IBM Decision Management Summary
IBM is one of only a handful of suppliers capable of offering an enterprise decision management solution. The term ‘decision management’ embraces the notion that operational business decisions can be automated and managed for greater efficiency and accuracy. Decisions permeate the operational environment in all businesses, including credit limits, supplier approval, best up-sell offers – and so on.
Decision management rests on three pillars – predictive analytics, business rules management and business optimization. Dig down into the obscuring layers of IBM terminology and you will find that this is precisely what is on offer. The term ‘analytical decision management’ is used for the technologies and methods which support the creation of decision models, and ‘operational decision management’ is used to describe the execution environment of such models. It’s an unnecessary distinction in many ways – since it’s all decision management.
IBM’s decision management capability is to some extent disjoint. Three separate products comprise the essence of the offering – SPSS, ILOG Optimization products and the ILOG JRules business rules management system. Please bear these three elements in mind as we explore the layers of obscuration that IBM has created on top of them to make the whole thing seem more business oriented (with the side effect that it makes it all seem more confusing).
To a large extent IBM is a services and infrastructure business, and its Decision Management offerings will be primarily of interest to businesses with an existing large investment in IBM technologies and services.
Here is a brief summary of IBM’s offerings:
IBM Operational Decision Manager includes IBM Decision Center and IBM Decision Server. The latter is essentially a run-time environment – a business rules engine in the main. IBM Decision Center provides a repository for business rules and the tools to create, modify and manage them.
IBM SPSS Decision Management software supports the creation of predictive models, which can be used to fully or partially automate business decisions. IBM tries very hard to convince that SPSS can be used by business users to create models, with its ‘three-click’ message. It has obvious appeal, but in many instances is simply not realistic, and potentially quite dangerous.
IBM ILOG CPLEX Optimizer is the core element in the IBM Decision Optimization Center. This concerns itself with the optimization of business resources given a set of constraints and objectives. An Eclipse based IDE is provided so that optimized models can be incorporated into production applications.
While these three components provide the ‘bare bones’ capability, there are issues such as deployment, model management and monitoring, integration with other applications and so on, that need to be addressed. Certainly IBM addresses some of these issues, and particularly with its Service Oriented Architecture (SOA) bias, but even so some considerable effort is needed to tie the whole thing together, and in this respect it is inferior to some other offerings.
IBM SPSS Decision Management
IBM SPSS is a large and comprehensive analytics toolkit. The Decision Management component simply extracts those elements that are most useful in building predictive models. It comes with two out-of-the-box applications. IBM SPSS Decision Management for Customer Interactions is targeted at front line business managers who need to manipulate business rules and models in order to obtain the most profitable outcomes from promotional campaigns and other customer contacts. IBM SPSS Decision Management for Claims – Empowers claims processing staff to run simulations that give them greater control when balancing efficiency gains against possible risks of fraud.
Decision Management 6 includes all aspects of automated decision design and deployment that an organization needs.
More generally IBM SPSS allows users to use more advanced analytical techniques to discover patterns in data. Simulation capabilities are also available so that different models and approaches can be evaluated. Collaboration features are also included so that various team members can work on a given model.
IBM Operational Decision Manager
This is effectively the environment to create, modify, manage and execute business rules. The rules can be created by expert business users, typically in the form of IF-THEN-ELSE type constructs, or via the implementation of decision trees or tables. Not all models created in SPSS will be suitable for deployment in the Operational Decision Manager environment, and particularly those that create opaque models such as neural networks.
IBM Decision Server is the run-time component and delivers a high performance environment that provides decision services to other applications. An Eclipse based development environment is provided along with event monitoring tools.
IBM Decision Center provides a non-technical rule language that supports business users and others that need to create rules through a user friendly interface. Social media style collaboration is supported and a decision governance framework which includes role based security, release management control, historical reporting and simulation capabilities. The centralized decision management repository provides a single unambiguous definition of business rules.
It should also be added that IBM is shifting its decision management capability to the cloud, and specifically Bluemix. IBM Business Rules for Bluemix simply requires the installation of the Rule designer onto the Bluemix platform.
IBM Decision Optimization Center
This embraces the ILOG CPLEX optimizer and provides an integrated environment for the creation, management and deployment of optimization models. These models find wide application in business, from deciding how to deploy sales resources, through to scheduling aircraft so that refueling costs are minimized. In decision management, optimization is often used to optimize the use of business rules based on current business conditions.
The IBM ILOG CPLEX Optimizer supports a variety of optimization styles, including linear, nonlinear and mixed integer models. The architecture supports distributed parallel algorithms for very fast processing.
These various IBM platforms make extensive use of the Eclipse Integrated Development environment. While this is an excellent platform, it does mean that developers have to work fairly hard to tie the various components in a decision model together. For organizations already heavily committed to IBM this may come as second nature, but for others it may all seem a bit daunting.
IBM Decision Management is primarily challenged by FICO and SAS. While it does provide excellent tools for model creation, optimization and model execution, there is less emphasis given to deployment and implementation. Of course these needs are met, but there is no well defined model management capability – although model management is accommodated. In summary IBM provides the necessary tools, but is weaker on the management and monitoring of decision models.