« Action "0" when building an analytic organization | Main | A growing EDM community »

Is Operational BI an oxymoron?

Amit Chembukar and Prasanna Keny of Tata Consulting wrote Trends in Operational BI in BI Review. It's a nice article and they make some good points about the characteristics of the kinds of systems they discuss:

  • Operational - delivering decisions in operational, transactional systems
  • Low latency - something I have discussed a couple of times before in the context of risk management and decision latency in active data warehouses
  • Granular - I would say transaction-centric as it is about using a single transaction to drive decisions not aggregating a bunch for reporting
  • High availability - obviously these systems cannot go down, even while the way they make decisions is changing constantly

The authors talk about a "BI service at a decision point in a process". Me I call these "decision services" and they are a crucial design element if you are to bring "intelligence" to bear on a process or system. They go on to give three examples and it seemed to me that these are close to what I call Enterprise Decision Management or EDM.

  1. Their first example is a TSA system and my question is why bother with the reports at all given the need to "cater to bookings made between the report generation time and flight departure time" that resulted in "a service that accepted the name of a customer and returned potential matches from the TSA lists." This service could have been used as someone booked to flag them in a more event-based way. Indeed, this story illustrates why traditional BI ideas (reports) are so useless in operational systems. The report was produced because producing reports is what people are used to.
  2. The second example is an EDM example through and through - rules and analytics working together to recommend actions to the field.
  3. The third is another great EDM example - next best action being the primary customer-centric use of EDM (I presented about this at Teradata's conference, for example). Analytics and rules driving recommendations (actions) while more traditional BI technology is used to display useful contextual information that will help the conversation go smoothly. The feedback loop is another common EDM feature and when the "NBA function was packaged and deployed as a Web service" it is clearly a decision service.

All in all a nice article with some good examples of EDM.

I don't think that "Operational BI" is the right name for these kinds of decision-centric applications any more than I did when Keith Gile wrote about it or when I consider "BI 2.0" By and large the BI vendors do not understand this kind of system and all the dashboards and reporting tools they have developed are of little use in building them - there are critical differences between BI and EDM. Even the data you collect in your data warehouse may well be aggregated and processed in a way that makes it less helpful when you are trying to build these kinds of decisioning systems. If you think of these as "BI" systems, even "operational BI" systems you will bring the wrong perspective to bear. Think of them, instead, as decisioning systems.

Technorati Tags: , , , , , , , , , , , ,

First time on the EDM blog?
Subscribe to the EDM blog feed or check out some other recent posts:


TrackBack URL for this entry:

Listed below are links to weblogs that reference Is Operational BI an oxymoron?:


Neil Raden

James, another thing to consider is that almost all BI tools are designed to be interactive. It is very difficult to take a piece of software with internals that are designed for a GUI and convert it to run in unattended mode. In addition, a lot of the intelligence of the tools is buried in the UI, not the engine.

BI place emphasis on ease of use and presentation/navigation, two things that are not really important in operational BI.

Even more importantly, the leading BI generally lack the capability to dispatch an action, in other words, to close the loop. They may be able to run scheduled jobs in the background and push out a report as an exception or an alert, but the process is pretty clunky and does not scale to the volumes that a true operational BI system would require.

And one last thing - performance. BI gets its instantaneous response time only through the laborious staging of data into aggregates, cubes and caches. A direct-to-datawarehouse BI tool using SQL could not provide the throughput needed for large-scale decision services.

Is it an oxymoron? I'd have to say yes.

The comments to this entry are closed.

Search Site



  • enter your email

Upcoming Events

  • FICO Tools & Analytics User Forum 2012
    BERLIN: September 11-12, 2012 LONDON: September 18-19, 2012 Gain new insights for improving business performance through advanced analytics and decision management tools.