By now, the monolithic factory was supposed to have given way to the virtual factory: a community of dozens, if not hundreds, of factories, each focused on what it does best, all linked by an electronic network that would enable them to operate as one—flexibly and inexpensively—regardless of their locations. This network would make it easy for companies with dissimilar computer systems to exchange information about inventory levels and delivery schedules. It would allow companies with different CAD systems to collaborate electronically on designs. It would permit potential suppliers to gain entry to the system in order to bid on jobs with minimal hassle and little or no investment. And finally, it would allow a small manufacturer to have the same access to information as a large partner.

For most companies, however, true electronic collaboration remains elusive. Networks for producing autos, textiles, and many other products do exist. But when one looks at how they share information, one is reminded of what Dr. Samuel Johnson said: “It is not done well; but you are surprised to find it done at all.” Even highly sophisticated companies have found—and continue to find—the task of creating seamless electronic networks of lean, computer-integrated manufacturing operations to be frustrating and difficult. Managers at most of these companies are still struggling to make their information systems more flexible. They are perplexed about why so much paper is still being shuffled around. They are desperate to figure out how to extend the network to more of their partners without causing costs and overhead to balloon. And they do not understand why their heavy investments in IT have not radically changed the way their companies work.

Clearly, the three main technologies that companies have employed to create the virtual factory—electronic data interchange (EDI), proprietary groupware (such as Lotus Notes), and dedicated wide-area networks—are not complete solutions. Although that conclusion is hardly a revelation, many managers do not understand exactly why these technologies are not delivering. The reasons become clearer if one thinks about the different demands a network must meet for a large-scale virtual factory to succeed.

Working with companies in industries such as electronics, white goods, paper, and aerospace, we discerned three basic demands on such a network:

  • It must be able to accommodate network members whose IT sophistication varies enormously—from the small machine shop with a single PC in the corner to the large site that boasts an array of engineering workstations and mainframes.
  • While maintaining a high level of security, it must be able to cope with a constantly churning pool of suppliers and customers whose relationships vary enormously in intimacy and scope.
  • It must give its members a great deal of functionality, including the capacity to transfer files between computers, the power to access common pools of information, and the capability to access and utilize all the programs on a computer located at a distant site.

EDI, groupware, and wide-area networks can each deal with some of these demands, but none can deal with all of them, nor can combinations of the three technologies. Does this sad fact mean that the virtual factory remains a mirage—a wonderful destination that can’t ever be reached? The answer is no. Real virtual factories are now being built. For example, AeroTech, a small, relatively young information-services company, has built one for McDonnell Douglas Aerospace that represents a radical departure from the approaches that others have taken. (See the insert “The Real Virtual Factory That AeroTech Built.”) AeroTech has created a networked manufacturing community that is open and friendly to even the most unsophisticated users, provides a very high degree of functionality, and works even though the community’s membership is constantly changing.

The Real Virtual Factory That AeroTech Built

AeroTech Service Group, which is based in St. Louis, Missouri, has built a highly effective virtual factory with McDonnell …

Two critical elements make this type of networked manufacturing community possible: a function we call an information broker and open standards based on the protocols established for the Internet. As anyone who has used the Internet’s World Wide Web knows, open standards make it relatively easy for members of a community to share information regardless of differences in their individual IT systems. In addition, they make it possible for members to use one another’s computing power. Finally, the Internet’s open standards permit each member of an internetwork to pick the communication channels—from a normal phone line to a high-speed connection—that are best suited for it to carry out its role in the virtual factory. In a real virtual factory, the network is the factory.

The information broker, which is an outside vendor in McDonnell Douglas’s virtual factory, performs a variety of functions. It signs up new partners. It keeps track of the network’s members and the number and level of relationships that each has with others in the network. It oversees security, constantly ensuring that each partner has the proper security clearance and access codes. Although AeroTech does not yet do so, an information broker could also serve as a converter, employing powerful conversion software to permit partners who have different formats or proprietary software to exchange information.

The Demands of a Virtual Factory

Before laying out the mechanics of an information-brokered manufacturing internetwork—a real virtual factory—it might help if we first explored why existing forms of EDI, groupware, and wide-area networks are inadequate. To that end, let’s examine how each can or cannot satisfy the three main demands on a virtual factory: (1) that it be able to incorporate partners at any stage of a relationship, (2) that it be able to incorporate partners with all levels of IT sophistication, and (3) that it be able to provide all required functionality.

