Oracle Real Time Decisions at Dell #oow10

I blogged a couple of internal Oracle sessions on RTD (Oracle Real Time Decisions in e-commerce #oow10, Oracle RTD Roadmap #oow10) and now we get a chance to hear a customer speak – Dell with some help from Objectifi. Octane is Dell’s brand name for their RTD project – the platform has a number of other elements but is based on RTD.

Dell originally looked for simple to integrate predictive analytics engine to model using existing real-time data, a multi-use tool for tech support, sales and more across all channels and started Octane. They wanted to be sure they could support real-time data, that the data captured in every channel could improve the results of decisions in every channel and, crucially, they wanted the decisions to learn. Octane changes Dell’s approach from static content and offers in email campaigns to a system that learns in real-time which meant a dynamic content and targeted/needs-based offer engine.  Octane is planned to support Tech Support, Sales, Marketing, Telecom and Online. They hope to use it to drive offers for the call centers, dynamic marketing messages, intelligent call routing and web front ends. Octane is initially supporting offline sales and offline service with a goal to expand out to support channel partners, email, social, chat and online sales/service.

The first use case was online service, support by Siebel CRM. 20,000 tech support agents use Siebel CRM and handle 70,000 requests/day across 15 countries and 30 languages. Dell wanted to drive offers into this environment to enable up and cross-sell. Started with 150 agents in Canada and were encouraged by the results while driving out to about 4,000 agents at present. The system delivers the top 5 offers along with the associated script as part of the Siebel CRM environment. Results have been very positive – close rates offline have gone up 15-20% and margin has risen 10-15% while revenue per call has risen 5-10%.

The second use case is a .NET application for 15,000 sales agents that handles 80,000 orders a day from non-web channels. 500 agents are already on the platform and once again Octane delivers offers and associated information into the UI. This is too new for much in the way of results but looks likely to generate significant incremental revenue.

Moving forward, the opportunities for Octane are moving online. 150M resolutions are served online today, for instance, but no cross-sell or upsell is performed. 13B page views on the main e-commerce site also offer an opportunity.

Octane uses customer profile data is made available based on 5 years of consumer historical data with 1Bn records is flattened into usable format for things like total spend.The architecture is a classic decision service with either JSP or ESB/Web Services access from the UIs. Data warehouse data is fed into a customer data store for use in Octane for performance reasons. Supports 2,400 concurrent users/hr running450,000 transactions/hr/JVM with clear opportunities for further optimization.

Lessons learned:

• Don’t get in the way of the agent – drive behavior with incentives and training for instance
• Establish KPIs and tracking a baseline/control group to make sure you could see the benefits
• Include the customer experience, don’t allow sales to overtake service
• People and markets change so continually review the models
• Data is key, cross-channels and normalized
• Think cross-channel from day 1 so establish offer governance and think about things like translation and markets
• Its not another rule engine but a decision framework for model-based optimizing of real-time behavior
• Remember that offers don’t have to be something you sell – can be something you want to learn or ask. Next Best Action not Next Best Offer.

Dell wrapped up by emphasizing that the overall ROI they got was extremely strong – just 2-3 months to get recoup their investment. They really like RTD as a product and felt that their partnership with Oracle and the RTD team was excellent.