Software developers have highly active imaginations. This lets them create innovative new products, but also sometimes makes it hard to distinguish what their products do from what they plan them to do.
It doesn't help that many vendors speak a language that sounds like English but makes no distinction between the present and future tense. So it's not unusual for a vendor to say a product does this or that and then calmly mention that the feature is due for release in six months. The uninitiated can find this confusing.
E.4 (E.piphany, 650/496-2431; www.epiphany.com) is the product of such creativity. E.piphany — yes, that's how they spell it — is actually the company's fourth name in two years, each reflecting a slightly different business strategy. E.4, released this January, is based on technology the firm acquired from Silicon Graphics and embedded in an initial set of products released in late 1997. That set of products officially included a campaign manager, although it never quite appeared. The campaign management portion of E.4 is scheduled to arrive this April.
The core technology of E.piphany automates the tasks involved in transforming data from multiple sources and loading it into star-schema databases. Because transformation and schema design require large amounts of skilled labor when done manually, this technology lets E.piphany systems become operational in a considerably shorter time — eight to 16 weeks — than would otherwise be possible. It also makes it easy for the system to adjust to new data elements: In fact, it can add a new attribute to an existing file without reloading the data from scratch. In a large database, this type of reload could take days. Once added, the new attribute is available to users.
The data transformation portion of E.piphany's technology also provides substantial productivity advantages. The firm has about two dozen templates for different kinds of data such as backorders, inventory levels and address changes. Each template allows the system to transform the particular type of input data into a transaction — for example, to derive the number of units sold during a period from the change in cumulative sales. This lets the system work with copies of operational files, which are usually easier to capture than the operational transactions themselves.
The generic templates quickly can be adopted to specific systems such as an SAP manufacturing module, Vantive customer-service application or Acxiom enhancement file. E.piphany adds to its library of system-specific transformations as they are constructed so they can be reused should the opportunity arise.
The third major element of E.piphany's core technology is a function to automate the creation of aggregate tables. Such tables contain summary data and speed response because they are read more quickly than the underlying detail. E.piphany keeps track of the data elements required by queries as they are run. By its April release, the system will use the statistics to prepare recommendations to create new aggregate tables and discontinue existing aggregates. If users accept the recommendations, the system will then build the new aggregates automatically and maintain them over time.
Queries automatically check for appropriate aggregates and use them when they are available.
The initial set of E.piphany products provided data analysis and reporting tools that read the files created by E.piphany's core technology. These also included some list generation ability. Under E.4, the tools are replaced by solution modules to support specific applications such as sales reporting, customer acquisition or marketing offer optimization. The firm has defined 16 solutions of which nine currently exist. Each solution contains a combination of prebuilt analysis, reporting and campaign management templates and will have a workflow capability to lead the user through the project at hand. The idea is to speed deployment while also helping users apply the system as effectively as possible.
The E.piphany system runs across the Internet, using a standard Web browser for access. End users are provided with fairly powerful tools for multi-dimensional reporting, graphics, data profiling, clustering and forecasting. It does not provide built-in regression or neural net analysis, however. The new campaign manager will provide multi-segment nested campaign trees, Nth and random selects, revenue and profit projections based on user-supplied assumptions and reasonably complex queries. It will let users run test counts against a sample database and translate the output into quantities for the full database.
A scheduler will control when a campaign is run and allow it to repeat automatically on a regular schedule. But all segments in a campaign are selected at once, so a promotion involving a sequence of contacts over time would need to be split into separate campaigns. The system keeps track of which treatments are applied to each segment and each customer, but does not maintain a list of treatments that is shared across campaigns.
E.4 will be able to send lists of customers and treatment IDs to touchpoint systems such as call centers, sales automation, e-mail and Web servers. But because updates and campaigns run as batch jobs — typically overnight — it will not be able to manage real-time interactions directly. Nor is the system able to access data stored outside of its own file structures. Users might want to do this to read detailed transaction tables that are not used often enough to justify importing on a regular basis.
The various incarnations of E.piphany systems have been installed at about 25 companies in total. Existing installations use Microsoft SQL Server and some are migrating towards Oracle. Some functions of the system, such as rules to control query execution, are tied to a specific database engine, so moving to a different platform such as Informix would require some work. Pricing for E.4 will be based on a combination of the solutions purchased and number of users. Cost begins at $250,000 per solution.
