 12309,
|
|
Keywords:
|
retention ; runoff ; survival analysis ; regression ; prediction
|
|
Abstract:
|
Effective customer retention is key for enabling a company to grow organically. Developing robust models with high accuracy on customer retention is difficult. A customer's past behavior as well as other factors such as customer interactions must be among the predictors for a model to be effective. Capturing these types of predictors correctly and making them actionable is not trivial, however. Multiple statistical techniques are used for capturing these metrics. Survival analysis is then used to create X's that capture the past customer behavior. Logistic regression is used to score customers and cluster analysis bucketizes them based on their likelihood to runoff prematurely. These scores provide a means to make retention actionable for the business marketing and sales teams. Sales and marketing campaigns have already been launched based on the results of this modeling work.
|
