Decision making starts with understanding the situation—what happened, what normally happens, what are the possible outcomes? This is the realm of data warehousing and business intelligence (BI). Over the past decade or two, the technology, practices and skills in this area have become, at this point, fairly "smart."

enlarge

Getting there from here Starting an enterprise decision management project involves several steps: > Organizations must focus on the identification and management of decisions at the operational level. To date, the focus has been on data and information, but decisions themselves have been somewhat overlooked. Because it is difficult to implement even moderately complicated decisions in operational code, these decisions are handled manually or with trivial logic. > Organizations must adopt some new technologies such as business rules management systems (and the skill for developing rules models), data mining and predictive analytics and possibly the orchestration of services. > IT staff must gradually cede rules development and management to domain experts, and analytics staff should consider the implementation of their recommendations, not just the presentation or mathematical validity of their models.

Running a modern business requires systems to keep track of events. Systems register events as they occur, safeguard their footprints as stored data and, albeit in a mechanical way, react to those events. Generally speaking, these systems are not very smart. As a result, many of the business decisions companies make tend to be hidden in systems that result in poor decisions, if they get made at all. In fact, most systems aren't configured to learn from the past and therefore struggle to keep pace with change.

Although many organizations believe the answer is to implement newer, "intelligent" systems, much of today's existing technology has the potential to be "smart enough" to effect change for an organization. What is needed is a new approach, called enterprise decision management (EDM), which combines the realms of operational and analytical processing.

While BI focuses on data, operational systems focus on business processes. This separation led to the development of data warehouses. It is clear now that "smart" systems require the convergence of data and business process technologies with a focus on decisions. Because business decisions influence customer, investor and other stakeholder perceptions, these decisions cannot be left to "dumb" systems.

Most enterprise applications are "dumb"—they manage transactions and store events. To the extent that they have embedded "rules" about operations, they are limited in scope and relevance because of the inherent complexity of decision making and the rate at which decision-making rules change.

Instead, when a decision is required, most systems allow for human interaction, which often slows the process. Data warehouses and BI have used the data stored in these enterprise applications to inform managers about the past, but only after the data is extracted and integrated into a separate set of models that cannot be related to the original schemas. No matter how "smart" the BI applications become, this smartness cannot flow backward to the original applications, which remain difficult to modify or enhance to reflect new understanding. This lack of learning disenfranchises executives, because the systems that run their businesses are not amenable to change even as those executives use their BI environment to better understand what change is needed.

As data warehouses are rapidly growing, the demand for fresher data—even approaching real time for certain subject areas—is rising. The BI industry is reacting to these challenges with better, faster, more scalable and more flexible approaches, but best practices remain focused on past events. While the use of analytics in "predictive reporting" is growing and making for better-informed knowledge workers, traditional BI tools—reports, dashboards, cubes—are not going to make operational systems smarter; EDM can.

The EDM framework
EDM is a framework for using existing technologies to unify an organization's analytical and operational systems, imbuing the operational systems with the "smarts" of BI and bringing analytical processes into operations. This approach aims to automate and improve high-volume operational decisions. In doing so, it:

	Develops decision services using business rules to automate those decisions
	Adds analytic insight to these services using predictive analytics
	Allows for the ongoing improvement of decision making through adaptive control and optimization

Beginning this process requires attention to five concepts:

1.	Focus on high-volume, operational decisions. Forward-looking, large-scale, contemplative decisions, such as new product strategy or acquisitions, are the domain of BI. EDM excels at quickly making numerous decisions that are relatively straightforward. Those involving the repeated analysis of large volumes of data are good EDM candidates.
2.	Externalization of the rules that drive these decisions. The rules that exist in operational systems are usually embedded in code or so implicit that they cannot be understood without examining the source code or program documentation. This makes them difficult to modify. The EDM approach abstracts business rules to a business rules management system where they can be managed in the open and shared across a wide range of needs.
3.	Integration of these rules and executable analytics into decision services. A decision service is a self-contained, callable component with a view of all conditions and actions that must be considered to make an operational business decision.
4.	Use of service-oriented architecture (SOA) to integrate these decision services into existing systems. EDM integrates these decision services into existing operational systems. Doing so requires some sort of SOA.
5.	Establishment of a mind-set and technology framework of adaptive control for continuous improvement and testing. Once decisions can be implemented this way, it is much simpler to monitor the results and adapt them continuously. This concept, called adaptive control, can be realized by various means. One common approach, Champion-Challenger, allows the application to run an alternative rules base for a small sample of cases to test a hypothesis. Because the rules are implemented as abstracted objects, not in code, the solution is straightforward and fast to implement.

Little decisions add up
"Intelligent" systems remain, except in specialized, esoteric fields, something for the future. But the tools and technology available today make your systems "smart enough." EDM works for those decisions that, on an individual level, do not affect the organization's future, but as a group need to be made fast, in volume and fairly accurately. BI is an important component of EDM, both in informing rules formation and in evaluating decision effectiveness. The investment in scalable data warehousing and BI environments will reveal its value in an EDM environment. T

Neil Raden and James Taylor are founders and partners of Smart (Enough) Systems LLC, a consulting and services firm providing analysis and consulting in enterprise decision management to technology providers and businesses. They are also the authors of "Smart (Enough) Systems: How to Deliver Competitive Advantage by Automating Hidden Decisions," Prentice Hall, 2007.

Teradata Magazine-June 2008