In their pursuit of lean manufacturing and demand-driven strategies, manufacturers are increasingly interested not just in improving their real-time view of manufacturing data -- such as job status, work in process, shop floor inventory, quality data and capacity utilization -- but also putting it in context with business data such as customer orders, product specifications and cost information.
Call it business intelligence, manufacturing intelligence or supply chain intelligence; it all boils down to a need to answer strategic business questions that require the juxtaposition of data from enterprise resource planning (ERP) systems with plant floor and manufacturing execution systems (MES).
"Manufacturers need to use a lot of real-time data but in a broader business context, such as relating plant performance to schedule adherence or quality data to production orders," notes Colin Masson, an analyst at AMR Research in Boston.
And whereas typical business intelligence systems pull data from a transactional system into a data warehouse to analyze historical trends, manufacturers need to view data as the event is occurring and respond before it is too late.
"If you're trying to use manufacturing data while you can still correct the quality problem or adjust the length of the production run, you have to look at that data in something closer to real-time," Masson explains. "The emphasis in manufacturing intelligence is to detect and correct things before you end up disappointing the customer."
With that in mind, here are some issues that manufacturers face in their efforts to build better intelligence into their processes (click here for additional online resources).
Don't Get Caught up in Semantics
The software products that help gather and analyze the required data for manufacturing intelligence don't fit neatly into a single category. Traditional business intelligence systems hail from the likes of Oracle Corp. with Application Server 10g, SAP America with xMII, SAS Institute Inc. with SAS Enterprise BI Server and Cognos Inc. with Cognos 8. But what manufacturers need is a layer of software that pulls data from ERP systems with data from MES and plant floor systems and gives it meaningful context, says Ken Brant, research director for manufacturing at Gartner.
That layer can come from the ERP vendor, the MES vendor or from a pure-play manufacturing intelligence provider, he explains. Examples of the latter include Activplant Corp., Informance International, and Incuity Software Inc., all of which enable multi-site performance analysis.
But manufacturers shouldn't waste time trying to get a handle on software categorizations and terminologies, Brant advises. Instead, they should spend time developing a clear strategic objective. "Don't pay attention to terms [the vendors are] using," he says. "Think about the kind of intelligence you want to act on and the kind of presentation of data you need and then decide which company has the right approach for your objectives."
For instance, if you've just gone through an Oracle or SAP implementation, it would make sense to build out from there. A lean transformation might require you to look at a different set of vendors.
Watch your Connections
To collect all the data you need, your manufacturing intelligence system may need to connect with many different types of systems, including device drivers, historians, SCADA systems, MES and ERP systems, and quality management systems, Masson points out. And only some of the available systems enable multi-site MES connectivity, he warns.
"You need strong device and proprietary systems connectivity, which makes it a lot more challenging than just using a library of ERP or database connections as you do in the traditional business intelligence market," he advises.
Unravel the Data Model Morass
Another challenge for the IT department is the variety of data models they'll likely encounter, particularly because most companies haven't standardized on a single MES. This is quite different from the more familiar world of business intelligence.
"Every manufacturing line has a uniquely defined data model, whereas with business intelligence, you have the luxury of leveraging a standard data model provided by the ERP vendor," Masson says. It may be necessary for IT to build a standardized data model, a job for which many IT departments lack the necessary skills.
Start with ERP ...
In some cases, you may be able to integrate your manufacturing intelligence system just with the ERP system and skip the MES integration, Masson says. This is possible with some of the higher end systems from mid-market ERP manufacturers such as Infor, Lawson Software/Intentia, IFS, QAD Inc. and Microsoft Corp. with Dynamics.
All of these companies offer strong manufacturing capabilities, as well as business and manufacturing intelligence modules, he says. "In many cases you can integrate these solutions with the shop floor and don't need another MES in between," he says. "They're built so you can get updates to the business intelligence solution on a real-time-enough basis to manage."
At the very high end, SAP's acquisition of Lighthammer Software Development Corp., brought the global ERP player a manufacturing intelligence tool that can connect with many different control systems and also integrates with SAP enterprise applications.
... Or build from MES
Meanwhile, MES vendors have also added manufacturing intelligence modules to their systems. Examples include The Invensys Group, with InFusion; Siemens, with its Intelligence Suite; Apriso Corp. with FlexNet; Pelion Systems Inc.; and Factory Logic with Pacemaker and Synchronizer. These systems perform multi-site performance analysis, integrate with enterprise systems and provide needed context for decision-making.
Work with a Sense of History
When architecting your system, you also need to consider whether you want to maintain an operational data store. Some systems, such as Lighthammer, maintain connections to disparate data sources, perform their calculations and display key performance indicator (KPI) statistics without storing data anywhere.
"It has the benefit of being quick to implement, but if any of those systems go down, the whole system is impacted, and there's no long-term history that can be archived away," Masson says. Other systems, such a product offered by Informance, maintain an operational data store, provide mechanisms for managing data models and archive data in a long-term archive, he says.
Make it Event-Driven
To maximize the timeliness with which you can act on the data provided by the manufacturing intelligence system, Gartner's Brant says the system should use a message-oriented or event-driven architecture.
"You want to get the data as a notification with the context built in," he says. "You're looking for particular data points that trigger a particular process or a decision to be made." This requires identifying several different scenarios and particular events the company wants to be made aware of and enabling the system to send alerts to a desktop or handheld computer while the event is still actionable.
Brant cites an example of a critical machine outage, which would adversely affect order management and on-time delivery. If the system detects that only 50 units were produced against an order of 100, it should send a message to the account executive, the production scheduler and the advance ship notice system. "It needs to be discriminating, and you need a decision-making hierarchy," Brant points out.