BACKGROUND: Cullen/Frost Bankers, Inc. is the largest independent bank based in Texas. Founded in 1868, Cullen/Frost is a multibank holding company with $6.9 billion in assets and trust assets of $11 billion as of April 1999. The Cullen/Frost family includes Frost National Bank and United States National Bank. Cullen/Frost has more than 75 branches in 11 Texas cities. Cullen/Frost offers a full range of personal and commercial banking products and services and an array of trust and financial management services through its Financial Management Group.

PLATFORMS: NT workstation.

PROBLEM SOLVED: The customer relationship management (CRM) strategy established at Cullen/Frost involved profitability, customer satisfaction and product efficiency issues. In addition, there was a need for real-time access to customer surveys. KnowledgeSTUDIO provided engines for CHAIDs (chi-square automatic detection systems) and neural nets to address these issues. Because data mining requires the availability of a broad range of statistical tools, integration with SAS and S-Plus was a major concern. KnowledgeSTUDIO had import engines for both packages and a pseudo-code generator for SAS. The final hurdle was the budget. The product fit very nicely and allowed completion of our toolbox in an economically feasible manner.

STRENGTHS: The engine behind KnowledgeSTUDIO gave excellent response time. Models were produced with relative ease. The learning curve for the CHAID product was very short. The ability to produce SAS rule sets and the presence of import drivers for SAS and S-Plus made the product an excellent fit for our shop. The low cost relative to other mining products was also a nice benefit. CHAID analysis is a very effective way to detect interactions in data sets with a binomial response variable. It has been used in database marketing for some years and is just now making its way into mainstream analysis.

WEAKNESSES: The SAS code generator dropped an end statement on the outer DO loops. This presented productivity problems until it was fixed in Version 2.0. The tree print facility lacked screen editing capability.

SELECTION CRITERIA: The three major reasons why KnowledgeSTUDIO was selected include:

  1. It provided efficient and high-quality tree analysis and contained neural net modeling capability.
  2. It provided flexibility in import drivers and rule generation.
  3. It was a cost-effective solution.

DELIVERABLES: During the last three months KnowledgeSTUDIO has been used to produce eight tree models and five rule sets built from external cluster analyses. The import engines and pseudo-code generator greatly facilitated information transfer in our data mining environment. The models were developed on statistically sampled data with a large number of variables. I was pleasantly surprised at how quickly the results were presented and, subsequently, at how easily the pseudo code was generated. With that kind of turnaround, additional relationships were examined.

VENDOR SUPPORT: The response to problems I encountered was excellent. This was true of both short-term and long-term fixes as with the pseudo-code generator. The principals at ANGOSS were not shy about jumping in and helping out when problems or complex questions were encountered. We elected to install the product in an NT environment. The ANGOSS system engineers facilitated this installation with prompt and courteous knowledge transfer.

DOCUMENTATION: Using the documentation, the learning curve was short compared to other packages I have used. The product was easily employed and results were obtained quickly. I depended on documentation for 95 percent of the learning curve and the help line for five percent.

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