BACKGROUND: Experian Corporation is an information solutions company that provides the direct marketing industry with lists of addresses and phone numbers, and analysis of their direct marketing offerings.

PLATFORMS: HP Vectra, 4/300, with 160-256MB RAM, 10MB hard drive.

PROBLEM SOLVED: As a systems analyst/modeler, my group is responsible for developing current models to estimate various values for information contained in the National Consumer Database and the Experian InSource Data Repository. WizWhy has been an important tool used in our data modeling efforts and helped to solve the problem of data modeling of incomplete information at the individual household level.

PRODUCT FUNCTIONALITY: While we have used WizWhy to develop some of our rules-based models, I have found it most valuable to identify key predictive variables to be used in logistical regression analysis and to find the natural segmentation of the data. Given hundreds of variables for hundreds of thousands of records, WizWhy trends analysis returns a predictive rating for each variable. This, used in conjunction with the rules analysis, allows the modeler to tailor models to the data at hand. This is especially useful when percent of fill for each variable varies widely. WizWhy then provides rules to segment the model based upon the fill for each variable.

STRENGTHS: WizWhy does a good job identifying "then-not" rules. Under certain parameters, "if-this" the result "is not" within the range specified. Our business prefers the direct relationships, "if-this-then-that," and WizWhy allows you to turn the "not" rule off. However, you still need to increase the minimum probability of the "if-then-not" rule. You can reexamine the data set and select a query that limits the total data presented more closely to the data range that WizWhy is looking for. If too tight, it jeopardizes the statistical viability of the model.

WEAKNESSES: WizWhy is quite CPU intensive. In complex data sets, WizWhy seriously tasks the processor so it is impossible to multitask other applications while it is computing the rules. If the user cancels in the middle of the identify rules run (i.e., after it has found 2,000 rules), the system will then go ahead and calculate the rules too. The user should also cancel this phase, as the results are based on incomplete information. We have found that for the average PC user, 128MB of RAM can handle most jobs.

SELECTION CRITERIA: We chose WizWhy because we were looking for a rules-based data analysis package and it met our criteria.

DELIVERABLES: Overall, we have found WizWhy to be user friendly, enabling the user to easily work with a Microsoft Access database on the first try. WizWhy provides the user with a look at the data from a perspective that is often missed in pure regression or neural networking environments. While not the all-inclusive answer to modeling, it is a valuable tool to understand our data and predict occurrences.

VENDOR SUPPORT: When we have had questions, the technical support team has been responsive, answering most of our questions immediately over the phone.

DOCUMENTATION: We found the documentation to be adequate.

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