BACKGROUND: Australia-based Morris International is a dynamic organization with diverse interests including direct marketing, property, manufacturing, information technology and venture capital. The direct marketing arm of Morris International promotes products via direct mail from Australia to most countries throughout the world.

PLATFORMS: GainSmarts is Urban Science's proprietary data mining tool that sits on top of the SAS Base System. Supported platforms include Windows NT, Sun Solaris, Compaq True64 and HP-UX. A version for IBM's AIX will be released next year.

PROBLEM SOLVED: The direct marketing industry is a competitive industry. The name of the game is to select people from a database that are likely to respond to your promotions. Traditional analysis methods such as RFM (recency, frequency and monetary) are now being superseded by more advanced knowledge discovery and data mining (KDD) software and methodologies. KDD software allows Morris International to build powerful predictive models that facilitate better selections of names than traditional RFM. These models consider many additional variables and sift through large samples of historical promotion data, looking for customer buying patterns. The goal is to reduce costs by mailing less to nonresponders and increase revenue by mailing more to responders.

PRODUCT FUNCTIONALITY: GainSmarts provides a number of modeling methodologies. Some of these include binary logistic regression, neural nets and decision trees. GainSmarts also performs an impressive number of variable transformations, interaction checking and statistical tests on data prior to any attempted modeling. It is the view of Morris International that the preprocessing functionality, which is largely automated, is what gives GainSmarts models the edge in performance. GainSmarts is open architecture, so complicated tweaking of parameters is quite possible. This functionality is hidden from view unless you need it. GainSmarts produces comprehensive text- based reports on the entire modeling process including gains tables and graphs. The graphical user interface is friendly and not overly complicated. GainSmarts imports SAS data sets and views, ODBC databases, flat files, delimited files as well as Oracle, Informix and SQL Server.

STRENGTHS: The biggest strength of GainSmarts is raw performance. The binary logistic regression models built with GainSmarts outperform all other response models we have built with competing software. GainSmarts has an excellent default mode for model building, which means you can build high-performing models without having to manually tune every parameter. Of course, you have the option of doing this if you so desire. The default mode performs transformations and carries out many statistical tests automatically. The output files provide a thorough analysis of the model building run. Being based on SAS, GainSmarts is capable of dealing with 100+ variables and several hundred thousand record training samples in reasonable time. In summary, GainSmarts provides a nice blend of power and ease of use.

WEAKNESSES: GainSmarts would benefit from better graphical representation of results, and we found that the interface really requires 1024 x 768 resolution to run effectively.

SELECTION CRITERIA: Our primary criteria when evaluating knowledge discovery and data mining software was the performance of these tools on a binary response model (buy/won't buy). We chose GainSmarts because it gave us the highest response rates and highest overall profit when tested over unseen historical promotions. We pitched GainSmarts logistic regression model against other data mining products using advanced methodologies such as neural nets, and it still came up superior. Another desirable quality of GainSmarts is the ability to build powerful models quickly due to a well-tuned default mode. GainSmarts did not impose any limitations on us in terms of the number of variables we could have in our models or on the quantity of samples we could use for model training. Models with several hundred variables and several hundred thousand samples could be built in reasonable time.

VENDOR SUPPORT: The support provided by Urban Science is excellent and sets a high standard for the industry. Urban Science generally releases quarterly upgrades of GainSmarts on CD-ROM. Urban Science addresses problems immediately and generally responds with advice or upgrades within 24 hours. They are committed to delivering a quality product.

DOCUMENTATION: The GainSmarts documentation consists of online, context-sensitive help and a glossy, 300-page, color manual. The documentation covers all functional areas of GainSmarts and explains how to use all its features.

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