In an airport at Frankfurt, Germany, a maintenance worker crawls through the cramped ventilation system, wearing a small device that reads data from chips positioned throughout the system to verify that he's done his job. At a U.S. theme park, a teenager on a waterslide wears a special wristband with an embedded chip that will remember how many rides he has taken, no matter how wet he gets. And at a large drug company's warehouse, a shipment of the regulated narcotic OxyContin gets tagged with chips to track its route, and discourage theft and counterfeiting.

What do all these chips have in common? Each contains special RFID (radio frequency identification) technology, which enables a device to read data stored on the chips at a distance, without line-of-sight scanning or physical contact.

After years of use in applications such as vehicle tracking, factory automation, and anti-theft systems, RFID is suddenly poised to break into the same league as the transistor and the microprocessor. The cost of RFID tags has dropped from several dollars to 30 cents per tag or less, making RFID increasingly cost-effective for much higher volume applications, especially in the supply chain. Driven by recent RFID mandates from several large buyers including Wal-Mart Stores, Target, and the U.S. Department of Defense, much of the corporate world is scrambling to test and roll out pilots of RFID technology.

"There's a lot of hype around RFID and maybe a little bit of herd mentality," says Jon Brendsel, VeriSign's director of electronic product code network services. "Lots of end-users don't know why they need to do it, but they know they need to do it."

The list of potential RFID benefits is seemingly endless: greater visibility and product velocity across the supply chain, better inventory management, automatic replenishment, reduced invoice reconciliation and labor costs on the receiving dock, easier product tracing and recalls, and reduced product tampering, theft, and counterfeiting. But to get these benefits, industries will have to navigate a host of thorny challenges involving hardware and software, standards, and even business models.

The Physics Test

The first set of issues facing RFID deployments is the physics of getting RFID readers to read tags accurately in real-world environments such as warehouses. Current success rates for tag readings run as low as 80 percent, explains Kara Romanow, a research director at AMR Research. "The tags just fail. The quality's just not there yet," she says. RFID readers have a hard time detecting the tags through interference from metal, liquid, nylon conveyor belts, and dense materials such as frozen meat and chicken parts.

Furthermore, as with other wireless technologies, the devil is in the details of reader infrastructure layouts and potential conflicts with other sources of wireless transmissions. "Every site's a little different. You can't just throw up antennae; there's a tuning aspect," says Tig Gilliam, a partner at IBM Business Consulting Services. "This is dirty fingernail stuff."

To further complicate matters, there are two types of tags: active tags, which contain a battery and can transmit further but have a shorter life span; and passive, which draw power from the reader and so have a shorter range but can live forever. And there are also two competing standards for transmission protocols, 13.56MHz and UHF, each regulated differently and with different effective ranges and abilities to avoid interference.

To pass the RFID physics test, experts recommend putting time and engineering resources into testing which RFID configurations will work best for your specific application.

"This is not a science; it's still an art," says Tony Sabetti, RFID products business manager at Texas Instruments, one of the largest tag producers. And it's a good idea to get cross-functional teams, including product, manufacturing, and supply chain experts involved in RFID planning, especially when RFID tags will be affixed to a product during the manufacturing process.

The Software Challenge

On the software side, the challenges to RFID deployment include getting the right data onto the tags in the first place, and then leveraging the data the tags generate as they navigate the supply chain across multiple enterprise systems. "Software is a huge issue," AMR Research's Romanow says. "Everyone's underestimating it."

Romanow says that for manufacturers, getting the data onto the RFID tags in the first place requires that existing order processing and fulfillment systems be capable of operating at the pallet, case, carton, or item levels, which many currently are not. For item-level RFID tracking, these systems also need to be able to handle sequential data (sequential serial numbers for a whole carton of otherwise identical products, for instance), but most existing applications are oriented toward classes of objects rather than sequences.

Once the tags are loaded with data, the real software challenges begin, starting with managing the reader infrastructure. Today's RFID readers are simple devices that typically lack an operating system, upgradable firmware, standardized drivers, or communication protocols -- although this is starting to change. "There are no consistent mechanisms to manage them, check their health, do software upgrades, and turn them on and off," says Javed Sikander, Microsoft's program manager for industry solutions enablement. "There's no standard interface across all the readers."

Next comes the middleware challenge of managing how the readers filter data coming from the tags as they move through the supply chain. "If I put a tag within the read field of an antenna, the reader reads it a couple hundred times a second," VeriSign's Brendsel explains. "You need a middleware layer which knows that the first time it sees it, it's an event, and the next 10,000 times its just garbage."

