Oracle Automates E-Commerce Personalization

The concept of mass personalization has long been the Holy Grail goal in marketing, predating even the 1993 publication of Don Peppers' and Martha Rogers' visionary book, The One to One Future. Technologies such as rules engines and predictive analytics have since emerged to optimize customer interactions. But with yesterday's release of Oracle Real-Time Decisions 3.0, the vendor says it's bringing unprecedented levels of automation to the task.

"Rule- and model-driven systems can present offers or targeted content to specific audiences, but those systems have tended to require a lot of manual labor," says Nicolas Bonnet, an Oracle senior director of product management. "Real-Time Decisions automates that process tremendously with continuous, self-adaptive learning that constantly refines the targeting in closed-loop processes in which the system sees how well previous suggestions have performed."

Based on technology originally developed by Siebel, Oracle Real-Time Decisions is said to be in use by several dozen customers in a range of call center, e-commerce Web site, and multi-channel customer interaction and marketing scenarios. Cross selling, up selling and best-offer optimization have become sweet-spot applications for the product, moving it into a market once dominated by e-commerce and application server vendors like Broadvision, ATG and BEA (the last now owned by Oracle). Oracle Real-Time Decisions also encroaches on Web analytics tools that report on site navigation and e-commerce success after the fact but do nothing to optimize interactions as they are happening.

Oracle Real-Time Decisions is said to complement rules engines and predictive models, adding the benefits of automated optimization. "It's an open system, and the majority of customers are feeding the engine with existing rules and predictive models," Bonnet says. "The system has a self-learning, predictive model that's based on traditional statistical techniques. But what's important is the degree to which the tasks of building, recalibrating and recycling the model have been automated." Thus, statisticians aren't needed to handle these data-mining-oriented tasks, Bonnet adds.

Upgrades to Real-Time Decisions 3.0 include more sophisticated built-in rule-writing capabilities as well as support for batch processing. The latter applies what's learned in real-time interactions to offline activities. "[Our] customers learn a lot about what customers coming to their Web sites are doing. But they want to be able to apply that intelligence to their [batch-oriented] outbound e-mail and direct mail campaigns and lead management efforts," Bonnet says.

The 3.0 upgrade also includes a library of prepackaged "decision patterns" to speed deployment of typical applications such as delivering cross-selling promotions on a Web site or supporting call center customer-retention programs. The upgrade's new decision templates can be used more generically; for example, predicting the time a customer will spend on a Web site. This insight can be applied across a range of use cases, including cross selling, up selling and customer-retention programs.

Oracle Real-Time Decisions has also been embedded behind the scenes in business processes to automate decisions. "We're working with a port authority in the Middle East that uses the product to predict which shipping containers to inspect for customers or security reasons," says Paul Rodwick, Oracle's senior director of product management and strategy. "In this case it supplements a rules-based system, because the bad guys are always testing the edges of those rules."

Embedded deployments of Oracle Real-Time Decisions are in keeping with the goal of automating routine, high-volume transactions espoused in the 2007 book Smart Enough Systems by James Taylor and Neil Raden. For now, however, Oracle's Rodwick and Bonnet say e-commerce uses are better understood and in greater demand than the emerging market for what Taylor and Raden have dubbed enterprise decision management.

Oracle Real-Time Decisions is available pre-integrated with the company's Siebel e-Commerce, Siebel Marketing and Siebel Contact Center CRM- and service-specific applications, but the majority of deployments are said to include non-Oracle front-end applications. The software starts at $100,000 and the 3.0 release is available immediately.