PREDICT EARLY MORTGAGE REPAYMENT
Early Mortgage Repayment affected 10.4% of the portfolio in a Dutch subsidiary of a large international bank. The goal of this study was to detect high-risk customers to prevent interest loss.
Using the DataDetective datamining software by specialists from Sentient, interesting insight, relations and explanations were quickly be found in the databases of our client. A scoring model was trained on 20000+ mortgages, and applied to the other 5000.
The DataDetective prediction model was able to select up to 7 times more customers that were going to end their mortgage early than average benchmark prediction models.
ANALYTICAL CRM SERVICE FOR INSURANCE
The aim of this project was to support insurance plus agents in marketing activities and customer relationship management in business-to-business.
Insight into the structure of a business relationship portfolio and a client potential list have been achieved with datamining techniques applied to a data vault managed by Sentient.
By comparing the customer base of a certain insurance office with all data in the data vault, behavior or interest of its customers were predicted.
DATA MINING FOR CRIME ANALYSIS
The application of DataDetective in law enforcement has yielded an impressive track record, when compared to conventional methods.
Analysts estimate that DataDetective has made their work 10 to 20 times more efficient. Searching a person using a description leads to the right match 50% more often.
A comparative study has shown that criminal activity has dropped over 15% due to measures that were based on DataDetective results.
CHURN PREVENTION FOR WEBSTORES
Customers’ churn or under activation affected 87% of the online customer base of a large department store chain. The goal of the project was to build a model which could predict customer behavior in online ordering.
Based on this prediction high potential customers can be approached for reactivation to prevent churn or under activation. A database with 630.000 transactions was used to build a prediction model.
Final results showed this model predicted with 68% accuracy whether a customer would churn. This amounted to a potential extra revenue of €25mln per year for our client.