Once the data has been acquired and filtered, the work of feeding it to the appropriate applications begins. Most companies have multiple applications that need access to the RFID data, notes IBM's Gilliam, including warehouse management systems, inbound supply chain systems, planning systems, order management systems, and data warehouse and analytics systems. "You've got an enormous amount of data that has to be moved around the network among applications," Gilliam says.

And that data must be moved quickly and assigned accurately to the appropriate business process. "Just dealing with that volume of information coming in is going to test the limits for a lot of the real-time software out there," Microsoft's Sikander says. "The software pipes in between the devices and the host need to be able to pump data at a very high pace."

Although most of the largest software vendors have made announcements about supporting RFID, very few have actually released products, although that is expected to change this year. "I'd give the (large ERP vendors) a C -. Some have done next to nothing" in deploying RFID-enabled applications, AMR Research's Romanow says, though she points to exceptions such as SAP and middleware vendors OATSystems and ACSIS, and warehouse management vendors Manhattan Associates and RedPrairie.

For the foreseeable future, RFID deployments will require a lot of customer integration. "The consultants especially in the early going will play a huge role ... this is a huge opportunity for the systems integrators," Romanow says.

Collaboration and Data Issues

To maximize the benefits of RFID, data generated by the tags must be sharable by business partners and with all companies along the supply chain that are using the same tag. Two potential standards have emerged to facilitate this maximization: the EPC (Electronic Product Code) standard managed by the EPCglobal consortium and supported by Wal-Mart and many large retailers and vendors, and an alternate standard developed by ISO (International Organization for Standardization) and supported by numerous European companies and the U.S. Department of Defense.

Most analysts expect the two competing standards to merge quickly, probably before Wal-Mart's mandated January 2005 deadline for its top 100 suppliers to support RFID. The larger issue is how uniformly the resulting standard will actually be implemented, and how much collaboration it will ultimately enable.

The EPC standard works on the concept that each RFID tag acts as a "license plate." Rather than try to load today's capacity-constrained 64- or 96-bit tags with lots of data about the specific item or shipment, the idea is to simply store a unique identifier that will refer to detailed information stored elsewhere on a global network. This identifier would include a 'manager ID' (to identify the manufacturer), an 'object ID' (to identify the class of object), and a serial number for that individual item, as provided by the manufacturer. The EPCglobal consortium envisions a federated network of providers of value-added services based on the license plate, similar in architecture to the Internet itself.

But there are challenges. "How closely can we align that very thin (license plate) representation with the keys that already exist in databases like VINs (Vehicle Identification Numbers) or GTINs (Global Trade Item Numbers)?" asks Sue Hutchinson, EPCglobal product manager. "How do we control access so it's appropriate to the business relationship; how does Toys R Us make sure Mattel doesn't see what Hasbro's sending?"

The Business Case

A final and crucial hurdle for most RFID deployments is coming up with a business case to support the required investment. Although in theory the cost-saving and strategic benefits from RFID are huge, getting those benefits requires many trading partners to participate, as well as significant investments in hardware, software, and human resources, an investment AMR Research's Romanow claims can reach as high as US$13 million to $24 million in the first year for a typical large consumer products manufacturer.

"The problem is, no one cares about the total business case, they only care what their share of it is," IBM's Gilliam points out. "The big hitch is, when do you sort of get critical mass?"

Gilliam says given the lack of immediate ROI and uncertainty about standards and the specific details of compliance requirements from large buyers like Wal-Mart and the Department of Defense, most companies are taking one of two approaches.

The first is the so-called 'slap and ship' approach, or doing the minimum level of investment to slap tags onto a subset of outgoing shipments to comply with the current mandates. A second, smaller group is making bigger investments in deployments further upstream, in an effort to both comply with mandates and wring out some internal operational efficiencies from the technology.

"Some companies think 'Gosh, can I just wait a little longer? What's the minimum thing I have to do to comply?'" Gilliam explains, whereas others "have done a business case and know which products (if tagged) can offer a return."

But even the slap-and-shippers may have to get more involved than they thought, sooner than they thought, EPCglobal's Hutchinson warns. "Even with a slap and ship approach, there's still a good deal of (work) involved to make sure you have your tags programmed correctly and that you're tracking the goods for this RFID pilot implementation versus the rest of your supply chain" she says. "I think people are coming to the realization that they're going to be looking at communication to their back-office systems and middleware challenges sooner than they thought."