Oracle Real-Time Decisions (RTD) is Oracle’s core Decision Management product and I have reviewed it before (see this First Look – Oracle RTD 3 and this one on the Oracle RTD Roadmap plus I have written a detailed Oracle RTD product review). With the release of “Oracle RTD Base Application Release 3.1” in May 2011 Oracle introduced their new “Decision Manager” application. Oracle RTD now addresses both decision management (allowing the business to manage their decision logic and the lifecycle of decisions by separating decisions from the main applications) and decision optimization (adaptive targeting and self learning for cross-channel decisions).
RTD’s execution model is as a pure play Decision Service. It provides a real-time closed loop self-learning component for use in business process execution – the business process provides data, gets answers or recommendations from the Decision Service and then continues to execute based on the recommendation. Feedback from the process is used to continuously improve the decision making. RTD Decision Services are designed to optimize business processes in real-time using both historical and real-time data by recommending the best asset or treatment in context.
RTD Decision Services have logic expressed with a business friendly framework, integration across technologies and it automatically handles control groups and continuous learning. RTD does not require the promotions, content, products or offers it deals with to be managed in RTD – it can pull them from third party or Oracle asset management systems. RTD handles goals, eligibility rules, predictive models, arbitration, automation and reports. Users define what is “best” using multiple competing priorities and can easily plug RTD in to their existing systems.
RTD is mostly used in marketing and sales processes around customer experience, next best action and personalization. However maintenance recommendations, operational process optimization, risk and fraud, service treatments and more are also emerging as use cases. In addition, more and more customers are using RTD for multiple decisions – often after an initial success. Several have selected RTD as a common decision platform across the enterprise. They have a wide range of customers including Dell (who presented at Oracle OpenWorld ), Betfair and others in telecommunications, financial services, insurance, high tech and retail. The closed-loop/self-learning capability is critical to its success and RTD is often able to generate 100% lift over business as usual.
With the new release, RTD has three main components:
- Learning Engine
To automatically learn from any interaction - Decision Engine
Uses rules and automated / offline predictive analytics to drive decisions - Decision Manager application
A collaborative environment to manage, analyze and refine decision strategies
The last of these is a new product – Decision Manager – that provides an enterprise view of cross-channel decisions and a collaborative decision management environment for business users. It has a number of new features:
- Role-based user interface for authoring and managing decision logic (business rules) and their associated data
- Search for products, offers and other data associated with decisions across multiple source systems
- Version control and audit
- Multi-user collaboration
- Visualization and reporting
Decision Manager extends the core RTD product (which focused on decision optimization) and is a web-based application for business users to manage their decisions. It does not require technical know-how to use, a key requirement from their large customers. The decision repository is displayed using a folder-based structure specific to the installation. Each user sees a different view of this, based on their role, and can get basic information about everything in the repository.
Offers are grouped into campaigns and are a core object with supporting creative for each (banner ads, scripts etc). These have basic eligibility information (start, end, regions etc) and users can define additional business rules for more focused eligibility. These rules can use any customer data attributes including session information as well as information about the choice or application context. A wide range of performance information is available with things like correlations drivers for success (what makes someone likely to respond to an offer), trends over time etc. Every change is recorded for a complete audit trail.
Multiple perspectives are supported. Viewing Offers by Campaign is how a marketing manager might view things for instance but a channel manager might see things differently and the repository can also be viewed with a Channel perspective with Placements (pages for instance) and Slots within those Placements being the organizing principle. Slots are of specific Slot Types to make sure the right creative is used. Slots and Slot Types can be analyzed in terms of performance, allowing the various contexts in which decisions are being made to be analyzed as well as the decisions themselves. For instance, the most successful 100×200 banner can be determined or the factors that are most effective at selecting the right ad to display in a particular slot or how a given offer’s effectiveness varies depending on which page it is displayed and much more. The focus on implementation details at the front end means that the decision service is not just returning, for instance, an offer but a suitable piece of creative without becoming channel specific.
Regardless of the perspective being used the product manages a core set of objects: Campaigns have Offers which have Creatives. Creatives are suitable for Channels and Slot Types. Channels contain Placements which contain Slots, each of which is a given Slot Type. This object model underlies RTD but rather than managing all this data itself, RTD uses a relational structure to allow instances stored in various systems to be mapped to this semantic model. This allows multiple source systems that contain Creatives or Campaigns etc to be mapped to the RTD model and then used in RTD deployments. Decision Manager can also manage and be the source of content directly.
Decision Manager also supports Projects to define units of work for RTD. Design elements can be bundled into a project and then versioned, tested, deployed and retired as a unit. This allows a non technical user to include their elements and collaborate with content creators or more technical users around a common project Every change made to every element of the project is visible in the audit trail. Finally tags can be used and these tags can also be used for analysis – so a user could see all offers tagged Travel for instance to see where they are being successful.
Once all this is done, RTD as a decision optimization engine uses this metadata to see what offers are possible in a given slot and then learns what works best for different kinds of customers.
RTD’s federated approach goes beyond using data and objects stored in multiple systems. RTD can also execute logic for eligibility by calling out to other systems in defined ways or load eligibility data if no API exists. If a source system has eligibility rules defined for a campaign, for instance, RTD has hooks allowing you to link to this system to evaluate eligibility for any Offer contained within the Campaign. While each such link has to be coded it only has to be done once (so future campaign eligibility rules can continue to be defined in the source system and automatically used by RTD) and RTD allows you to define the propagation approach to be used – do campaign eligibility rules in a third party system impact all offers in the campaign or do rules about offer eligibility impact all campaigns in which they appear. In addition the user interface for managing the repository is based on Oracle technology that supports mashups so you can also link to rules display user interface components in your source system.
RTD is at its heart a self-learning system. The reporting throughout the product shows what the engine has learned. RTD learns what decision making approach works and which approaches do not work in all these different contexts – not just overall effectiveness of an offer but also the effectiveness of an offer with different creatives and placements. All this information is displayed within the tool and can be exported. So the predictive drivers, correlating attributes, ranking etc can be pushed into a third party product or an Oracle product for further analysis or to drive the behavior of some other system.
The management of decisions in Decision Manager is well thought out with good information presentation as you work on the various elements of offer management. The federated approach means that existing content and campaign management systems can be leveraged while taking full advantage of the self-learning approach that is at the core of RTD.
You can get more information on Oracle RTD here – including a detailed Oracle RTD product review that I wrote. Oracle is one of the vendors listed in our Decision Management Systems Platform Technologies report.
Comments on this entry are closed.