In the early 1980s, IBM faced a critical problem. The company had installed its mainframe computers at customers’ facilities all over the world, and it needed to manage its inventory of computer parts in a way that would let engineers fix malfunctioning machines as rapidly as possible. That was a mind-boggling logistical exercise. It involved hundreds of thousands of parts stocked in thousands of locations worldwide.

Even as IBM executives wrestled with the issue, Morris Cohen, a Wharton professor of operations and information management, stepped up to help. As the principal scientist working on the problem, he and a group of colleagues from Wharton helped IBM develop Optimizer, a decision-support system that let the computer giant map out a global parts supply chain. Result: IBM was able to cut inventory investment by $250 million, while reducing annual operating costs by 10% and increasing the level of customer service.

IBM was hardly an isolated case. During the past two decades or more, Cohen has continued to study how companies use supply chains to support after-sales service operations in industries ranging from computers to automobiles. In the process, he says, he has learned fundamental lessons about the way such supply chains should be set up and managed. Among them: It would be a mistake for a company to set up a parts supply chain in the same way that it does its production supply chain. “These logistics problems are very different than planning production,” Cohen explains. “They are like high-stakes gambling problems, or like portfolio management problems. You have to solve them by understanding the risk tradeoffs involved.”

What exactly does that mean for companies? According to Cohen, one implication is that parts supply chains must be designed in a way that takes into account the criticality of a company’s products, and the cost to the consumer if the product fails. Consider, for example, a component in a computer system used by air traffic controllers. If the air-traffic control system were to go on the blink, the results could be devastating. Ideally, the system should be repaired in minutes, if not seconds. Similarly, a stalled machine in a semiconductor fabrication plant could bring the entire manufacturing process to a standstill. Cohen explains that components for such critical products or services must be served by supply chains that differ dramatically from parts supply chains for non-critical products. Example: if a home hair dryer doesn’t work, the worst result for the user would probably be a bad hair day.

Another key issue, according to Cohen, is determining the locations in the supply chain where a company should stock the parts. Should parts be stocked in a centralized fashion—say, a single warehouse or a small number of central warehouses? Or should the company have a far-flung network with multiple stocking points? The answer, Cohen notes, depends on the criticality factor. In the case of computer systems for air traffic controllers, it makes sense to have critical components stocked over a wide network, so that plenty of points in the supply chain can back up one another when the parts are urgently needed. In contrast, a busted hair dryer could probably wait an extra day for parts to be shipped from a distant central warehouse or disposed of altogether by providing the customer with a replacement product.

Cohen, who has overseen research by several Ph.D. students over the years on such issues, developed many of these principles while studying parts availability in the automobile industry. Between 1985 and 1987, Cohen and his colleagues worked with General Motors to study the company’s parts supply chain network—at a time when the company was setting up Saturn as an independent firm. Cohen says that many of the recommendations that he and his colleagues made showed up as part of Saturn’s service-support system.

Today Saturn has the highest off-the-shelf availability rate for parts of any car maker, according to Parts Monitor, a trade publication. In a paper published this year in the Sloan Management Review, Cohen and three colleagues—two of whom worked for GM and Saturn—point out that Saturn’s performance was based on two key factors: Matching the company’s supply chain strategy to the criticality of its customers’ needs, and involving dealers (or “retailers”) in its supply chain strategy.

Where is parts supply-chain management headed in the future? Cohen believes that with the coming of e-commerce, the field is moving towards a revolution. The main reason is that “technology has reached a point where large-scale optimization has become possible in real time,” Cohen says. As a result of the Internet, it has now become possible for companies to create vast networks connecting their operations with those of their suppliers and customers. In the past, the hurdles to setting up such networks were so enormous that companies developed their own proprietary supply chains at enormous costs in time and money. As web-based technology takes off, however, the process of setting up such supply chains will become both faster and less expensive.

Cohen believes that it is now possible for him to commercialize some of his research. With a view to doing that he last year launched MCA Solutions, an application service provider that offers supply chain services to clients. According to William C. Ross, the company’s director of software engineering, the company’s products give users a global bird’s eye view of optimal inventory.

In a recent article written for the Financial Times Mastering Management series, Cohen and Vipul Agrawal, chief operating officer of MCA Solutions and a former professor at the Stern business school in New York City, explain how e-commerce will transform supply-chain management. Cohen and Agrawal believe that as companies’ access to sources of supply increases as a result of web-based exchanges, “the dream of always providing the right product to the right customer at the right time and place and at the right price will very likely become a reality.”

A good example of a company that is using the Internet to manage its supply chain is Dell Computer. According to Cohen and Agrawal, the company’s web site allows customers to specify the configurations of their computers. This means that Dell can procure and assemble components at lower cost and with shorter lead times. Dell’s supplier network can provide components to the company’s assembly plant in Austin within hours of an order being placed. Dell’s sales over the Internet now amount to more than $30 million a day.

Over time, more companies will migrate to such web-based supply chain management systems. This may not happen as rapidly as some people hope, considering the setbacks this year to the Internet economy, but the supply chain revolution is on its way—and there’s no turning back.