REVIEWER: Jorge Portugal strategic marketing director for Banco Espirito Santo.

BACKGROUND: Founded in 1880, Banco Espirito Santo serves more than one million customers at 586 branches in Portugal, 22 in Spain and another 32 offices in 12 countries. The bank's commitment to customer service led its strategic marketing group to articulate a clear mission: implement analytical tools and adopt relationship-building techniques to precisely predict customer behavior and increase customer loyalty.

PLATFORMS: Clementine runs on a Windows NT workstation powered by a Pentium III processor.

PROBLEM SOLVED: Banco Espirito Santo (BES) fights the spread of an eroding customer base every day. As customers gradually reduce transaction activities, they also begin diverting assets to other banks. With Clementine from SPSS BI, BES identifies key behaviors of customers who are likely to leave the bank. My strategic marketing team and I dynamically profile these relationships and build models to test intervention tactics and keep customers happy.

PRODUCT FUNCTIONALITY: Clementine is the bank's main workbench for data mining and model-building activities. The nonstatistical modeling techniques included (neural nets and decision trees, among others) are widely used together with classical statistical models (such as regression and clustering). Future applications will use other modeling options included in the package such as Clementine's various rule induction models.

STRENGTHS: The main strength of the product is the inclusion of state-of-the-art nonstatistical and statistical techniques for customer behavior model building within a friendly and robust development environment. Producing a model in the Clementine environment is faster than in other similar packages.

WEAKNESSES: The main weakness is a lack of functionality for data preprocessing and analysis.

SELECTION CRITERIA: The product is quite easy for beginners to learn and use and provides adequate control of the included models' parameters. For example, the default parameterization of the neural net produces acceptable predictions if the right modeling approach to the problem is taken. This level of accuracy allows the modeler to concentrate on the data approach initially while still allowing some fine-tuning of the model parameters in a second stage.

DELIVERABLES: With Clemetine, we produce models (in C++, Java, etc.), reports and graphics.

VENDOR SUPPORT: SPSS BI took a rapid prototype approach, fully demonstrating the benefits, features and functions of the product at the pre-implementation phase, at no cost to BES. After the implementation, the vendor has provided efficient and helpful answers to users' questions. Special-purpose workshops conducted by the vendor's technical staff have been key in accelerating our users' learning. True partnership, support and confidence in technical competence of the vendor have been key for adoption and loyalty to the product.

DOCUMENTATION: There is physical and online documentation available, both of which are sufficient to operate the product. However, some of the technical descriptions of model output interpretation and meaning (be they neural nets, decision trees or logistic regression) could go into greater depth. For instance, the documentation does not indicate in which situations the model output should be interpreted as a probability and in which situations a confidence or how to convert such measures.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access