David Raab Article Archive

Retailers have enthusiastically explored database marketing in recent years, but their industry poses some unique challenges for database marketing systems. One is the sheer volume of data generated by capturing each point-of-sale transaction. While a typical catalog might average two to four items per order, a retailer–particularly in high volume industries such as groceries–might have more than a dozen. For chains with many outlets, the result can be enormous files that occupy huge amounts of storage and need expensive multi-processor systems to operate effectively on a standard relational database.

Most vendors’ response has been to employ specialized, proprietary database engines that are more efficient at handling large volumes. This works well for data analysis and direct marketing programs, but conflicts with another retailer requirement: the desire to distribute marketing data back to the actual point of customer contact. There, at a cash register or customer service terminal, retailers want to identify high-priority customers and provide customized services. Proprietary database engines are difficult to integrate with operational systems for this type of task.

Open MarketWorks (STS Systems, 514-426-0822) takes the alternate route, sticking with a standard relational database. To gain the necessary performance, the system relies heavily on user-defined summary tables that reduce the amount of data processed for most tasks. Open MarketWorks also addresses another problem–definitely not unique to retailers–by providing non-technical users with a visual tool to generate complex queries.

Open MarketWorks files are built primarily with transactions extracted from point-of-sale systems. The system resides on the customer’s in-house computer and can be updated daily if required. To simplify maintenance, STS provides a “Smart Load” utility that can automatically load the new transactions into the system as the transaction files become available. User-defined summary tables can be rebuilt automatically with each load or updated on demand.

Data other than transactions, such as new customer names, address changes or rented lists, can also be loaded in batch processes. The system uses proprietary name matching mechanisms that can be adjusted by the customer. It incorporates Group 1 Software’s PC-based Accumail address standardization, automatically moving the data from the Unix server to the PC and back. The system also includes data entry screens that allow users to edit single records directly.

Household-level processing is not currently available, although the system does allow customers be placed in “affinity groups”. These groups could be households, but are really intended to define larger groupings such as employees of the same company or members of an organization. STS plans to add true householding in the future.

The system’s query combines selections with data analysis. Users choose a variable, are shown a bar graph of the number of records with each value for that variable, and select records by pointing to the appropriate bars on the graph. Several variables can be chosen in sequence to further refine the selection. The resulting query logic and specific list of records can both be saved.

More complicated selections can be built by defining two lists independently and then selecting records that appear in both lists, either list, or one list but not the other. The interface cannot build queries involving ad hoc calculations or other complicated data relationships, although an appropriately skilled user could write these directly in SQL. Open MarketWorks can limit the number of records selected for a list, but does not currently allow Nth or random selections. STS is working to add mapping scripts that allow users to make selections by defining regions on maps.

The system provides a strong set of marketing analysis features. It stores a table of marketing campaigns, and allows each campaign to be assigned several feature codes such as season, offer, and product type. Several mailing lists can be assigned to one campaign, with a separate key code for each list if desired. A list can be designated as a control groups, and then will appear separately on response analysis reports.

Response to a campaign can be defined in terms of coupons returned or activity within a specified date range, and reports can split results by purchase location or merchandise. The system also stores the budgeted and actual costs associated with a campaign and can calculate the profit margin based on promotion and merchandise costs.

Open MarketWorks comes with about ten standard reports. There is no custom report writer, although STS plans to integrate a third party product for end-user reporting. Calculations, scoring and mapping would also be done outside of the system.

The system is written primarily with the PowerBuilder system development tool and C programming language. It requires a Unix server and Microsoft Windows on the user’s desktop. The system can process reports and queries on the server, leaving the user’s desktop machine free for other tasks. It does not allow the user to define batch jobs or schedule tasks to run at some other time.

All Open MarketWorks installations use the same data structure. Customers could make changes but would have to modify the source code to view any new fields on the screens. The standard structure supports individual and transaction level data but lacks a separate a household level. Initial implementations use the Sybase relational database.

Open MarketWorks was introduced in late 1995 and currently has one active installation plus two under way. It is STS’s second-generation marketing database, succeeding Customer Profile System, which was launched in 1988 and has about 120 installations. STS is a major provider of point-of-sale, distribution, merchandising, order entry and other operational software for retailers. The 500-employee firm also offers hardware purchase, system integration and data center operations for its clients.

Prices for Open MarketWorks are based on the number of customers stored in a system and will be charged as an annual license. Fees range from $75,000 to $300,000 per year.

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.