Different Stages.

Like people in a romantic relationship, manufacturing partners typically move through progressively closer stages—from dating through engagement to marriage. During the exploratory stage, companies learn about each other and establish norms of interaction and bases for further involvement. For companies, exploratory activities include requesting, sending, and obtaining information about products and services; distributing requests for bids and receiving quotes; and establishing contracts and purchase orders. Collaboration entails more sharing and planning. Companies at this stage have agreed, for example, to work together as customer and supplier and therefore want to exchange and review more detailed data, such as CAD/CAM files and manufacturing-process documentation. Integrated manufacturing partners expect a continuing relationship—for example, in a joint venture. Their activities include sharing data about production, inventory, and schedules and accessing the information and applications resident on each other’s machines.

To support partnerships at all stages, information systems must be secure yet easy to enter.

At any one moment, it is common for a company to be at different stages with its various partners. It may have several joint-venture partners and at the same time may be constantly shuffling customers and making changes in its relationships with suppliers. To support many relationships at all stages, information systems must be easy to enter and leave: a network that is expensive or difficult to join will discourage exploration. In addition, networks must be secure: potential participants will not join a system that will expose their internal networks or their transmissions to spies and hackers.

Different Levels of IT Sophistication.

Regardless of the stage of their relationship, participants in a virtual factory will vary greatly in their level of sophistication about information technology. That level is a combination of several factors, including the type and power of installed hardware and software; the average and highest level of computer expertise among site personnel; and the degree to which people on the site are already connected to one another by an internal network.

Within the typical large manufacturing company, different groups are themselves at different levels. A design group might have a cluster of linked workstations running advanced drawing and modeling software. The production control function might access mainframe manufacturing-resource-planning (MRP) software from dumb terminals. A sales manager might have a lone PC. And it would not be unusual if none of the three could interact with the others electronically, even if they were all on the same site.

There is also tremendous variation across smaller companies. For example, a subcontractor specializing in finite element analysis or other modeling techniques is likely to have an advanced computing environment, whereas a supplier of packaging materials may have only rudimentary systems. The wide range means that any technology underpinning a real virtual factory must be easy to implement, even on low-end hardware.

High Functionality.

Manufacturing groups have both heavy and complex information-sharing requirements. By heavy, we mean the great volume of information involved in any manufacturing process. For example, consider the body of documentation needed to specify how to build even a relatively simple part, the different information formats typically used, and the information required to move the item through the production process. As the part moves among partners’ sites, this ever mounting pile of information must go with it. For complex parts, the amount of information required can be staggering. The paper required to fully describe the production of a Boeing 747 undoubtedly would not fit inside the airplane itself.

Information-sharing requirements within manufacturing are also complex. The IT process required to send a purchase order, for example, is very different from the process required to send an assembly. The former is a simple transmission of text, whereas the latter is an interactive manipulation of graphical data. Both, however, are common activities that a real virtual factory should support.

Sharing and manipulating all information on a network involve three distinct types of functionality. In ascending order of complexity, they are: simple data transmission, data access, and access to applications, or what we call telepresence.

Transmission is the most straightforward; it is simply sending a packet of information from one place to another. Partners in a virtual factory need to exchange all kinds of information, from E-mail to purchase orders to numerical control programs. Transmissions such as videoconferences need to occur in real time, but most do not.

Data access capabilities permit members of the community to share common pools of information in more sophisticated ways than simply by passing messages to one another. For example, by creating virtual bulletin boards and file cabinets that authorized users can open, companies can make sure that all participants in a product development project are abiding by the same schedule, that updated CAD files are always available to suppliers, or that regulators can monitor emissions levels. A real virtual factory must be able to permit some users to add or change information in databases in addition to viewing it.

The highest level of required functionality is telepresence, a capability that allows all authorized people to see and use the programs resident on a given computer, whether the users are company insiders or outsiders and whether they are on-site or far away. This attribute is one of the key advantages of the Internet, which lets cybernauts jump from machine to machine around the world, making use of the information and applications on each.

Telepresence allows people at a distant location to see and use programs on any computer.

Telepresence is extraordinarily useful within a virtual factory. For example, it allows small companies to use the number-crunching power of a large partner for computationally intensive simulations, and it permits a customer to check the status of an order by logging on to a supplier’s order-management system. That kind of activity can even be performed by means of a home page on the World Wide Web. (See the insert “Virtual Factories, the World Wide Web, and Java.”)

