The Direct Marketing Association Research Council recently sponsored a very useful seminar in which leading vendors of analytic tools for marketing presented their most up-to-date technology. Rather than the usual vendor presentations where there is an attempt to avoid overt sales pitches, this was intentionally set up as a "bake-off" where each vendor touted the capabilities and advantages of their products. The vendors represented were Urban Science, Unica Corporation, SPSS Inc. and SAS America.

While I am not going to try to evaluate the pros and cons of each product, there are some clear trends in the industry that I will describe.

Modeling Methodology

The leading analytic software packages all include the leading modeling methodologies (regression, tree models, neural nets, etc.). While a close examination would show some advances and differences, this is not the area in which the vendors are trying to differentiate their products. The idea, of course, is to offer an all-inclusive package so that users do not have to purchase specialized software packages for each methodology. While no package is going to be best in class in every methodology, they all seem to be pretty good. For example, I was interested to see that genetic algorithms are now being included. I have been using a specialty software package for this functionality.

The exceptions seem to come from vendors that specialize in modeling software rather than offering a broad suite of marketing software tools. GainSmarts, the modeling tool offered by Urban Science, includes a specialized methodology for finding the optimum transformation and selection of variables. KXEN (which was not represented) also offers analytic tools that are positioned based upon the modeling and variable selection methodologies they employ.

Interface and Automation

One trend in marketing is the desire to conduct marketing campaigns that are differentiated across many market segments. There has been a corresponding increase in the ability of marketers to collect and access data. However, with the old modeling tools an army of statisticians would be needed to analyze all of this data. While this is not a new direction for modeling software, the vendors are making major improvements in the user interfaces and the automation features of their tools. This enables a small analytic shop to build and maintain a large number of models.

The key is striking a balance between ease of use and control. It is important for the analyst to explore and transform the data prior to modeling, to understand and control the methodology used for modeling and to verify and check the results. Using too many defaults inside a black box process is dangerous. But these tools are headed in the right direction. They attempt to give the user control over the details while providing interfaces and options that make execution quick and easy.

Model Implementation

This is where the real action is taking place. Anybody with experience in modeling to support marketing campaigns has numerous horror stories about the difficulty of incorporating the results of modeling into the marketing operational systems. It has required complex cooperation between the modeling analysts, IT departments and marketers. I have worked in companies where the process of scoring and using a model was so inflexible and burdensome that it routinely took three to four months.

Now this delay can be virtually eliminated by integrating the modeling tool into campaign management tools and creating other standard interfaces. Because all modeling steps (from data extraction, data transformation, estimation and validation) are done within one software package, a record of the complete modeling procedure can be captured. This modeling meta data can then be provided (automatically) to a campaign management tool that can use it to score and use the model. The manual processes of documenting the model, explaining this documentation to an IT group and coding the model into the operational systems are eliminated. The presentations at the seminar included demonstrations of Unica modeling results integrated into Unica campaign management, SAS modeling integrated into SAS campaign management and SPSS modeling integrated into Siebel campaign management.

These modeling packages are also making it easier to create custom interfaces between model results and other marketing applications, such as real-time Web site applications. The details stretch my knowledge of IT applications, but this is definitely an area where modelers need to expand their understanding of how their results can be used.

In summary, I thought the seminar was an excellent format for learning about software alternatives. I hope we will be able to hear from additional vendors in the future.

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