Predicting customer churn in insurance using SPSS Modeler

Post Comment


To learn more about how SPSS Modeler and predictive analytics are helping companies gain competitive advantage, read the white paper Insurance customer retention and growth.


0:00 begin this insurance example in cartons workspace an easy to use unified 0:04 workspace which empowers the typical business user 0:06 with the self service experience 30 footprint web-based infrastructure 0:10 with my customized dashboard a combined historical customer information for my 0:14 data warehouse 0:14 with data exported from my predictive model created in SPSS modeler 0:18 or consuming a state in my car was portal I can specifically see 0:22 which of my customers are most at risk defecting from their insurance policies 0:25 what's up for renewal 0:26 I can also see what the impact of losing his customers will be 0:29 to my overall business based on their customer lifetime value as well as 0:33 which retention offers they're most likely to accept in my data mining and 0:36 analytical work bench 0:37 known as SPSS modeler I can see the model is comprised of a series of 0:41 interconnected nodes 0:42 which visually shows the floor the data starting on the left 0:45 we're looking at the days or snowed which in this case contains my 0:48 transactional 0:49 and demographic customer information like policy type marital status 0:52 and vehicle size as well as customer lifetime value ourself like to find 0:56 which indicates whether customers current her has churned in this field is 1:00 what will use the target heart shaped remodel algorithm 1:03 they were using this model with the library almost forty different 1:05 algorithms to appease even the expert modelers 1:08 it can be difficult to know which algorithm to use so to help the less 1:11 experienced users get started 1:12 there are even some auto modeling notes which will run and compare several 1:16 veterans 1:16 once the open at the modeling algorithm is the predictive model 1:20 which is the Golden Nugget you see here the graphing allows me to view the 1:23 results of the model 1:24 as a graphical output I can go to the end the stream 1:27 where the output is generated and run it where can see the two columns were added 1:30 by scoring model 1:32 active at the customer's predicted remain a customer and canceled if the 1:35 customer is likely to churn 1:37 the column side it shows the confidence are accuracy of this prediction 1:41 finally now that completed the model I wanna share this analysis for 1:44 consumption by my business community 1:46 in my comments be I portal their number of methods of exporting the data 1:50 which include publishing the results as a package which can be consumed by 1:53 report authors in Report Studio 1:55 using the IBM Cognos BI expert note I can use this information to focus on 1:59 those individual customers who are predicted to churn 2:02 I can see this particular customer has a high probability 2:05 and his customer satisfaction is low but he does have a strong customer lifetime 2:09 value 2:10 will want to make your attention offer in order to retain its customer