The hype surrounding so-called real-time analytics has prompted many organizations to
consider upgrading their information supply chains and analytic capabilities. However, the road to
real-time is not so simple.
Most enterprises stop short with the development of high-latency analytic solutions that direct integrated information at enterprise eyeballs, rather than business processes. Yet executives throughout the business community speak readily about their need for real-time analytic solutions that optimize business process performance. (See
It's high time to define real-time
IT professionals throw around the term "real-time" to indicate how quickly new data is made available. Instead of absolute definitions (e.g., "sub-second"), we recommend that enterprises adopt a more meaningful definition of "real-time" as a relative degree of latency that ensures the availability of fresh data representing a current business state. Similarly "near real-time" should refer to a degree of latency that ensures availability of data "fresh enough" for the individual or process using it. (See Figure 1.)
The key is to define "real-time" relative to a business process, not a clock. Case-in-point: for information needed to close monthly financial records, real-time may mean, "since our last transaction last month." This may be days ago. Conversely, for Web personalization it may mean, "before the visitor clicks anything else." This clearly is sub-second.
Setting BPM objectives
Even before tending to information supply chain upgrades, objectives for business performance management (BPM) must be established. The approach is straightforward:
- Identifying business process, particularly those that can be better managed via analytics
- Determining performance measures that assure harmony across all business processes
- Selecting mechanisms to consume the analytics and configure them to regulate associated business processes
Second, all performance measures should relate to core enterprise goals to ensure they are well coordinated. Conflicting goal-oriented analytics can be avoided by focusing them on:
- Efficiency or effectiveness
- Speed or precision
- Quantity or quality
- Risk reduction or innovation
- Savings or returns
- Derivative indicators or direct (usually financial) indicators
- Top line or bottom line
- Balanced or focused
- Short-term or long term
- Internal or external (market) conditions
Third, and often most difficult, is to ascertain how, where and when business performance analytics will be assimilated into strategic decision-making. Most business applications are not designed to leverage analytic output, and most individuals still prefer not to be told how to do their job by a computer. Some business functions may not be good candidates for BPM because training or application modifications may be deemed too prohibitive.
Interesting versus important information
Another analytic trap organizations fall into is relying exclusively on user requirements to fashion analytic initiatives. Business professionals are notorious for either not knowing what information could help them do their job better -- or worse, not being able to apply analytic output. We have seen multi-million dollar data warehouses be vastly underutilized to the extent that they are eventually shut down. DW managers have resorted to inserting outlandish dummy data into reports to see who is really leveraging them. DW project teams often fail to employ common sense to understand candidate business processes and how they might be impacted through analytics.
Tightening information supply chain links
Delivering real-time applied analytic solutions demands addressing the infrastructure components that comprise the information supply chain (ISC). Enterprises must appreciate that their real-time capabilities are no better than the least efficient component of their overall ISC. For example, if their data quality mechanism operates in batch, either suspect data may be delivered in real-time, or clean data may be delivered periodically. Or, if their data integration tool has no message queue listener or DBMS trigger insertion capabilities, then latency is limited to how often the tool takes to instantiate and to identify new data. Even if DBMS inserts or updates must succumb to an indexing or referential integrity process, this introduces a discernable degree of latency.
Bottom line: While real-time analytic aspirations are a noble objective and can provide operational optimization, enterprises must be pragmatic about determining how and where to apply such solutions. Business over-steering and information "engine flooding" are frequent unwarranted outcomes. Moreover, real-time analytic solutions can incur extreme costs that may outweigh their benefits.
Mr. Laney is an experienced practitioner and authority on business performance management solutions, information supply chain architecture, decision support system project methodology, consulting practice management, and data warehouse development tools. Prior to joining Meta Group in February 1999, he held positions with Prism Solutions as a consulting practice director for its Central US and Asia Pacific regions, as a methodology product manager, and as a consultant to clients in Latin America.
This was first published in September 2002