When you hear the litany of reliability and readability woes that plague radio frequency identification (RFID) pilot projects in the supply chain arena, you may wonder how anyone is making any progress with this technology at all.
Problems range from nonfunctioning tags to environmental conditions, such as temperature, humidity and radio frequency interference, that can render tags unreadable. The challenges continue as containers move through the warehouse and onto trucks -- a process in which tags can be damaged or thrown out of alignment so they can't receive reader signals. And the list goes on: Forklifts cut cables, tag printers can't keep up with conveyor belt speeds, or adhesives that bond tags to containers fail at low temperatures.
"There's still a lot of immaturity in the technology," says Kara Romanow, research director at AMR Research Inc. She points to the 10 percent to 12 percent of tags that are "dead on arrival" -- meaning they arrive at the user site in nonworking condition. Even when you weed out the bum tags, average read rates are still just 80 percent to 90 percent, she says. Some of this is the result of tag failure, but other common problems are incompatible tag/reader combinations and products that aren't suited for RFID. "Companies are trying to tag RF-unfriendly products, like soup cans and baby wipes," she says. "It really varies by product."
And yet, if you talk to RFID users like Gene Obrock, vice president of operations at Henkel Consumer Adhesives, or David Adams, senior vice president of corporate strategy at TrenStar, you find that they are more optimistic. Yes, they agree, RFID systems are fraught with reliability pitfalls, some of which never go away. But the potential payback is so great, they say, that building an RFID-based supply chain that takes these issues into consideration is worth the trouble -- even if the benefits are five to 10 years away.
"The equipment is becoming more durable, the technology more reliable, the tags are improving, and at the same time, we're more knowledgeable about the relationship between the technology and the packages we sell to our retail customers," says Obrock, whose company is testing RFID with Wal-Mart Stores.
Obrock is clear that reliability will play a major role in determining Henkel's eventual return on investment, which will be achieved in part by obtaining more accurate and timely inventory data from retailers. But valuable data can be collected only when the volume of Henkel's RFID-tagged packages increases. "It may take five or six years, but eventually it will grow into the fabric of our organization, like UPC bar codes," Obrock says.
Obrock and other experienced users advise companies to learn as much as they can about the relationship between RFID and their own products and processes and then build a system that anticipates and resolves reliability problems that arise.
DOA tags are common enough that experienced users have found ways to work around them. They arrange with tag manufacturers, for instance, to ship extra tags to compensate for those that don't work. Many printers today also include a readability tester that rejects bum tags, minimizing the chances that one will ever be applied to a container.
Of course, that doesn't resolve the issue of tags that test fine but fail later. That's why users that require the highest reliability guarantees must work directly with tag manufacturers on quality assurance -- and pay a premium to certify that the tags will work.
This extra cost is acceptable to Atul Salgaonkar, founder of RFID Solutions, which develops RFID systems for clients in the pharmaceutical and health care industries, where accuracy is paramount. "If I have 10 samples and only eight are read, I don't want to have to do a song and dance in front of my client," he says.
Salgaonkar proposes that manufacturers sell different flavors of tags with prices reflecting variances in reliability. The first step down that path would be for the tag vendors to publicize bit-failure-over-time rates, which they currently don't do. In the meantime, Salgaonkar advises clients to attach not one but two tags to pallets -- which is fine for his clients, whose high-end products warrant the extra cost.
Some observers see tag reliability becoming less of a factor as bigger-name companies such as Texas Instruments and Hitachi Data Systems enter the business. "There's a good chance the reliable-tag issue will be resolved by the end of the second quarter next year," predicts Steve Banker, service director of supply chain management at ARC Advisory Group.
For their part, retailers seem willing to accept some level of inaccuracy. "We're receiving rules that say, 'If every week you send me a pallet with 80 cases, and this week I only read 79, I will understand it as 100 percent,' " Obrock says. Indeed, Banker expresses doubt that the technology will ever reach 100 percent accuracy at the pallet level, where tightly packed goods and other causes of interference can lead to readability problems. Retailers "will have to make a leap of faith that if they can read 19 out of 20 SKUs on the pallet, that the 20th one is there," he says.
What will continue to suck up lots of R&D dollars, however, is studying the nature of the products being tagged. "In the dry-goods space, people are getting close to 100 percent reads, some at a rate of 600 feet per minute," says Eric Peters, CEO of True Demand Software, an RFID startup in Menlo Park, California. But with metal containers and anything containing liquids, read rates are a much lower 60 percent to 75 percent, or even 15 percent on some metal containers, he says.
Companies need to get creative to overcome these problems. When the Department of Defense had trouble tagging 55-gallon metal drums, Peters says, it finally succeeded by placing weatherstripping behind the tags, which minimized interference.
With Henkel's metal tape products, tags have to be placed in a very specific position on the box, or readers will pick up reflections from the tape. Meanwhile, a completely different type of tag works better with the company's shelf-liner products. That's where the cost really comes in: Companies are spending research dollars trying different tags and tag placements with all of their products -- and they're documenting the results and making sure workers abide by the explicit instructions. They might end up requiring a variety of tags, with different form factors or different antennas, to accommodate all of the variables in their product lines.
