Apr 18, 2004fternoon, everyone. It’s always a daunting challenge to get the post lunch shift, but I will do my best to keep it entertaining. Just a quick note about the title. It was "RFID, the Future of Manufacturing," but I decided to temper it a little bit because, in a sense, there are a lot of very important short-term issues, that impact the manufacturing as well. And so I just thought it might be nice to balance some near, near-term and longer-term issues as we went through this.
I’m really going to just give a little bit of introduction to what the Auto-ID Labs are, and what the Auto-ID Center was, before talking a little bit about the issues that RFID is facing in manufacturing. Okay, so we have to understand what the Auto-ID Center was and the Auto-ID Lab is. The Auto-ID Center terminated at the end of October last year, so it was a set of research labs with six satellite centers. And it breaks into two parts; one section is EPCglobal, an organization run by EAN-UCC and tasked with promoting the use of EPC Technology. The Auto-ID Center was made up of six laboratories, one of which is Cambridge in UK. The others are in the United States, Australia, Switzerland, Japan and China. These labs are tasked with the ongoing research issues. (Download presentation.)
So, just to recap what the Auto-ID Center set out to do. This was back in 1999 and it was a 25th birthday for the barcode. Some people thought it was time to look at the next generation of bar code and so the Auto ID Center started with that in mind. What is it that the next generation of bar codes should be. People realized that to achieve better synchronization than we had at that time between physical flows and information flows in the supply chain, clearly some form of very easily automatable reliable technology for linking information to physical objects was needed. And importantly a single system was being envisioned to the whole supply chain and obviously for the reason we are all here today, RFID was identified as the initial key element for getting information of physical objects into information systems.
So, its important just to pick up what the key thrusts of the Auto-ID Center were, and the reason is when I start to talk about manufacturing, you will see that manufacturing has been doing some of these things for longer than Auto-ID Center, in different ways though. The first0 thrust was creating low-cost tags. Just drive the price down as low as possible by reducing the amount of silicon, therefore minimizing the amount of information on the chip and coming to a conclusion that an idea on the chip was kind of the minimal, that you needed in order to get the point to a database entry where you might hold other information.
The second equally important issue was that it still was very hard to see business justifications for single applications. So the idea of having multiple applications based around the same tag was very attractive, but what if there were some implications of that as well. Firstly if you are going to use the system both within and among the companies, then clearly a standard interoperable set of solutions around tags is required. Secondly, it's not sufficient to have standards for the hardware. You also need standards around the software, the collected information and also transferred information potentially between organizations. The overriding model that developed was that this idea of the EPC network, the Electronic Product Code network, which is the output of the Auto-ID Center, which really seemed not as a standalone RFID system but as an extension of the Internet.
So, whereas the Internet initially connected all computers, the vision of the Auto-ID Center is to connect things to the Internet. So it's a top down vision rather than just two or three individual tagging jobs. The outputs from the Auto-ID Center as of the end of the last year were a set of building blocks and I’m just going to illustrate those building blocks by way of an example taking a tea box. First thing was called the Electronic Product Code, a unique number system, which was developed to give manufacturing information, product type information and serial number information. A tag attached to the packet of tea. A reader interrogating that tag to retrieve the unique serial number. A system called the Physical Markup Language which, which encapsulated the EPC into a standard language format integrated with other information such as time and, physical location information. There is a software element called Savant which is a data filtering and logging component. And then finally, there is something called the EPC Information Service, which isn’t a database in itself but it provides a specification for a wrapper around existing databases, so the queries could be for example in a retail outlet, you might want to run a query as to the nature of the way a particular piece of fresh food was manufactured, and that clearly could be using EPC Information Service structure to interrogate a database potentially in a different form and model within the manufacturer’s organization.
So they are essentially the six elemental building blocks. Why did we build all six? In order to build a network of RFID systems rather than just standalone RFID systems, these pieces were necessary. In order to reduce chip prices, we built tag specs of the minimalist variety. In order to put the minimal information on a chip, we have built this Electronic Product Code structure. With that ID on a chip, we needed to hold all other information on a database and therefore we developed the EPC Information Service.
