Retailers are especially demanding users of database marketing systems. In addition to the complex campaign design, integrated scoring and sophisticated data analysis, retail systems must deal with huge volumes of data, customize promotions for individual stores and evaluate results without response codes that link transactions to promotions.
They also often want their marketing databases to run loyalty programs, distribute messages to the point of sale and let store personnel look up and change information about individual customers. In many cases, they want to extend their systems still further to coordinate promotions across all media; to service separate retail, catalog and credit operations; and to support merchandising, site selection and inventory analysis. It's no wonder there are more database marketing systems targeted exclusively at retailers than at any other industry.
Ceres (Ceres Integrated Solutions, 919-785-0575, www.ceresios.com) provides a set of retail functions that is unusually complete for a new product. The system was introduced earlier this year and has one current installation with three underway.
Like most of today's marketing systems, Ceres works with standard relational database engines including Oracle, Teradata or DB2. The underlying marketing database is built and updated outside of Ceres, although most users adopt Ceres' data structures for promotion history and campaign data.
The system connects with its data either directly or through open database connectivity drivers. In either case, a “metadata” layer controls how the information is presented to the user. The metadata allows different users to see different views of the data and lets the system adapt to complex data structures such as multiple levels of household, individual and account data. The metadata also lets queries automatically access summary tables and use specialized query language extensions provided by the database engines. These capabilities are important for performance on a standard relational database dealing with huge transaction files.
Ceres can further improve scalability by splitting processing among several application servers. Each server can be dedicated to a specific task such as selections, reporting or statistical scoring.
To create a promotion, Ceres users first define market segments using an interface that lists the available data elements and lets users fill in blanks to define the desired data ranges. While this type of interface is usually limited to simple queries, the Ceres version allows complex logical relationships, user-defined calculations and custom value ranges. Users can limit the quantity selected via random sampling or by ranking on a user-specified data element. The system can display a statistical profile of records in a segment as a tabular report or on a graph. The vendor is adding the ability to browse a sample of the selected records.
Once the segments are defined, users can place them in a hierarchy to eliminate duplicates and can define test/control splits. Contents are assigned on a matrix with segments and splits down the side and promotion offers and versions (i.e., physical components) across the top. The user defines the physical package that will be sent by selecting offers and versions on each row. Column footers show the total number of customers who will receive each offer and each version.
Users can also assign up to five tracking codes per row. The system can split versions by individual stores without creating separate segments for each store. Users can import an Excel spreadsheet that specifies the quantity by store for each segment. The system stores the offer, version and tracking codes on the promotion history file created when a campaign is executed. It also stores the list of products associated with each offer. It will automatically create a coding dictionary that tells the mail house exactly which components should be sent to each code.
Promotions in Ceres are organized in a hierarchy of divisions, event types, events and campaigns. A single event can include several campaigns, such as direct mail, television and newspaper promotions. Campaigns are assigned start and end dates and can be displayed as bars on a calendar that shows all promotions planned for a specified period. The display can be color-coded to highlight media, event types, merchandise, or other differences.
To set up multistep campaigns, users open a promotion on the calendar and drag one of its segments to a later promotion. This will automatically ensure that only people who received the earlier promotion are eligible for the later promotion. Users can schedule a campaign to execute at a specified date or at regular intervals, but cannot specify that one campaign will execute a set number of days after another. The system does let queries use relative dates, however.
Reporting is done through custom reports written with Crystal Reports and Crystal Info technology, which provide advanced formatting, graphics, scheduling, alerts and other functions. A separate set of customer, product, response and decile analyses work from active data sets, which lets users view the underlying detail, display results on graphs, create ad hoc groupings, and save user-defined groups as promotion segments. Maps are created using components licensed from MapInfo, so that map selections can also be saved as segments. Ceres has also licensed statistical components from SPSS, which it plans to use in an automated modeling module in 1999.
The system includes a customer contact management module that lets users look up an individual customer record and add information such as birth date, buying preferences and address changes; but it does not store the callback dates or detailed text notes required for true contact management.
Pricing of Ceres is based on the size of the purchasing company and on the modules purchased. A minimum installation of basic modules for a retailer with $1 billion in revenues would cost about $300,000. This could grow to $1 million for a large retailer with all system modules included. Cost in later years is approximately half the initial cost and includes both system upgrades and custom programming to meet client-specific needs. The system runs on Windows 95/98 or NT workstations.
Ceres was founded in 1995. The company is now 25 percent owned by Acxiom, which is offering the Ceres product as part of its own retail solutions.
David M. Raab is a consultant specializing in marketing database system
evaluation. His e-mail address is [email protected].
