CATEGORY: Data Mining and Visualization

REVIEWER: Mark Stewart, senior statistical project manager for Ameritrade, Inc.

BACKGROUND: Based in Omaha, Nebraska, Ameritrade, Inc. is a world-leading provider of online brokerage services and is a leader in leveraging technology for financial and operational growth. As a result of its use of technology, the company is able to offer outstanding customer service and commission fees that are among the lowest in the brokerage industry, while continually expanding the investing tools it offers. Customers can manage their investment portfolios via the Internet as well as through personal digital assistants, wireless phone services and interactive voice recognition systems. Ameritrade, Inc. is a subsidiary of Ameritrade Holding Corporation.

PLATFORMS: Ameritrade utilizes best-of-breed solutions for storing and analyzing the data needed to support its marketing efforts. An Oracle data mart is used as a generic repository, queried and sampled through SAS on high-powered workstations running Windows NT. The resulting sets are mined and modeled through KnowledgeSTUDIO from ANGOSS Software.

PROBLEM SOLVED: Response rates and customer quality were considered to determine the effectiveness of marketing initiatives. Using KnowledgeSTUDIO, we were able to quickly identify variables in a data set to determine if an individual is likely to have the means or inclination to open a discount brokerage account. This information is being used to improve the effectiveness of a substantial marketing budget.

PRODUCT FUNCTIONALITY: KnowledgeSTUDIO has the features expected in today's data mining tools. It has decision tree capabilities that combine the power of CHAID and CART into a tool providing fast, easy-to-understand explanations of data phenomena. This functionality allows quick variable selection to feed into the model-creation engines. There are several options available for model creation including linear and logistic regression, three flavors of neural networks and the ability to build a model directly from a decision tree. Two clustering routines are also available. All have user-definable options as well as logical defaults for people who do not need to know what a hidden node is. Reporting is easy to produce through lift reports, pie charts and bar charts. These reports can be segmented quickly based on any variable in a data set. The aggregate data behind the report can be quickly copied and pasted into a presentation system such as Microsoft PowerPoint.

STRENGTHS: The greatest strengths of KnowledgeSTUDIO are the user interface and its low cost. The program is straightforward and easy to understand. One helpful feature is the sample exercise available during installation, which offers useful examples and walks individuals through the program's process to familiarize the user with the tool. The program is also extremely flexible, providing multiple execution options. For example, the user has the ability to perform sampling, weighting and variable selection during data import, which gives the user increased control in determining the in-formation used for the construction of a decision tree. Additionally, the reporting features make it easy to transfer information into other software programs. Finally, the program is priced considerably less than similar products, offering many of the same features found in data mining tools that cost more than 20 times the price.

WEAKNESSES: Although KnowledgeSTUDIO contains a good set of the most popular analytical tools, some of the classics such as time series analysis and covariance matrices are missing. The product offering continues to expand with each version; hopefully these issues will be corrected as the product matures. The documentation is a bit sparse. A printed manual with several real-world examples would be a plus.

DELIVERABLES: KnowledgeSTUDIO has several forms of output. Charts can be copied directly to formal reports or the aggregate data behind them can be copied into another system. The resultant models can be presented in SQL or SAS code for direct implementation on production systems. The decision tree results can be printed, but be prepared to use multiple sheets of paper and tape if you do not have a large format plotter.

VENDOR SUPPORT: KnowledgeSTUDIO is supported well. On the few occasions we have needed to ask for assistance, it has been only a phone call or e-mail away. If a function not present in the software is desired, it is easy to speak to a developer and outline a plan.

DOCUMENTATION: The product is very intuitive to someone who is familiar with various data mining platforms. For those who are not, training is offered by ANGOSS, and the training materials make a good manual. The electronic help system is good, but (as with most program help systems) can be a bit awkward.

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