The second thrust was to use the same tag for multiple applications. Again, we needed to tag and read the standard specifications to ensure that one company, a distributor, can use the same tag as the manufacturer and the retailer can. In order to standardize the way data is managed and communicated, we built this data filtering specification and an open markup language. And then seeing the EPC network as a extension to the Internet deployed the ONS on the EPC Information Service. So, apologies to those of you that know that story, but that’s just a quick resume on what the Auto-ID Center did. It seemed like a lot more at the time. Actually, that is facetious, we did do a lot more, probably.
RFID existed before the Auto-ID Center and particular in manufacturing it definitely did, and if you look at this, this set of applications, what I have done is just picked up some characteristic of systems, complexity of information, single and multiple applications, volumes, and expected life, and if you draw a red line, then in a sense the red line picks up pre- and post-2000. And you can see that really what drove the Auto-ID Center’s activities was a set of characteristics that had not existed before and that is kind of a key piece of information there, and I will keep coming back to that.
Our laboratory in Cambridge wasn’t an RFID laboratory. It was a laboratory designed around what you do when you get information from RFID. And what we reduced that down to was, what you do when you get enhanced product information. Because really the value of the EPC network system is about enhancing the quality of information and then particularly, allowing you to make better decisions than you had made before.
So a lot of our work these days is looking at what would we mean by improved information quality, and it essentially comes down to the three dimension you see there. Either improvement in terms of accuracy, or what we call completeness, so, being able to access all information about an item, not just product type but perhaps it is specification, or perhaps it is temperature history as the DOD people were talking about this morning, and timeliness. Decisions are generally time driven, so being able to get information in a timely manner has a massive impact and, for example, some of the work we are doing in enhanced shelf-replenishment policies, in retail is massively impacted by timeliness.
You see where you might like to be with the ideal product information system and where we were previously is somewhere down at the bottom, where there were the massive trade-offs generally occurring between timeliness and accuracy, whether it was in warehouse management or in retail or in manufacturing, the same applies. And really what we have been trying to do with the Auto-ID Center work was to drive up towards that top corner, and the RFID deployments with EPC network is seen as the key way to get not only accuracy in timeliness, which you would get with any RFID deployment, but with a network element as well being able to lead to a greater level of completeness.
So there are the three dimensions I talk about. I just wanted to emphasize two things. The benefits are only extracted when information is turned into action. So just because you have got RFID read, does not mean you have any benefits. And the second thing is, essentially the underpinning element for a lot of the RFID work was that it is hard to automate bar code reading, but it is easy to automate RFID. I’m not explicitly saying it, but automation is a key element as well.
Just some comments on issues relating to RFID in manufacturing, and I think it is a nice discussion we are having this afternoon after this morning’s focus on retail and DOD compliance which, really for lot of manufacturers, is kind of after the event in terms of their being able to get benefits as well of costs. Now in my mind, manufacturing is kind of a worthy recipient of RFID attention. It’s actually got, as I said, little direct attention lately, but in fact, it’s the most technologically advanced element of supply chain and therefore most amenable, probably, to a high-tech deployment like RFID.
RFID systems are real time inherently, and the IT systems in manufacturing are actually well geared towards that where probably the same cannot be said for retail information systems in general. Certainly in the UK and Europe and Japan and increasingly in the US, the manufacturer is responsible for not just his own operation but for his products throughout their entire life cycle. I will come back to that. In some industries, such as the auto industry, the manufacturer is often the central hub of the supply chain, so if the auto manufacturers do not adopt RFID, then it is kind of unlikely that in any event supply chains are going to adopt it. You would not see the same pressure coming from the dealerships, for example. Privacy was a big issue for a lot of us in the Auto-ID Center last year; it continues to be an important one. Manufacturing is rather decoupled from the privacy debate, because the consumer bodies are decoupled from the manufacturing activities.
The game is coming around two areas. I am going to talk more about that. Short term gains in terms of cost of production, and longer term gains in terms of value add. So actually, what I am going to do in the next few minutes is just pick out the drivers for manufacturing involvement, on how we are going to touch the zero term, what I call the zero term driver, compliance driven driver, but I will talk a little bit about the near-term, where there are some massive gains to be had cost wise. In the medium to longer term perhaps enduring value is the key driver.
