Table of contents for Information on Demand 2011
- Opening Keynotes at #iod11
- Business Analytics Optimization Keynote #iod11
- Analytics Keynote #iod11
- Transformation in the era of big data and analytics #iod11
- Business Analytics – the power to meet your priorities #iod11
- Decision Management Orchestrating Consistent Enterprise-wide Decisions #iod11
Erick Brethenoux, Pierre-Henri Clouin and Asit Dan presented on IBM’s Decision Management approach – a nice chance to see both the Business Analytics and WebSphere bits of IBM talking about the same problem. CEOs consistently tell IBM that volatility, uncertainty and complexity are continuing issues. CIOs with a mandate to transform the business are responding to this by looking to drive better, real-time decisions (75%). Meanwhile CEOs say that reinventing their customer relationship, getting closer to their customers and better understanding the, is critical. But this only works if companies have the operating dexterity to respond to this.
Like me IBM sees a range of decisions – from big, complex strategic decisions about differentiation say to front-line, operational decisions about a single claim or transaction. Operational decisions are focused on identifying opportunities to increase profitability to better manage risk or fraud and to ensure compliance. IBM uses the OODA (Observe, Orient, Decide, Act) loop as an organizing framework for these different decisions (the OODA loop is one of the enablers described in my new book, Decision Management Systems).
Decision Management is a business discipline for automating, optimizing and governing repeatable decision decisions. IBM talks about the combination of Operational Decision Management and Analytical Decision Management to put together Decision Services – business rules and predictive analytics complementing each other to automate decisions. The graphic below shows how this fits together. They had a great example of detecting rare diseases using business rules that were derived using data mining and predictive analytics.
The Analytical Decision Management component is focused on using the data you have to build predictive models so that the insight in the data is available to a decision service. But this insight needs to be complemented by business rules to turn predictions into actions, hence the operational decision management. And, of course, all this needs to be tied to you performance management and business intelligence systems so you can see how much impact it is having.
Eric gave a great demo of a claims system that uses SPSS Decision Management and its support for rapid model construction, integration with SPSS Modeler and for business rules. He also made the point that these decisions are being embedded into existing systems with a minimum of disruption – to Siebel CRM systems, to chat messages, to green screens and websites.
Pierre Henri talked about the Operational Decision Management side of the house, IBM’s newly integrated business rules and business events offering. This allows you to manage much larger numbers of rules and to integrate these rules with the various kinds of predictive analytics being discussed at IOD. They have three focus areas for this product – Adapting to rapid change, Align business and IT to improve collaboration and Act with precision and reliability. This empowers the business to own this decision-making, externalizes decisions for reuse and manageability and delivers real-time decision making.
Integration between the two can be done using SOA or by bringing PMML from SPSS Modeler into the Operational Decision Management platform.
Asit came up at the end to discuss some patterns for deployment. Obviously the key ingredient is a repeatable, operational decision – either helping someone identify the best course of action to take next or driving a system to act appropriately. These decisions often get embedded into operational business processes, making them very hard to change independently and hard to improve analytically (see this paper on the decisions at the heart of your process). Managing decisions separately dramatically simplifies processes and makes it possible to change and manage them independently. Many legacy systems and processes have to be altered to take maximum advantage of decision management technologies.
IBM has identified patterns for this:
- Tactical patterns such as fast tracking simple transactions or coordinating across multiple processes using events and decisions.
- Incremental patterns such as externalizing and automating a decision after removing it from a legacy system or automating manual decisions to increase straight through processing.
- Strategic patterns include reusing core decision services across the enterprise (for customer next best action say), process redesign around customer treatment decisions etc.
Tactical ones involve little or no process change while strategic ones are big impact and disruptive. The incremental patterns fit in between with some impact on processes but at a manageable level. Asit gave an example of an insurer that quickly got started by fast tracking claims to reduce fraud exposure. As they externalized decisions with rules and analytics they started to see how this could be used to increase straight through processing by automating the approval decision. Finally they started to see how this could be used to redesign processes around a more customer- and decision-centric approach.
I have a webinar coming up this Thursday to discuss how to use business rules and predictive analytics to build Decision Services.