David Raab Article Archive

Empower (Saxe Inc., 303-770-3300) may not be the last word in database marketing systems, but it is surely the latest. With object oriented programming, advanced indexing methods, multidimensional data models, and multithreaded parallel processing, Empower is a virtual showcase for the newest techniques. The results are genuinely impressive.

Stated simply, the benefit of Empower is that it allows large-scale database marketing on a standard relational database. This implies two main capabilities: adequate speed for complex analysis and selections; and an interface with sophisticated query, segmentation and reporting capabilities. Both are difficult to provide with standard relational database tools. Other vendors who have tried with reasonable success include EDS/dbINTELLECT (303-275-2440), Dun & Bradstreet Information Services (201-605-6270), and Mercantile Software Systems (908-981-1198).

Like most of these competitors, Saxe has chosen to improve relational performance through specialized software techniques. Empower employs a proprietary technology, called “Hoops”, which both maintains lists of records with specific data values and establishes linkages among data elements in different tables. This enables Empower to resolve queries without the time-consuming joins that slow down conventional relational databases. Hoops can also assess the level of detail required by a query and point it to a table of summarized data when appropriate. This improves speed by reducing the amount of data to be read.

Actual performance depends on the task at hand. A query that requires only a record count is returned in two or three seconds even when it involves several million underlying records. Creating a list of the records themselves might take two or three minutes against the same file. An extract of the data associated with the records would take longer still–perhaps twenty minutes for two million records with a couple hundred fields each.

A typical Empower database will include separate customer and transaction data levels, although more complex structures are possible. The data is stored in denormalized “star schema” structures that separate the analytical elements–things like customer type, transaction type, time, product, demographics and location–from the underlying details. Analytical elements can be summarized in multi-dimensional “cubes” that resemble large, many-level cross tabulations. These cubes can be accessed extremely quickly.

Happily, all of this complexity is well hidden from the Empower user. The system’s other major technology, object oriented programming, has enabled Saxe to build an exceptionally integrated and flexible graphical user interface.

Queries are created either with a point-and-shoot builder or by writing Structured Query Language (SQL) code. A drill-down multidimensional interface is expected by the time this column is printed. Users can build complex selections incrementally, by limiting subsequent queries to the records already selected. Although the basic query interface includes only the standard logical selection capabilities (equals, greater than, less than, including, excluding), more complicated formulas can be written in SQL, saved as custom functions, and embedded in selections when needed. Users can store the list of selected records, browse the underlying data, and extract specific columns or subsets of records in a variety of formats.

Reports can combine data from separate databases and can be limited to groups created in a new or preexisting query. The system is delivered with standard reports that are customized for each client. User-defined reports include line lists, a 5-dimensional cross tab and a columnar format. Cross tab cell contents are limited to record counts, while the columnar report can show other kinds of information and allows each column to be limited by a different query. Object-oriented programming allows report output to be viewed in tabular formats, as graphs or on maps without any data conversion and lets the user select, sum, sort and perform other operations on the rows of a report as if it were a live spreadsheet. Objects to support statistical modeling will also be added in the near future.

Empower provides sophisticated campaign management. It supports multi-segment promotions spread over different drop dates, distinguishes test from control segments, and allows Nth, random, ranked and other cell splitting mechanisms. The system captures budget and actual expenses, allows the user to define return on investment calculations, and will take snapshots of response at user-defined intervals by multiple, user-defined groupings of segments. The system can assign priorities to different promotions, can assign an independent query to each segment, and will control duplications across segments. However, there is no graphical representation of the campaign hierarchy. This makes it somewhat difficult to control deduplication and priorities across segments.

The system will execute an entire campaign selection at once, but the actual queries for different segments are executed independently. This is probably less efficient than making a single (though complex) pass through the file, although Empower actually will run several queries simultaneously if multiple processors or processing threads are available.

Empower databases are all currently updated at Saxe and then either stored at the vendor or loaded on the customer’s own machine. A customer-based loader/builder is due at the end of June and a generalized version with a graphical user interface is due by September. The system can handle incremental updates, processing 4 million transactions against a 7 million customer, 30 million record database overnight. This includes address standardization, geocoding and householding. The system permits online access to individual customer records through lookup and editing screens.

Existing Empower installations run on the Oracle 7 relational database, although Empower has been tested with Sybase and can link to other ODBC sources. The system can run on Sun Solaris, Microsoft Windows NT or Windows 95 servers, and on the same platforms plus Windows 3.1 for client workstations. The largest installation has over 50 GB of data on 28 million customers and over 100 million transactions.

The system has been in development since 1993 and was officially released in June, 1996. There are two operational installations with three more underway as this is written. Empower is priced at $250,000 for a server license and five workstation licenses. Implementation costs at least $75,000 and can reach many times that figure depending on complexity. Update processing at Saxe and the client-based update software are additional.

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.