The phrase slap and ship was used a number of times today, and clearly, slap and ship fits around the outbound category and has minimum impact in terms of the manufacturers’ business. I think one of the key areas that has largely been untouched by this latest wave of RFID is production. We talk about some asset tracking applications but very little product tracking is done during quality control. I look forward to the day when there is a manufacturer making an announcement along the lines of the retail announcements that everyone is scrambling to meet. In the longer term, we see real value in being able to integrate design and go right back to where the concept and the product specifications are developed.
I guess the ultimate issue for the manufacturers is how to make the most of the information coming back from the customer, whether it is the positive type of information—demand forecasting—or whether it is negative,. But if product recalls are on the cards, then being able to do effective track and trace and post deployment quality control clearly is a very big issue.
Another characteristic that I want to talk about is the way people are thinking in terms of RFID tagging. It is quite skewed towards the retail operation, so we are typically talking about these three levels—item, case and pallet—and obviously most of the focus is case and pallet to date. If one were to abstract that, and as academics, we love abstracting and forming general theories of RFID relativity or something, I think that these three levels are really about entity aggregate and reusable asset. And the reason I try to use that description is that then gives us a nice feed into how manufacturing might use it. In fact if business is kind of skewed towards auto manufacturing, but the equivalent entity might actually be a component in the case of an auto operation, that the aggregate could be a single car body assembly, and that the reusable asset rather than a pallet might be a carrier used in an automatic production line.
So that the key point to pick up there is that actually, you have got two entirely separate tagging strategies. On is a product based tagging strategy, and one is a resource based tagging strategy. To emphasize that a little bit more, what this diagram tells you is what happens to those different sets of tags. Product tags, whether aggregates or items, follow a product life cycle. They go from company to company, and asset tags tend to follow a circular path, often within a singular organization or a within a set of organizations.
Whether you go down one route or the other massively influences the type of output you are likely to get. And issues such as cost, the flexibility and range of, the robustness of the particular tag, whether it has got to be used many times, or across many locations has a big impact, and also the issue of standards, because if you have got a reusable asset, just used within your organization, you’re not so heavily driven to meet someone else’s standards at least from a compatibility point of view.
I saw the numbers this morning in the Forrester, really very good talk, from Forrester, which had integration down to it being about $300,000 out of a ten million dollar budget for a one-year deployment. That seemed to be quite low. My experience with integration, seeing other people’s integration efforts and their own experience, it has been that it is quite a high target. And there’s lots of integration challenges, not just that. Integrating of physical hardware, integrating into the existing information systems—this middleware debate is not going to go away for quite a while. And then ultimately, integrating within your business processes.
We like to talk about three levels of integration:
Connected—simply just replacing bar codes in a stand-alone mode.
Coordinated—enhancing existing processes
Coherent—actually redesigning processes from scratch potentially to leverage gains from RFID or the EPC network.
So now I was going to talk a little bit about near-term deployments, and as I said, I am really just scraping over the surface some of these issues, but hopefully there are a couple of points to take away. I said I wouldn’t talk about compliance issues because it’s a well discussed at this meeting, so I will jump over that. I think everyone has identified that there is little direct benefits to the manufacturers, but there are some positive sides apart from the fact that there is little choice. There is an opportunity to influence adoption pathways, because it is quite certain that the way in which tags and readers develop over the next year or two will be massively influenced by these compliance requirements. So if manufacturers are not involved now, they will not be able to influence that. And we can identify quite a long list of issues that people might want to influence both in the hardware and software areas.
Okay, so the near-term drivers are in areas where there might begin to be some return on investment, either through cost reduction or improved performance. Cost reductions come under two categories, generally. One is a safety stock reduction. Safety stocks are due to demand forecasting uncertainty, but also in part to the general uncertainty in supply chain. Hopefully, EPC will be able to reduce that. Also, many of the inventory manager's, warehouse manager's approaches are based around manual stock counting, at least partly.
