Tech2Tech
Insider's Warehouse
Prepare for impact
Automate for simpler, smarter and more agile decision making.
by James Taylor and Neil Raden
Making the right decisions at every level of an enterprise drives a business to success. Yet while organizations and executives focus on high-impact, strategic decisions, those made by operational and front-line personnel are often neglected as they appear to lack impact. This is a mistake because the seemingly little decisions—which add up—ultimately define a company’s brand identity, as in these cases:
- Customers consider a product either expensive or a good value based on the price offered to them.
- A customer service representative’s cross-selling techniques determine whether the customer feels valued or over-sold.
- The options offered at a Web site, kiosk or ATM suggest to a customer whether that company is easy or hard to do business with.
While each of these operational decisions makes a small impact on the organization, the cumulative effect is huge because they occur so often
Enterprise decision management
Enterprise decision management (EDM) focuses on the operational decisions that create value in your business as reusable assets that drive results. The EDM approach makes these assets widely available via decision services. This is a method in which decision points in processes and systems are made simpler, smarter and more agile through automation. Additionally, the service constantly monitors and improves the way operational decisions are made to keep up with regulations or updated business processes.
EDM is a management discipline, not a technology stack. But to change the way organizations automate and manage their day-to-day operations, the approach relies on technology and its effective deployment. It necessitates additional investments in business intelligence (BI) systems and data warehouses to capture and understand all of a company’s data, and it ensures that repeatedly proven analytic and business-rules technology are integrated into mainstream development approaches.
A range of robust, cost-effective and well-understood technology, available today, is required for EDM to effectively manage operational decisions. Some fundamental technology infrastructure requirements and reference architectures must be used to ensure a successful adoption of EDM.
Infrastructure requirements
The key infrastructure for EDM falls into three broad categories: data infrastructure, data analysis infrastructure and decision management infrastructure.
Data infrastructure
The data layer must be able to deliver accurate, up-to-date and fine-grained information for use in managing operational decisions. This data might be used offline to develop analytic models, to mine for business rules or to understand what has happened in the past. It might be used inline to drive a specific operational decision, support a manual decision-making process, or show the state and results of decision making.
Data analysis infrastructure
EDM requires data to be analyzed and delivered as machine-readable artifacts. Taking the form of data mining outputs or predictive analytic models, these artifacts allow operational systems to use insight that can be gained from historical data much as visualization helps business users.
Decision management infrastructure
The most widely deployed decision management infrastructures are:
- Decision management applications. Focus on the automation and improvement of a specific class of decisions
- Business-rules management systems. Can also be used to manage decisions in which analytics are not important, such as compliance, or by including executable analytic models as “black boxes”
- Decision management platforms. Make it possible to manage a variety of models and rules for many different decisions; these platforms are becoming increasingly popular
Service orientation
Broad adoption of EDM requires that this technology supports a service-oriented or component-based approach. Suitable technologies that allow access from service-oriented applications, or are merely compatible with this type of approach, have broader applicability to EDM than those that do not. The ability to package rules, optimization models and analytics into coherent and self-contained decision services is important; ensuring that all of the necessary data, integration, visualization and reporting are available is critical.
EDM solutions require multiple technologies to be used in combination, and they must be integrated with systems of record and transactional environments. Using a service-oriented approach allows decision services to be managed as reusable assets that are easily integrated with other systems and helps ensure that ongoing changes in the decision-making logic are isolated to a single coherent component.
An EDM architecture
When considering technology for operational decision making, it is useful to have a reference architecture or framework that covers how decisions are designed, put into operation, and deployed or executed.
At design time, a central decision repository is updated with the definitions, rules and models of how decisions are to be made. The repository might be physical or logical, and it is implemented by several linked repositories, each supporting a different aspect of development. Decisions are identified and modeled in the repository, where they are linked to key performance indicators (KPIs), business processes and other aspects. The rules and models needed for a decision are based on policies and regulations, which can be verified, validated and simulated using historical data. Decision designs that are useful and complete can then be deployed.
EDM has much to offer companies with high-volume, operational decisions that matter
to their business.
As an automated decision is designed, requirements such as throughput, data and regulations should be considered. Service level goals and business objectives are used to make design trade-offs and technology choices. Depending on the decision design, successful implementation may require analytic insight or business rules—or both.
Before the analytic models or business rules can be mined, however, the data must be understood. What is it telling us? What insights might we gain? Is all of the data needed available, accurate and reliable? Rules mined from data and legacy code, along with those modeled explicitly, are integrated and combined with analytic models to define the logic needed to make the best decision.
Finally, the automated decisions must be verified and validated to ensure all circumstances are considered and that nothing contradictory is being proposed. They should then be simulated to confirm that the outcome is what is expected and desired. Simulation and verification, along with modification, will likely take multiple iterations to achieve the optimum decision function design.
Once these decisions are deployed, they must be operational in a typically complex IT environment. As such, they must support event-driven architectures and enterprise services buses. Additionally, to ensure the decisions are consistently made, it is critical that they can integrate with legacy applications. This puts decisions, and the decision services that implement them, at the heart of the IT architecture.
During the course of their use, decision services will need to be modified. First, the effectiveness of existing decisions needs to be assessed. This typically involves applying corporate performance management or BI tools to analyze business-results log files and map those results to KPIs. If the effectiveness of a decision is satisfactory, then little, if any, change will be required. If the effectiveness is not satisfactory, then new approaches must be devised and tested, new rules written, and new analytic models developed. Of course, any external regulatory change could require an update to the rules involved in a decision.
Analytic competitor
EDM involves an unrelenting focus on the decisions that create value in a business, especially the operational decisions that drive front-line processes and systems. An automated decision services tool is created by extracting and consolidating decision-making functions out of applications consistent with current service-oriented architecture (SOA) design principles. Building on the basics of a data infrastructure—including the data warehouse, operational data and unstructured data—EDM uses business rules management systems to manage these decisions and adds data mining and predictive analytics to enhance them. Because the definition of a good decision changes constantly, EDM includes a feedback adaptive control loop to discover the best ways to improve each one.
EDM has much to offer companies with high-volume, operational decisions that matter to their business. Companies focused on improving their operations and compliance methods, competing with analytics, and enhancing their strategic agility would do well to adopt EDM as an approach. These strategic organizations will become increasingly decision-centric, focusing explicitly on the effectiveness of their decisions rather than on aggregated historical metrics.
The recognition of decision making as a competency and the allocation of resources to elevate, understand and continuously improve decisions will make these organizations true analytic competitors.
James Taylor and Neil Raden are co-authors of “Smart (Enough) Systems,” the definitive text on enterprise decision management. James Taylor is CEO of Decision Management Solutions (www.decisionmanagementsolutions.com), and Neil Raden is the founder of Hired Brains (www.hiredbrains.com), offering services in BI and decision sciences.
Photography by iStockphoto
