| |||||||||||||||||||||||||||||
 
As the time required to make business decisions decreases, BI solutions are evolving from historical analysis to real-time event-driven tools. 
For manufacturers, the notion of the real-time enterprise is somewhat old hat. Since the Japanese made "just-in-time" a watchword in the 1980s, manufacturers have been motivated by the need to drive production not by plan, but by actual demand. In many cases, that need has dictated practices that combine the benefits of long-range enterprise resource planning (ERP) with the contingencies of choreographing deliveries of fresh inventory, sometimes by the hour. Transaction systems, such as ERP and customer relationship management (CRM), can provide excellent snapshots on business activity, such as status reports on inventory levels or order backlogs. However, business intelligence (BI) provides the ability for businesses to look back, examining how well the organization has worked to achieve such goals as meeting customers' expectations for price, quality, or service. Not surprisingly, in an era of thinning margins, the notion of faster analytics should appeal to manufacturers. However, until now, notes IDC analyst Henry Morris, BI has typically focused on historical rather than real-time information. "A lot of BI is still conducted on a scheduled basis," he says. Looking at Today, Not Yesterday Consequently, manufacturers who are analyzing the timeliness of incoming or outgoing deliveries, the quality of a particular supplier, or the profitability of a production line are more likely to work from information fed to data warehouses in batch loads conducted during off hours. From a practical standpoint, that means they are working from information that may be a few hours or even a day, week, or month old. Nonetheless, many of the back-end data transformation tools used for loading data warehouses have recently added the ability to trickle in data from production systems on an ongoing basis to make data warehouses more current. A new variant of this approach might bypass the data warehouse altogether. "Business intelligence should be all-pervasive," explains consultant and database veteran Jnan Dash, formerly a group vice president for advanced product development at Oracle. He suggests that a publish-and-subscribe architecture (also known as "PubSub") similar to what Sun has developed for its internal systems could provide the mechanism for producing real-time analytics on an event-driven basis. "Business activity monitoring (BAM) systems are pointing to the fact that event-driven technology can be useful," he says. According to Dash, the key to making such an architecture work is to have "lightweight, zero-install browsers that filter events and notify (users) in real time." For manufacturers, an event-driven architecture that provides event-based notifications based on preset thresholds should sound familiar. Veterans of process control technology are used to working with process control and process historian systems that flag events in manufacturing processes before they careen out of control. In effect, real-time business intelligence would take the same threshold-based logic and apply it to manufacturing or supply chain business processes. For instance, if an account manager were alerted whenever a customer shipment was delayed more than a day, it could prompt an analysis of the company's service history to that customer. According to Dash, while the idea for dispatching real-time alerts isn't new, existing technology has not been suitable for sophisticated analytics. "We used to do a lot of this with database triggers," he says. "But those triggers were so heavyweight that they were limited to low-level events." At this point, solutions for delivering BI on a real-time, event-driven basis are just beginning to emerge. "The line between transactional systems and analytics are blurring," says Dash.
|
 
 | ||||||||||||||||||||||||||||
|
| |||||||||||||||||||||||||||||