But also, there is the delivery issue. to be able to move towards aggregate tracking where what-you-see-is-what-you-get, is clearly a big target. And quality assurance has enormous cost associated with it. So we have identified the competitive advantages to early adopters in the manufacturing domain. Do not just go to slap and ship, but actually start to do this sort of thing now in conjunction with these slap and ship activities.
In terms of systems integration, I just want to make a couple of brief points. I think I have said it several times, just thinking about RFID as being a sensing activity is really very limited. In fact, one needs to consider the whole set of decisions and actions coming out of that, and we have identified a number of issues linked to sensing issues. So bear in mind, a very important point is, you don’t get location from RFID, okay, you get some very poor sense of location, you don’t get location. So if you are going to automate, do not assume RFID is going to give you that.
The route to integration of EPC data is certainly non-trivial. So there are three issues that need to be addressed in order to achieve what we call coordinated structure. If I now just pick out a couple of examples, and in fact these are illustrative rather than even best practices within the individual organizations involved, but just to give you an example of near term deployment. This is a case with an [fast moving consumer goods] manufacturer in Europe who had looked at bar codes and decided that in order to meet their requirements in their outbound processing, they needed to increase speed and accuracy simultaneously. They came up with a really rather novel approach to managing the quite high costs of tags at that time. They embedded tags not into pallets, not onto cases, not onto items, but into the ground. And put readers on their forklifts and actually by linking information on the pallets to the forklifts, were able to trace items really quite well.
I call that type of deployment really a connected deployment, because it was built to be a standalone solution for a one-off problem within that organization. But the focus there is on asset tagging. Another well reported auto parts manufacturing example was to develop tag and receipt carriers for automotive OEMs in trying to insure high accuracy, because the failure to deliver the right seats to the right organization has a relatively big impact.
So the driver from a product information point of view was accuracy here, again it is a high-frequency, not a UHF solution. The deployment was again in outbound, although, I think very cleverly in this case, they were actually able to use the fact that they were tagging their outbound seats, and work that back into production to get some gains in terms of increasing the use of production resources, because they knew where their items were, and very accurately were able to actually increase some of the flexibility of their operations.
So to summarize where manufacturing near-term is heading, it is down the route of reducing the cost and increasing the accuracy of product information. People will focus around ID tags and readers, the hardware elements, and outbound and warehousing are likely to be the early deployment areas, and tags deployed at an asset level, or in some cases a limited aggregate level.
Now, if you think about asset deployment, as I said before it doesn’t fit in with any of the drivers of the Auto-ID center. So we would like to see a merging between the existing asset tagging strategies that manufacturers have deployed today, and the slap and ship model being developed on aggregate tagging. The integration at best is going to achieve coordination.
In contrast to the sort of benefits—short term is probably up to three years, perhaps even up to five years, particularly given that most of the energy is heading at the outbound zone at the moment—the longer term areas are of greater interest, really, because they represent the value adding gains and I know in America and also in Europe over the last five or ten years, the great holy grail in manufacturing facilities has being responsiveness or agility, and to a certain extent, the early deployments of RFID were doing nothing to help that, but in the longer term, there are some really attractive gains to be made in terms of using RFID as a means of actually changing the business model to get greater responsiveness and particularly greater customization.
People can do customization now, but it is very expensive. So how do you customize and do it as business as usual for reasonable cost? RIFD has a role to play there. Or enhancing the offering, so again, the RIFD network means that you got have an ability to extract information about a product to the product itself, as it moves through it’s life cycle so the user can actually access information about his product as well as about the product at purchase point. And there are a lot of people looking at product extensions. Some of them are attractive from a business point of view—service life management—and the others are actually becoming legislatively necessary in Europe, I suppose the laws are extremely tough now, and each manufacturer must show a return route on quite a range of products—auto and electrical being two examples.
So one of the gains we see there is perhaps where cost reduction and performance improvement are going to be reasonably standard things people do. There is an enormous opportunity for significant differentiation in the second phase. It is going to require alignment with core business strategies and probably require some reengineering, and it will be a longer but more rewarding phase. And again, if they go back to the integration question, whereas all the coordination issues will be clearly required in the second phase, for us the big issue is right over on the right hand side. How is EPC, RFID information going to influence decisions, and how do the physical operations change to enhance, to directly exploit the viability of that type of information?
