October 30, 2008
First Steps To and Beyond Operational Business Intelligence
This article outlines some steps organizations can take as they move from "traditional" business intelligence (BI)
toward operational business intelligence and enterprise decision management.
First, let me recap my July article. Operational business intelligence (BI) has a focus on day-to-day operations and so
requires low-latency or real-time data to be integrated with historical data. It also requires BI systems that are
integrated with operational business processes. However, while operational BI might be part and parcel of operational
processes and systems, the focus is still on changing how people make decisions in an operational context. To compete on
decisions, however, you must recognize that your customers react to the choices made by you, your staff and your systems,
and that you must manage all the decisions you (or your systems) make - even the very small ones. This is the basis for
enterprise decision management or EDM. Five main areas of difference exist between operational BI and EDM - a focus on
decisions (especially operational ones), organizational integration, analytic technology change, adoption of additional
technology and adaptive control.
In this article, I want to outline some steps organizations can take as they move from "traditional" BI towards operational
BI and enterprise decision management. Some of these steps would be a good idea if operational BI was your goal. But
hopefully you are more ambitious than that and want to really begin to compete on decisions.
Start Small
Despite the dreaded 'E' word (enterprise), EDM is ideally suited for starting small and growing over time. It is enterprise
decision management because decisions must be managed as an enterprise asset. Like most new things, starting small is a
good idea when moving to competing on decisions. You need to pick an operational problem (some limitation in a day-to-day
process), but it should not be something super-important. You also want it to be a well understood process, perhaps one
already being managed using a business process management system, and one where there are one or two clear decisions to
make.
Begin With the Decision in Mind
This brings us to the second point - begin with the decision in mind. EDM is completely decision-centric, so it is vital
that you identify and understand the decisions in the process on which you are working. How are the decisions made and who
makes them? Who decides how they should be made and how they are regulated? How long do they take now and what is the
business value of making them faster? Without a good understanding of the decisions that matter to your target operational
process, you won’t get very far.
Understand the Difference between Decision Support and Decision Management
Decision support means delivering the information and insight someone needs to make a better decision - supporting a person
when that person is the right person to make the decision. Decision management means automating the decision and giving
control of how the decision is made to business owners. If it makes sense to your organization that the current decision
makers are the right ones, then you are going to be looking at decision support - something like most people’s definition
of operational BI. If, on the other hand, the decision is being made by someone who really should not make it so much as
pass it on, or if the decision is highly constrained and the person making it is not applying any real judgment or
expertise, then you are looking at automating it and, therefore, at decision management.
Bring People Together
Decision management and operational BI require three groups to work closely together - in some organizations four. You need:
- Businesspeople who understand the decision being made, the regulations concerned, measures and objectives.
- IT people who understand the operational systems through which the information flows and into which decision support
or management must be integrated.
- Information management people who know what’s in the data warehouse, how it is aggregated, which systems feed which
data to whom, etc.
- Analytic people who know how to mine the available data and derive new predictive insight from it. These may be the
same folks who are managing the information, but in many organizations, they are their own group and not necessarily
accustomed to working with information management or IT.
Assess Aggregation
One of the critical issues when moving to operational decision making, both in decision support and in decision management, is that of data aggregation. If your data warehouse or data mart, or your ETL processes, focus on aggregated data, you may well have to reassess them. For instance, if you store daily usage totals for reporting, they won’t be much help when you start trying to improve an intra-day operational process. Aggregation is, by and large, the enemy of both operational BI and of EDM.
Consider Performance and Workload
Operational systems, whether embedding decisions with EDM or adding reporting and dashboards with operational BI, will require you to reassess acceptable performance. Real-time, or near real-time, is likely to become a requirement. Rapid responses to events and low latency in high speed processes will be the order of the day. In addition, the workload for your data infrastructure will become much more mixed and more demanding. This will affect your data warehouse, ETL infrastructure, etc.
Plan for Adaptive Control
The consequences of decisions often take time to play out. To find out what works and what does not, you must constantly
try new approaches, test them and learn what works. Creating an environment in which you can constantly test and learn,
what we call adaptive control or champion/challenger, is going to be critical if you are to focus on improving operational
decision making. You must understand how you can try new challenger strategies on a small subset of your transactions and
compare their results with your champion strategy. This means building some decision-centric performance management
capability and building suitable infrastructure for splitting off some transactions and processing them differently. It also
means managing a change in mind-set away from just one "right" way to many possible ways so that you keep learning.
Evaluate Business Rules Technology
Finally, you need to evaluate business rules and business rules technology, especially business rules management systems. As you move from a mind-set of presenting information to people and hoping they make the right decision to one in which all or part of the decision-making is handled by the system, you need a way to represent the regulations, policies and best practices that should be applied to these decisions. Business rules in a business rules management system are the best way to approach this today. While this is a new technology for BI people, it is important to understand it if you are to move beyond passive operational BI and into more active decision management.
Hopefully you can see that this list overlaps with the kind of recommendations you are hearing for moving to operational BI. Even if operational BI is the next step on your journey, I hope you will aim at EDM so that moving to operational BI is the next step, not the last one.
By James Taylor
Business Intelligence Network
