REVIEWER: Richard Wherry, SVP market information manager, GWIM Client Analytics and Modeling, Bank of America.

BACKGROUND: Bank of America is one of the world's largest financial institutions, serving individual consumers, small- and middle-market businesses and large corporations with a full range of banking, investing, asset management and other financial and risk management products and services. The Global Wealth and Investment Management division focuses on providing wealth management solutions to clients.

PLATFORM: Bank of America GWIM division has implemented KXEN on a Windows Server 2003 platform.

PROBLEM SOLVED: A major focus is retention of deposits. Due to the current economic climate, the company has seen greater assets move into FDIC deposit products. The bank is focused on retaining these deposits by identifying clients likely to move assets and then creating offers conducive to retention.

PRODUCT FUNCTIONALITY: We used KXEN Analytic Framework, and in particular its robust regression component (K2R). This component allows automated model building and model validation for predicting either binary or continuous variables. Data preparation is automatic for handling missing variables, outliers, adaptive binning, recoding variables, and randomly splitting the data into estimation and validation subsets. Prior to KXEN, we had been using a product from a large vendor to build analytical models, but it was a time-consuming and labor-intensive process to construct derived variables and transformations. The time spent on this process has been greatly reduced with KXEN.

STRENGTHS: KXEN can operate with different data sources within a consistent framework. Data may reside in multiple databases and SAS data sets. In our work, we tested robustness of solutions, using out-of-time samples. The deviation detection functionality tests whether modeling data is significantly different from data to be scored. This is a very appealing utility within KXEN, which has proven useful. Sales and marketing processes can have large lead times, so data can become stale and/or a model may run once a year before needed again, and hence, deviation detection becomes very useful. KXEN’s rapid build time allows us to experiment with regional models and to focus on delivering models that meet the needs and requirements of the business lines we support. With clients across the U.S., the competitive environment for the bank in different regions, therefore, models that can account for regional differences are viewed as more powerful. Due to time constraints imposed by business requirements, we were not always able to fully investigate and develop innovative solutions. With KXEN, we can now build retention models per region and therefore create more targeted offers. The time savings can be utilized to enhance our understanding of the underlying business problem, develop insights and do a deeper analysis on the back end.

WEAKNESS: While KXEN has enhanced its data manipulation in the latest release, we cannot merge SAS data sets directly.

SELECTION CRITERIA: We are not a high-volume modeling shop; however, the GWIM clients’ needs are complex, which translates into complicated and often disparate data. Being able to build models rapidly allows us more time to analyze models and to structure the data in a meaningful way. This leads to better and more actionable solutions. We chose KXEN for its focus on robust models, producing consistent results over time while not overfitting; ability to read and write SAS data sets, export SAS code and SQL code; and rapid build-time. Developing insights from our data is critical to our marketing efforts, and KXEN plays a crucial role in this.

DELIVERABLES: We are using KXEN to develop models within GWIM Marketing, deployed in the form of SAS code for ongoing production retention efforts. Reports and measures on model validation are utilized in our model compliance review and documentation process.

VENDOR SUPPORT: KXEN provided three days of training, consisting of theory, hands-on work, as well as a walk-through of a real modeling problem we are addressing. The quality of the training was excellent. KXEN was accommodating in our need to respect budget concerns and, as such, delivered the training on site at two locations. We have associates with differing levels of expertise and background exposure to data mining methodologies, so KXEN did an excellent job adapting the content of the training to the audience at each location.

DOCUMENTATION: We have found the documentation to be adequate. With training and direct vendor support, our need to go to documentation has been minimal.

KXEN, Inc.

201 Mission Street, Suite 1950

San Francisco, CA 94105-1831

Tel: (415) 904-4160

www.kxen.com

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