Virtual Factories, the World Wide Web, and Java

Advances in internetworking technologies are making virtual factories easier to build. First, there is the much vaunted …

A real virtual factory, then, provides all this functionality to its partners, regardless of the stage of their relationships or their level of IT sophistication. The diagram “Three Factors Determine the Ease of Information Sharing” provides a framework that can help managers analyze whether their current solutions are adequate for their needs. The figure shows, for example, that it is much easier for companies at the integrated stage to agree on and build an information sharing structure than it is for companies that are only at the cooperative stage. Similarly, it shows that it is much harder to connect naïve IT users than sophisticated ones or to provide access to applications—real telepresence.

Three Factors Determine the Ease of Information Sharing

Current Approaches

To compare different approaches to electronic collaboration, note how much of the cube each fills. A system that fills a lot of the space, especially one that reaches close to the back corner (where unsophisticated companies at the exploratory stage can transparently use others’ machines), makes real virtual factories possible. Let’s take a look at the three different categories of technology currently in use and see how they stack up.

Electronic Data Interchange.

EDI is the oldest form of electronic collaboration among manufacturers; it grew out of a need to simplify the paperwork for administering the Berlin airlift. Today’s EDI uses a collection of common formats for communicating data between companies. It is used most frequently to exchange data such as purchase orders, to execute transfers of electronic funds, or to provide delivery information to customers. EDI standards specify how each of these information transfers should be structured so that any party using those formats can accept transmission from any other party using them.

However, conventional forms of EDI cannot satisfy all the demands of a virtual factory. Despite the existence of some common standards, many systems are still inflexible and proprietary. As a consequence, it is expensive and time consuming both to add new members to such a network and to expand the types of information exchanged on it. Depending on the particular network, it can cost tens of thousands of dollars to add an EDI link and to mold one’s own computer protocols to those used by the dominant customer. Such attributes mean that conventional EDI is best suited for linking the members of a relatively small, stable community—particularly a community in which one member is powerful enough to demand adherence to its communications standards. Conversely, it is ill suited for communities with a large number of transient members or members with limited IT resources. With traditional EDI, every time a new member is added to the existing system, a dedicated line—and in many cases, a special terminal on the member’s premises—must be installed.

Conventional EDI has other limitations. It does not easily permit members of the community to exchange information with one another, because the system has to be specially configured to create each link between each pair of members that want to communicate. EDI networks tend to be used only to send information in batches and are awkward for creating real-time links between sites. Also, current EDI systems are not designed to allow members to operate a partner’s computer from a remote location in order to use its applications software and computing power or to access its files.

Current EDI fills very little of the virtual factory’s requirements. It provides only a low level of functionality and, because of its expense and fixed costs, is appropriate only for integrated partners. (See the diagram “Electronic Data Interchange and the Virtual Factory’s Needs.”)

Electronic Data Interchange and the Virtual Factory’s Needs


The class of software known as groupware addresses some of EDI’s drawbacks and has become popular for building collaborative environments. Groupware applications help coordinate work in three ways. First, they make available a common body of information so that, for example, a salesperson on the road can check the in-stock status of an item for a customer. Second, they track work flows so that group members can—from a remote location—collaborate on documents and projects; all members of a design team, for example, can use proprietary groupware to make sure they are working with the most recent version of a drawing. Finally, the software provides a platform for communication and interactive discussions, from E-mail and bulletin boards to on-screen video. A major advantage of groupware is that all links do not need to be preestablished; authorized users can access and leave the system at will.

On the downside, groupware can be expensive. Each individual user must purchase a copy of the groupware application, and training and administration expenses for the new platform are high. Lotus Notes, for example, costs between $1,000 and $5,000 per user over a three-year period (PC Week, January 30, 1995).

A further drawback is that groupware cannot be used to gain access to remote computers that are not groupware servers. For example, it is not possible to use Lotus Notes to connect to another site’s manufacturing information system to review inventory policies or to use its CAD software to work on the drawings of a part.

