David Raab Article Archive

Readers who have spent the past two years in a cave or are not part of the computer industry may need to be told that a data warehouse is a collection of detailed historical information made accessible for business analysis. The rest of you have not only heard the term but probably noticed that these expensive systems are often justified with little more than a promise of “better analysis.” Not surprisingly, many companies later decide they want more specific applications. Database marketing is often at the top of the list.

ValEX (Exchange Applications, 617-737-2244) lets data warehouse owners add database marketing capability. The system does not offer the file building tools of traditional marketing database products, such as address standardization and household identification, on the assumption that these will already be part of the existing warehouse. (If not, Exchange Applications will happily add them using third-party products such as Postalsoft.) Instead, ValEX focuses on the specialized capabilities needed for marketing implementation: segmentation, selection and response analysis.

In fact, the system wraps these functions around a formal methodology, called “Value Exchange” and developed by Exchange Applications director Alan Grant. Because the methodology focuses on improving the profitability of customer relationships, ValEX capabilities extend beyond traditional selection and campaign management to include segment-based investment strategies, customer behavior measures, and profitability calculations. But the system can be used without adopting the Value Exchange superstructure; only a few embedded buzzwords betray its origins.

ValEX databases are typically updated on a daily or weekly basis. The product can read changes in the underlying data structure and automatically incorporate these into the lists of tables and fields that appear on ValEX screens. ValEX can also create its own supplementary data tables to hold sets of selected records or user-calculated values and could post these back to the warehouse if desired.

The product’s real functionality begins with the definition of customer file segments. This is done through an embedded version of Pilot Software’s Analysis Server, a multidimensional data analysis system that itself includes an automated segmentation tool using the tree?based “CART” (Classification and Regression Trees) technique. Investment strategies for each segment can be planned on a spreadsheet and then linked to specific marketing campaigns and measurements of customer behavior.

The system has a graphical query-builder which offers advanced marketing functions such as subqueries, embedded calculations, ranked selects and maximum names per household. Users can also write SQL code directly or have the Pilot CART tool automatically identify the best candidates for a particular promotion. Once written, the same query can be executed against households, individuals or accounts. There is no online ability to review the actual records selected in a query, however.

After the queries are built, they can be used to define groups of records that will be sent a marketing campaign. Another graphical tool lets users build a diagram to define the relations among the groups and how each group will be processed. The system can eliminate duplicates across groups, set a maximum number of records per group, and randomly split a group into test and control panels. However, it cannot use queries to split groups into a more refined segmentation and cannot copy a splitting scheme from one group to another. The system also lacks an explicit way to limit the number of many times a customer is selected across different campaigns, although this can be accomplished through carefully written queries.

Offers and outputs are assigned independently to each group or subsegment, so that members of one segment can be sent a telemarketing campaign while others receive a different offer by mail. The same output can be assigned to several groups and each group can be assigned to more than one output, so a single selection might produce promotion files, statistical summaries, snapshots of each selected record, and promotion history updates. The system tracks a variety of performance data for each segment, including budget and actual figures for revenue and cost.

The actual measures are customized for each installation and are placed into summary tables during each update. These tables can be analyzed with a conventional report writer such as Crystal Reports or through the Pilot multidimensional tool. The system does not include a custom report writer, although the extract module has a “profile” tool that makes it easy to generate three-dimensional crosstabs to load into a database or spreadsheet. Extract itself can perform more complex calculations and aggregations.

ValEX will work with any ODBC-compliant relational database. The application runs on a Unix or Windows NT server and on Windows NT or Windows 95 workstations. It is written in Visual C++ and allows close integration with other Microsoft OLE-compatible applications such as mapping software, spreadsheets and workflow control systems. Performance is determined largely by the underlying database and hardware, although the vendor has parallelized some resource-intensive tasks such as splitting a file segment. ValEX also saves time by storing the list of records found by a given query so they can be used without running the query again.

The product is sold for a one-time license fee based on the number of customers in the buyer’s database plus a charge for implementation services. Total cost for a minimal installation, handling about 500,000 customers and not including development of the data warehouse itself, would be around $300,000 in the first year and about $35,000 for maintenance in years after. Software and installation for 10 million customers would cost about $800,000 in the first year. Building a warehouse itself can run into the millions–Exchange Applications offers a “FastTrack” diagnosis project that alone costs $100,000 to $200,000.

Exchange Applications was launched in 1993 to develop marketing database software that took advantage of open technology and supported the “Value Exchange” methodology. The firm offers consulting in data warehouse planning, systems integration and direct marketing as well as customer management strategies. ValEX was installed at three of the firm’s clients before its official launch in June, 1996. At this writing, three additional installations are under way.

David M. Raab is a Principal at Raab Associates Inc., a consultancy specializing in marketing technology and analytics. He can be reached at draab@raabassociates.com.