It is my pleasure to share the registration link below for your chance to meet active members of the Oracle Real-Time Decisions Customer Advisory Board.
Join us to hear how leading brands across the world have achieved tremendous return on investment through their Oracle Real-Time Decisions deployments and do not miss this unique opportunity to ask them specific questions directly during our customer roundtable.
Gone are the days when “electronic billboards” for targeted merchandizing
programs were leading edge.
Over the course of the last few years we observed a dramatic
qualitative shift in how companies have applied Analytical Decision Management
techniques to drive Customer Experience Optimization programs. It used to be
the case that marketers were happy when they were allocated a dedicated piece
of real-estate on their company’s web site (or contact center or any
interaction channel for that matter) that they could use at their discretion for
the purpose of one-off targeting programs. What companies now want is granular control
over the whole user experience so that the various elements composing this cross-channel
dialog can be targeted and relevant. Such a shift requires a new approach to
analytics, based on understanding how the various elements of the user
interaction relate one to the other.
Let me introduce the concept of Decision Graph in support of
To move from electronic billboard / product spotlight optimization
to customer experience optimization, analytics must shift away from focusing on
“the right offer for the right customer”. The focus of the Decision Graph is to
identify “the right user experience for the right customer”. This change of
focus has a critical impact on your analytics requirements as one dimensional
targeting approach for matching customers with offers won’t address the need to
optimize multiple dimensions at once.
This is where the Decision Graph comes in. Let’s consider the following graph eliciting
the relationships between the various facets of the user experience to be
optimized in the context of a Marketing Optimization use case.
Now imagine that for every offer presentation on any
interaction channels, your analytical engine can record and identify the
characteristics of the customer interactions that are associated with success (say
click or offer acceptance) across all those dimensions.
Let’s take an example.
You see a nice picture of
bear cubs on a forest background with a punchy banner stating “please give
us back your share of the 20,000 tons of annual account statements” call
to action to sign-up for electronic bill payment on the “recommended for
you” section of the login page of your financial service web site and … you
decide to click on the “one click wildlife donation” link.
Our Decision Graph, can
then record the fact your customer profile is positively associated with
“positive responses” to marketing messages in the following context:Channel (Web), Offer / Product (Electronic
Bill Payment), Creative (The Bear Cub image), Tags (Environmental, Wildlife,
Donation, Provocative), Slot Type (Image), Slot (Recommended for you),
Placement (Login Page). As predictive models are attached to the Decision Graph,
this means that such a business event updates 10 predictive models that marketers
can now use for reporting and decision management purposes.
You can now generalize the idea and imagine that this graph
collects information about all marketing events across all channels and you
end-up with an analytical system that let all the actors of customer experience
optimization discover the relationship between the different facets of user
With the Decision Graph
will learn about customer segments that are receptive to eco-centric marketing
messages and which customers in the right context will step out of their
standard routine (why they came to web site in the first place) to
subscribe to specific causes.
Web user experiences
stakeholders will learn about which type of marketing messages are
appropriate and for whom at the start / at the end or throughout a logged-in
Content owners can focus
their digital agencies on the most effective creative themes as they will
be able to correlate response rates based on associated tags
And the company as a whole
will have learned who is receptive to eco-centric marketing messages when
displayed in a given context of a secured dialog from which it will be in
a position to dynamically tailor user experiences across channels based on
such empirical evidence
Now contrast this with a system that would only record the
fact you’ve subscribed to the Electronic Bill Payment option as part of the “Go
Green” Campaign and you will get a sense for the power of the Decision Graph. The
bottom line is that companies need analytical systems that operate at multiple
levels of the Decision Graph if they want to delight their customers with relevant
My next post will be on how the Oracle RTD Decision Manager
product enables you to create and configure such graphs and to automatically identify
the predictive drivers of response across the whole spectrum of the user
It is our pleasure to let you know that Oracle just
released a new version of the RTD platform addressing some important
scalability requirements for high end deployments.
The driver for this new version, released as a patch on
top of the RTD 184.108.40.206 platform was the need to address increasing volumes of
“learnings” generated by ever-increasing volumes of “decisions”.This stems from the fact several of our high-end
customers now have production environments with multiple cross-channel Inline
Services supporting multiple “decisions” which make use of multiple predictive
models using hundreds of data attributes to select from potentially thousands
of “choices”. Addressing those high-end business requirements required increased
RTD Learning Server capacity than was provided by RTD 220.127.116.11.
To address those needs, Oracle re-architected its RTD
Learning Server engine to enable some level of parallelization. This new
architecture relies on multi-threaded models updating and asynchronous learning
records reading/deletions operations. This change provides a 150% improvement
in learning record processing rates, which enables RTD to now process more than
58M incremental learning records per day with a deployment configuration
consisting of 3 concurrently active inline services each with 900 choices, 200
data attributes, and 4 choice event predictive/learning models. This was
achieved on a machine with 4 core / 6GB RAM allocated to the RTD learning
This new version of RTD is an important release for companies
setting-up Big Data Analysis & Decision platforms in support of real-time
and batch targeted customer experience enterprise deployments.