"If you have 10 different SKUs and 10 different tags that you're trying to test, without fail there won't be one tag type suitable for all 10," says Brian Higgins, director of global RFID solutions at BearingPoint.
That, of course, increases costs. "You can't get volume discounts if you're spreading out tag procurement purchasing power over 10 different tag types," Higgins says. And in addition to increasing costs, using different tags disrupts and adds complexity to your operations. Custom tags are also more expensive than the 30-cent variety, Romanow points out. "Whatever business case you have really goes negative when you start getting into custom tags," she says.
The challenge is to find the lowest common denominator, or the tag that works best on the widest swath of SKUs. For all these reasons, Higgins says, there's a gap between working through readability concerns and true deployment.
All in all, users have to walk carefully on the cost/performance balance beam when they engage in all this experimentation. "There are ways to enhance reliability, certainly. But the challenge is keeping it on a cost/performance curve so it's useful," Higgins says.
What happens next
So, you've worked through all the experimentation, and you've gotten an RFID pilot up and running. Don't celebrate yet -- there's room for more trouble down the road. "Pilots are neat, where you get 100 percent read rates and give each other high-fives," says Trenstar's Adams. "But when you turn on the conveyor belt, and you're reading 1,800 containers an hour, it's a whole different ballgame."
Adams knows what he's talking about -- TrenStar, a mobile asset management firm, tracks tens of thousands of RFID-tagged containers. The company is so comfortable with the reliability rate of its RFID system that it builds a key performance metric into customer contracts, promising 98 percent to 99 percent read rates. But the only way it can do this is by building intelligence into its network to pick up deviations and degradations quickly enough that someone can respond to them.
"Reliability is so much about the quality assurance process you have in place so that when things start to not work, you can pick up on that before it turns into a train wreck," Adams says.
TrenStar's RFID system includes a series of checkpoints that containers move through, each of which signals whether the container was read accurately. If a checkpoint is missed, the system alerts someone who can physically investigate whether the problem was a one-time anomaly or something systemic like a cut cable or burned-out battery. "We've put in business logic that says, 'If this occurs, take this action or e-mail this person, and if it happens 100 times, escalate the problem,' " Adams says.
Warehouse to worker
Following the RFID reliability trail of tears eventually leads to the most important point of all: ensuring the integrity of the data collected by this technology. Few companies have progressed to the point of integrating the RFID data into their warehouse and ERP systems, so it's too soon to tell whether the data will be accurate.
But companies like Henkel are preparing for the problems that will arise. "If I ship a pallet with 80 cases, and the number that gets read is 76, and if every pallet is read that way, what do I do with that data?" Obrock says. You can round up to 80, assuming a misread, but what if a large number of pallets were truly four containers short?
The fact is, all the assumptions that you build into your business rules need to be reviewed before using this new data -- and they need to continue to be reviewed as the technology improves. "You can't tell your data folks to assume data is off by 10 percent, and eight months later, it's actually only off by 3 percent," says Obrock. "You have to go back every so often to validate the difference between the real situation and modify the conventions around how they use that data."
While reliability and readability may not top the list of factors that will affect RFID adoption, they will, in the end, affect how meaningful the resulting data will be. One reason is that a large volume of data is needed in order for it to be useful.
"In a lot of these pilots, people are tagging 5 percent, 10 percent, 20 percent of their overall SKUs," Higgins says. "It's difficult to do any planning when you're only capturing that small of a fraction."
Still, experienced users are enthusiastic about the advances being made in RFID reliability, and they point out that the more users experiment and learn -- particularly in these early days when implementers are most likely to be open about their experiences -- the better off they'll be.
"Those who started three years ago will stay three years ahead. Others will catch up, but they'll be learning it on their own dime," Obrock says. "Once some of the retailers and manufacturers get good at it, I probably won't share all my insights because it will be hard-core knowledge." 50759
ARC Advisory Group recently conducted a survey of 24 manufacturers and distributors engaged in RFID implementations. Here are some findings from the study:
-- Twenty-three of the respondents said they didn't believe they would achieve ROI within two years.
-- Respondents said that, through experimentation, they were able to achieve 100 percent accurate reads at the individual case level. However, at the pallet level, "experience was all over the map," says Steve Banker, service director of supply chain management at ARC. Some respondents with hard-to-read materials were achieving just 50 percent accuracy, while users with large items and pallets with no interior cases fared better. "Most folks were able to read every case on a pallet only 70 percent of the time," Banker says. Respondents said they'd need 95 percent to 99 percent accuracy at the pallet level to see benefits of RFID such as reduced inventory levels.
-- Respondents attempting to automate tag application found that printer/encoders couldn't keep up with their conveyor belts. Successful alternatives, such as robotic applicators, were difficult to find. "Hardware suppliers think they're two generations away from resolving the speed issue," Banker says.
-- Automated tag application is also complicated by nonfunctioning tags. Many printers reject bum tags and generate voided-out ones, but if there are two of these in a row, it disrupts automated operations.
-- None of the respondents said they were willing to make a high-volume tag purchase commitment. "They're waiting for Gen 2 standards to be approved, and there's a new raft of hardware providers," Banker says. "The wait is partially in hope that they can negotiate a better price, but it's more about quality and reliability."