In summary, traditional groupware has many of the transmission and data access capabilities that an effective virtual factory needs. In addition, it requires a relatively low level of IT sophistication: people comfortable with PCs, for example, can use it without much difficulty. But groupware lacks adequate telepresence capabilities. Partners cannot use each other’s applications. And because group-ware entails a significant amount of administration and overhead, companies typically will not choose to use it to collaborate until it is clear that their relationship will continue. It will not be used by companies that are at the exploratory stage or anticipating a short relationship. If the community consists of a small number of partners that need to exchange only basic information such as orders, then EDI, because of its relative simplicity, is preferable to groupware. Although groupware is superior to traditional electronic data interchange in filling the requirements of a virtual factory, it nonetheless is far from perfect. (See the diagram “Groupware and the Virtual Factory’s Needs.”)

Groupware and the Virtual Factory’s Needs

Wide-area Networks.

This class of technology provides dedicated high-speed links that connect individual local-area networks. Unlike groupware links, wide-area networks are permanent and usually provide members with all the transmission, data-access, and telepresence capabilities that a real virtual factory requires. They provide universal access to all data and applications resident on members’ local-area networks.

Membership in a wide-area network, however, is exclusive and expensive to obtain. The high-bandwidth telecommunications lines that make up their backbones, for example, typically cost more than $1,000 per month. In addition, administration of a dispersed network is complicated, and a group needs a relatively high degree of IT sophistication to participate. Consequently, wide-area networks usually exist only within a company and are rarely extended to other partners.

Wide-area networks provide access to all applications but are expensive to obtain.

Because of those constraints, even large manufacturers build wide-area networks between only a few of their large sites and exclude smaller sites and smaller partner companies. Wide-area networks fill a tall but narrow slice of the total requirements of a virtual factory and cannot be extended to support companies that are at the cooperative or exploratory stages. (See the diagram “Wide-area Networks and the Virtual Factory’s Needs.”)

Wide-area Networks and the Virtual Factory’s Needs

These limitations mean that proprietary wide-area networks are like an exclusive club. It is very difficult, for example, to use a wide-area network to share CAD data files with potential partners who might be interested in bidding for work. It also is difficult to add a new member to the system quickly to exploit a new opportunity to codesign a product—a severe limitation in this age of rapid product development.

The Information-Brokered Internetwork

Could someone patch together EDI, groupware, and wide-area networks to create a mosaic that would fill the needs of a full-fledged virtual factory? No. Proprietary and disparate standards make such a network extraordinarily expensive, complex, inelegant, and, in the long run, dysfunctional. Moreover, none of the three conventional technologies accommodate exploratory relationships.

But a more flexible and less expensive alternative for carrying out collaborative manufacturing has just emerged—the information-brokered internetwork. The confluence of several trends now makes this approach possible: the emergence of widely accepted, open standards; ever cheaper computing power; increasingly abundant bandwidth; the development of essentially unbreakable computer security; and accumulated expertise.

Open Standards.

Standards among computers are simply agreements about how data should be formatted or transmitted. The most important for the emerging virtual factories are the TCP/IP protocols developed for the Internet, which standardize how dissimilar computers and networks pass data among themselves. The TCP/IP protocols allow the three levels of functionality we have described: transmission, data access, and telepresence. As the unbelievable growth of the Internet attests, these increasingly dominant protocols have been helpful because they are comprehensive, open (published), and nonproprietary (free).

At the same time, standards for text files, spreadsheets, CAD drawings, and other electronic documents that make it easier for companies using different applications to exchange data also have emerged. Consequently, partners who share these standards on how to send information and what it should look like can communicate with confidence that nothing will be lost in the translation.

Cheap, Powerful Computing.

Because of the phenomenal increase in computing power available per dollar, sufficient computing muscle for almost any information-sharing task in a virtual factory is now well within the financial reach of even the smallest companies.

Abundant Bandwidth.

The information-carrying capacity of a communications link, or its bandwidth, has been increasing at least as fast as computing power. For example, standard modems today provide as much bandwidth as the highest-speed links did ten years ago, and dedicated links between two partners now provide enough bandwidth for full-motion video conferences, a very high bandwidth application. And several emerging technologies promise to increase bandwidth dramatically in the near future.


Information security is understandably a major concern for companies seeking to build virtual factories. But despite the continuing tales of break-ins and hacker exploits, tools such as network fire walls—computers that act as bouncers and check all incoming data—provide the means to keep outsiders out and to limit or customize each insider’s access. In addition, essentially unbreakable data-encoding schemes now guarantee that information sent over the Internet or any other network remains unreadable until it reaches its destination.

