Intelligent Enterprise

It may not be deployed in your manufacturing plants or warehouses as yet, but radio frequency identification (RFID) technology is fast becoming pervasive in the plans of many companies and industries. It's not difficult to comprehend the potential benefits; RFID can provide accurate, real-time data about the location of goods and the state of manufacturing, physical assets and logistics. The challenge lies in integrating the technology in a well-thought-out way and, most importantly, tapping into the data for analytics and business intelligence.

Mandates from the likes of Wal-Mart, Target and the Department of Defense have spurred RFID deployments, yet many early adopters have too little to show for their investment in tags, readers, middleware, application software and integration services. In their rush to meet mandates, many companies spent vast sums on cosmetic "slap-and-ship" RFID deployments. These firms are still waiting for returns on their investments, and they're likely to view RFID as a cost of doing business. Meanwhile, a few pioneers have treated the mandates as an opportunity to achieve deeper levels of RFID integration, and they're gaining process efficiencies and making better business decisions as a result.

This article explores many of the technologies, challenges, opportunities and trends in RFID, focusing particularly on data management, process integration and business intelligence. Along the way we'll share examples of successful deployments that have gone beyond the quick-and-dirty approach.

Start at the Edge

Deploying RFID is inherently complex due to its multidimensional nature. In one dimension there are the various disciplines and processes in which RFID can be exploited, including manufacturing efficiency, asset management and logistics execution. Then there are industry-specific demands, as in consumer packaged goods, pharmaceutical and defense mandates. Technology requirements include integration with ERP and legacy systems as well as Electronic Product Code (EPC), Uniform Code Council (UCCNet) and Web services standards. Thinking through all these process issues, mandates and technology questions is hard enough, but then there are also more fundamental, practical challenges implementing the RFID technology itself.

It all starts with the RFID tags, which are passive transponders that are either preprogrammed or "commissioned" by printers/encoders. The tags typically encode 96 bits of data, including company identifiers, global electronic product codes (EPCs), serial numbers and other advanced features such as electronic locks and brand protection. While the goods are being picked, packed, loaded and shipped at the supplier, transited and received at the customer location, RFID interrogators (readers) use radio frequency energy to query the tags and collect the information encoded as well as the reader location and time.

In some cases, just getting optimal read rates from the edge devices (such as readers) is akin to a root canal due the physics of RF technology, tag placement and difficult materials (including liquids, metal goods and foil packaging). There are other confounding factors such as RF interference, but much has been learned in recent years, and "it's possible to achieve 100% read rates using multiple readers, locations and application logic," says Justin Hotard of Symbol Technologies, a supplier of RFID devices. These technical difficulties have also been addressed by EPCGlobal, the consortium that sets global RFID standards, which has ratified a "Gen 2" tag standard that supports dense reader modes (multiple readers in close proximity), faster read rates (of 500 to 1,500 tags per second versus 100 per second for current tags), improved security and selective reading, so, for example, you could issue commands to read only products that have expired.

Gen 2 tags aren't being produced in volume as yet, but manufacturers are supporting the standard and some, including Intermec, SAMSys Technologies and Zebra Technologies, are providing firmware upgrades for existing readers and printers (although the upgrades may not support all the new features of the Gen 2 protocol).

Manage the Data

Once you've nailed the basic data collection problems, the more involved device control and data management challenges emerge. Device control software is the layer that integrates with edge devices. The software gathers data, converts it to standard formats, such as XML, and pushes it or posts it to applications and enterprise systems.

Some view the middleware device controller software that stands between devices and applications as expensive and difficult to deploy. Thus, SAP and Intel have announced a partnership whereby Intel will incorporate features in RFID hardware that will enable direct communication with SAP's RFID software applications in plug-and-play fashion.

In a midsize to large RFID operation with scores or hundreds of edge devices, the device control and data management demands can get quite sophisticated. Vendors such as Oat Systems and Infineon cover these bases and even get into data analytics and trend analysis. Oat's Foundation Suite, for example, includes a data management system with built-in adapters for automated import of data from trading partners, a consistency engine for cleaning data and analytics and reporting that let you track and trace, report on exceptions and drill down into velocity or inventory changes by product, location or trading partner.

RFID event streams churn out a veritable flood of data compared to the trickle of information in barcode environments, but data volumes vary by business scenario and depth of deployment. The data might be managed in real time, post-processing or both, depending on the requirements of the scenario, and the sheer volume often requires RFID-specialized technologies.

German forest management company Cambium-Forstbetriebe is using RFID to track timber. In this high-volume scenario, nail-shaped tags are pounded into each log once they're cut, and they're scanned at several points along the supply chain to ensure accurate accounting of logs harvested and delivered to saw mills. Cambium-Forstbetriebe harvests some 70,000 cubic meters of timber annually, and the application is generating a massive amount of data that needs to be converted to XML, filtered and directed to numerous back-end systems including ERP.

"Relational databases just could not meet the [data volume] requirements, so we turned to in-memory databases," says Dr. Michael Gross of Dabac, a software and integration firm on the project. Queries are usually handled by database servers that are separate from, and often physically remote from, the application, but this approach requires that data be transformed from relational to application data structures. Dabac chose an RFID engine from Progress Software's Real Time division that uses locally accessible, in-memory caches to speed data requests that would otherwise restrict performance.

Cambium-Forstbetriebe spent approximately $600,000 on the project, and it expects a return on investment in just one year thanks to reduced shrinkage and improved payment reconciliation. Previously, the company lost track of 10% to 15% of logs somewhere between the forests and the saw mills; it expects to reduce this figure by as much as 70%. Reconciliation, meanwhile, was a laborious, manual process that averaged three months. Now the company has a highly accurate, granular record of what was harvested, what was shipped and what was delivered to each mill. The time and cost associated with exception handling are expected to be reduced by 50%.

As part of its effort to comply with Wal-Mart mandates, Pacific Cycles has integrated RFID data with a back-end ERP application. The outbound logistics scenario generates a relatively small volume of data just 10,000 RFID events per month--but the effort has already improved inventory control and warehouse management (see the "Field Report").