Active ingredient
- Reduce text size Decrease text size
- Increase text size Increase text size
- Print article Print
- Jump to comments Comment
- Share this article Share
- Email article to a friend Email
Over the past two years, vendors have been augmenting their BPM software with so-called BAM technology, trumpeting that this emerging technology is not merely another segment of the analytics market but the key to creating what some call the 'actively-managed enterprise'.
Real-time analytics has been the Holy Grail of business software for over a decade. The quest is for a technology that pushes the capabilities of business intelligence (BI) beyond its traditional 'rear-view mirror' model of analysing data in retrospect to one where the business can analyse 'on the fly' and react to events as they occur.
Unlike traditional BI, where the emphasis is on extracting data from operational systems hours or even days after transactions have been completed, the key to a more reactive management system is to sew the analytics directly into the business processes themselves. Although BI vendors are moving in that direction, the different analytical touchpoint of processes means, arguably, that the technology companies most capable of delivering on such a vision are not from the world of BI but are the suppliers which provide the platforms that control the process flow - companies such as IBM, webMethods, BEA and SeeBeyond.
| ||
Over the past two years, these vendors have been augmenting their business process management (BPM) software with so-called business activity monitoring (BAM) technology, trumpeting that this emerging technology is not merely another segment of the analytics market but the key to creating what some call the 'actively-managed enterprise'.
Active view
Whether a major and separate market for BAM software will develop is still up for debate, but few observers doubt it will play a significant role in the next decade of the software industry.
BAM is conceptually relatively straightforward. It uses the visibility into business processes provided by BPM to enhance the performance of operations; it directly measures the efficiency of processes so modifications can be made accordingly; it manages interactions with those processes and addresses exceptions; and it delivers appropriate information to individuals, groups or systems at different points in the business process chain so threats and opportunities can be addressed.
According to Graham Glass, CTO of webMethods, that makes BAM a natural extension of the process management fabric. "We want to leverage the infrastructure to provide visibility and control of real-time decision making," he says.
That involves tracking business process performance through the use of data collection agents to gather and analyse events within processes - to identify, collect, filter, correlate and provide baseline metrics that comprise key performance indicators. The goal is to spot exceptions to the rule. As Ross Altman, CIO of integration software vendor SeeBeyond, puts it: "BAM tells you what processes are out of range."
At the heart of all BAM technologies is an event-processing engine. It monitors processes continuously for pre-defined events - a sudden slowdown in order processing in the call centre or a failure of the product shipment department to meet its hourly targets. The engine combines events data and compares that actual performance to a baseline of previous behaviour in order to decide if it needs to flag up an abnormal trend or activity. The engine then sends an alert to a relevant group or triggers an automated process designed to remedy or exploit the situation.
To those in manufacturing, such structures may seem old hat. Robotics and other machines on the production line have for years included programming to trigger an alert if they go near particular thresholds. But there are a couple of critical differences with BAM systems, says Bill Gassman, an analyst at technology advisor, Gartner. The focus on events in BAM is process-relate. Moreover, the events originate from multiple systems, not just one.
There is echo of BAM in systems management event-tracking and resolution too, where system monitoring agents trigger alerts on issues such as critically low levels of disk space, a failed back-up or a sharp fall in transaction throughput. That tells an administrator what system to fix; BAM tells a manager what process are out of kilter.
BAM's fundamental architecture is not complicated: messages containing a description of an event are collated and the result tested against a set of thresholds that, when reached, trigger an alert or a systems action. Gartner splits that structure into three stages: absorption, processing and delivery. The 'event absorption layer' is responsible for collecting events, filtering out irrelevant ones and transforming events into a usable format for the event-processing layer.
This is where the real technical challenges lie for BAM. "Because a BAM system works in real time, it may need to make sense of thousands of events per second," says Gassman. The trick is to be able to assess their importance individually and in combination with each other against a string of pre-ordained rules in order to determine if an alert should be raised - and all in real time.
Once the alert is raised, the task of delivery is relatively simple - at least for the time being. BAM tools provide numerous options for delivering the message to individuals via dashboards, portals, SMS, email, instant messages, pagers, taskbar icons and other instant feeds. As BAM becomes more sophisticated, however, alerts will be delivered directly to applications, triggering a reaction.
Reacting to events in real time is all very well, but a more desirable outcome is for the system that catalogues the patterns of these events to predict problems before they cause serious disruptions. By using statistical analysis techniques, a BAM system can not just observe when an exception occurs but it can learn when an exception situation is approaching.
Some of the pioneers in this area have taken that one step further by developing 'fingerprints' - unique identifying patterns associated with the behaviour of previous events - which automatically cause key performance indicators to go out of range.
Given it is a technology that runs so close to live processes, many early users of BAM are reluctant to describe their experiences - after all, how many organisations want to highlight serious 'exceptions' in their operational processes? But a few have seen enough benefit to be convinced to make a case for BAM deployment.
One example is Motorola. Among other areas in its supply chain, the communications systems giant has implemented BAM to optimise its 'order entry to order book' process for direct customer sales. It set out to use BAM to address several challenges: to decrease the discovery and resolution time of problems it was having with order processing; to reduce the amount of manual monitoring involved; to raise the number of orders fulfilled online versus through call centres; and to provide early intervention and warning capabilities.
Using BAM technology from webMethods, the company says it experienced a reduced problem discovery and resolution time of 85%; a 75% drop in the number of 'stuck' orders - something that will produce a $3 million ROI annually as a result of the decrease in cancelled orders; and a 15% increase in the number of orders fulfilled online.
At this stage, many such BAM implementations are fairly piecemeal. But the wider picture suggests that BAM is an enabler of the real-time enterprise, alongside service-oriented architectures and BPM.