Accumulated Expertise.

It takes time to master any emerging technology. But companies now are comfortably familiar with current open standards and the new technologies in computing, bandwidth, and security. Indeed, many businesses now have the skills required to innovate with these technologies and therefore the power to construct real virtual factories.

The Brokered Internetwork and the Virtual Factory’s Needs

A network with the attributes listed above elegantly satisfies the demands of a virtual factory across all stages of relationships and all but the most naïve level of IT sophistication. (See the diagram “The Brokered Internetwork and the Virtual Factory’s Needs.”) Transient or prospective partners as well as small companies and unsophisticated IT users will be willing and able to join the community because the costs and overhead of membership are very low. Once all these companies are connected, each one can access a wider pool of computing power and information. For example:

  • A large company can quickly and inexpensively send bid-request packages to a much wider range of potential subcontractors than it could before.
  • People can access other companies’ on-line catalogs, get all required product specifications, and place orders from their desktops.
  • A machine shop can use the three-dimensional modeling software resident on a distant mainframe to place the part it plans to make within a digital mock-up of a larger assembly and check that it fits—before the first one is produced.

These examples represent only a tiny fraction of the total uses that members will find for a virtual factory. Robert Metcalfe, inventor of the widespread Ethernet networking standard, contends that the utility of a computer network increases exponentially, not linearly, as the number of users expands. We are confident that real virtual factories will provide solid support for Metcalfe’s Law.

The Information Broker

Of course, it is one thing for a networked factory to be possible or desirable, and it is quite another to create and maintain it and help it evolve. Designing the system, administering it, updating its technology, maintaining security, and exploiting new opportunities as they arise are collectively a huge job—and one that most companies whose primary business is not IT will be loath to undertake. Indeed, McDonnell Douglas has turned to what we call an information broker to perform this function.

McDonnell Douglas’s spin-off AeroTech developed the aerospace company’s virtual factory and, in its role as information broker, keeps everything running smoothly. AeroTech maintains a database of all users and of all the information to which each is allowed access. It also has developed software packages and training manuals that allow small companies to join the community easily and quickly. For larger, more established partners, AeroTech builds gateways to their existing networks and establishes high-bandwidth links. Finally, AeroTech could eventually convert or translate the various data formats that exist across the virtual factory. For example, if one manufacturer’s CAD data were written in Catia and another’s is in Pro/Engineer, AeroTech could translate the data for them so that they could work together. More important, it could build a system that would perform this type of service automatically.

Potentially, an information broker can provide much more than the computer security, maintenance, and translation functions for the virtual factory. It can help the partners identify which information has value to particular constituents and how revenue might be generated from this information. For example, one company may have test data on the performance characteristics of its valve components, and that data might be tremendously useful to another company in its design process. Or the design department of a particular company might have ready-made CAD drawings of a wide range of electrical connectors and might choose to sell them to a partner that wanted to avoid drawing them from scratch. And a larger company might keep a list of suppliers that have been certified for meeting the ISO-9000 quality standards—a list that could be useful to others looking for such vendors.

Very few IT contractors presently have the kind of relationship that would allow them to explore and exploit opportunities like these jointly. Why? Identifying information-sharing opportunities on the network requires industry knowledge that is both broad and deep. Current systems integrators rarely have such knowledge.

Identifying information-sharing opportunities requires deep knowledge of an industry.

The first generation of internetwork-building information brokers will probably be either spin-offs of larger companies, such as AeroTech, or completely new companies composed of computer networking experts and veterans of the particular industry who have a deep understanding of how the industry works. Real virtual factories are likely to proliferate first in environments where there is a large, dominant partner that can provide the impetus and the funding. McDonnell Douglas Aerospace, for example, had a clear idea of its requirements and was willing initially to assume all the costs of filling them, even though others also would reap benefits.

Eventually, however, many different models for building virtual factories are likely to arise. They may be star-like structures with dominant centers or manufacturing communities in which groups of small manufacturers band together for the same kinds of benefits available to large traditional factories with abundant resources for expanding information technology. The rigid formulation of traditional electronic data interchange will give way to a world of greater fluidity.

Once the benefits of the real virtual factory have been demonstrated, they will create a new manufacturing world. In this new world, companies that insist on remaining loners or that cling to today’s closed, proprietary systems will find it increasingly difficult to survive.


Leave a Reply