Okay, so a couple of examples. And these examples are current. An area manufacturer in the U.S., has actually announced that they are starting to tag not only their assets but actually parts and the related physical paperwork associated with components, reading this through their supply chain. So they’re doing this mainly to tackle the complexity, but actually getting gains in terms of better asset utilization. Interestingly, if they were to go to the RFID network level, they may have actually done away with some of that paperwork, but there are other restrictions on that.
So here is an accuracy issue predominantly, but actually moving towards completeness, being able to have complete information associated with any one component being used in their supply chain. Deployment is inbound and in production, so that is again a different domain to the typical deployments in manufacturing. There are tagging reusable containers at present, but there is a sense of coherence about the strategy in this example which is quite interesting.
Another recent announcement for an auto sports vehicle manufacturer who is looking to use RFID within their operations to actually carry cassettes of customization requirements directly with the systems, directly with the physical equipment that they are building as they go, and so the driver here is not accuracy or timeliness, it is about completeness of information, being able to associate for every single order, an entirely separate set of requirements, and having those requirements available in real time. Right now again, they are deploying on a carrier system, so again there are no gains further down the supply chain from that deployment.
Finally, within our laboratory, we are developing solutions, I guess we try to pick up all those characteristics in the longer term system. And you are all welcome to visit the laboratory in Cambridge. We have developed a laboratory scale, but a serious industrial deployment of a packaging system, and we now have a machining assembly and packaging outfit developed there and the idea behind this environment is being two fold. One is to test in hardware the use of EPC Network information and real time control, so connecting EPC information in real time to the functions of this robotic assembly cell. And the second one has been to try some novel strategies around developing solutions for orders sent by customers to actually drive their way through production. So rather than an order being passively handed over to a manufacturer, the order arrives, has a piece of intelligent software linked to it, it physically connects itself to the raw materials as part of that order through the RFID tags, and actually the order sets up all the control tasks that it requires to make itself. So that means that the system has to be entirely adaptable to any order that arrives within constraints. So in terms of this system, what we’ve focused on is timeliness and the real time aspect, but also completeness, the ability to capture an entire order and link it to the physical product.
It is an HF application, but we are doing a UFH changeover at the moment that focuses on inbound production. Importantly, we are looking at item and aggregate tagging in manufacturing, and I do not think there are many other places where that type of application is being considered. Our target is coherence.
In terms of long-term characteristics, we see completeness of product information as being perhaps the most important issue. The system components which are connected to this are really in and around middleware in the networking aspects, so we are seeing that in the short term, the RFID physics is going to be sorted out. We see more gains in production and inbound for a manufacturer than at the outbound, obviously. Tag deployment in terms of getting the real gains and the vision for the Auto-ID Center should be at the aggregate or item level, not carrier level, and integration coherency must be the target.
Just a couple of concluding comments. Firstly I have kind of jumped over it, but there are a number of barriers, non-standard deployments being one that is talked about a lot today, being able to solve complex tagging situations, effective integration and that’s still a way off. The next thing, there are some constraints in the way our supply chain is constructed today, which make it harder to extract benefits than we need to. Also, I just wanted to make one short point — if you are thinking about the gains of manufacturing, they do not just occur by implementations within the manufacturing domain. There are gains from upstream and downstream deployments for the manufacturer, if the manufacturer works out how to best exploit those.
Multiple applications, the shared product orienting information view is really important, so sharing information is key to getting that wide a deployment going. And therefore, viewing RFID as a network extension. And I will just make a point that in many industries, various manufacturers are becoming more and more held responsible for their own products. So just a finishing pitch, if you are in the manufacturing domain or have interest in the manufacturing domain, the Auto-ID labs is running a set of what we call special interest projects, and there is one being launched in the manufacturing domain, specifically around these near and longer term issues. And the website is down at the bottom there, I understand that the overheads will be made available from this talk, and if you are in any doubt, go to www.autoidlabs.org will get you there eventually anyway. Thank you very much.