Last summer, Kemet Electronics Corp., Greenville, S.C., one of the world's leading manufacturers of tantalum capacitors, a key component in the electronics industry, found itself at the center of a growing shortage of parts that curtailed production at numerous high-tech firms while causing others to redesign their products to get around it. As a result, Kemet moved swiftly to boost capacity, laying plans for an expansion of its Greenville plant. Unfortunately, just days before he was to preside over the groundbreaking ceremony for the expanded facility, Chairman and CEO David E. Maguire got a call from Solectron Corp., signaling the end of the shortage. Says Maguire, "John Chambers [CEO of Cisco Systems Inc.] sees his warehouse is full, he calls Solectron and says stop, and Solectron calls us. It jerks all through the supply chain." Kemet's case illustrates the importance of timely, reliable data to a supply chain. Supply-chain-planning software can help, but any manufacturer that believes software alone can prevent or solve its supply-chain woes should think again. The high-tech industry's experience is a case in point. Last summer dozens of computer and electronics manufacturers, including Dell Computer Corp., Palm Inc., and Agilent Technologies Inc., found themselves in a scramble to purchase tantalum capacitors. "It was like pulling up behind a bus at a fast-food restaurant and finding yourself at the end of a long line," says Ian Ross, Agilent's vice president and general manager for worldwide order fulfillment. In an effort to move up the line, Palo Alto, Calif.-based Agilent sent top executives to the offices of suppliers. "That high-level interaction makes a big difference," admits Ross. It also found new sources, but sometimes at a cost of 100 times what it normally would pay. And when it failed to find enough tantalum capacitors and other components, Agilent and other firms slowed production, changed product design, and blamed parts shortages for lower-than-expected sales. The market punishes companies that suffer from supply-chain woes. Last July when Agilent warned that a lack of parts would hurt earnings, its stock fell by 35%. In February when Nike Inc. said a computer glitch would damage its sales, the footwear company's share price dropped 19.5%. And the damage still is being felt. Cisco Systems, San Jose, responded to last summer's parts shortage by making commitments to buy and stockpile components months before they were needed. When the high-tech slowdown hit, Cisco was caught holding a big bag of inventory whose value was shrinking. On April 16 the company took a monster $2.5 billion writeoff of its swollen inventory. Crippling Shortage Just about any student of high-tech history could have predicted another shortage of tantalum capacitors -- the only question was when it would occur. Maguire, who has been making these devices since 1959, wasn't surprised. "I've seen about 10 shortages of tantalum capacitors in my career," he counts. The inability to predict the next shortage meant, in effect, that most companies simply did nothing until it was too late. Like a leg on a centipede, a capacitor is just one tiny piece in a supply chain that includes hundreds, even thousands, of parts. The only difference is, a centipede still can crawl with one leg missing, but a Nokia cell phone or a Dell laptop cannot function without that part. Complicating matters was the fact that so many different industries use these capacitors. Finally, with just-in-time manufacturing squeezing inventories, companies have little buffer stock to protect against interruptions in supply. Even those manufacturers that manage to collect all the pieces necessary to make a product in good times can do nothing about changes brought on by unpredictable events such as earthquakes, power outages, labor strikes, or political events such as the Asian financial crisis, which contributed to the deficit in capacitors. During the so-called Asian flu, manufacturers on the Pacific Rim slashed prices for capacitors so low that almost no one could afford to build new factories to make them. Then, as demand for cell phones, palm devices, PCs, and other electronics gear soared last year, suppliers including Kemet were flooded with orders. To guarantee parts, manufacturers such as Kemet forged long-term supply agreements during last year's shortage. Now that demand has slowed, these bonds are breaking apart. "The capacity's sitting there, but the customer can't take the parts because he can't afford them," Maguire says. Often the problems stem from bad or insufficient data. Forecasts wallow in the sales department, for instance, and then dribble on to the factory floor. In other cases, customers fail to provide their suppliers with usable information. Even in the high-tech business, known for its expertise in managing bits and bytes, suppliers view customer data with suspicion. "Forecasting information is really important, but it is never right," observes Ross. Long leadtimes for critical components are expensive, so companies try to buffer assembly operations with spare parts they know could run short. A PC manufacturer, for instance, might stock up on motherboards, skewing the demand data for that part. When the motherboard manufacturer obtains weekly or monthly forecasts, it doesn't know whether or not it includes the stockpiles, and that leads to second guessing. "We just don't have enough information. No one trusts the supply chain," summarizes Maguire. Nor is high tech alone in its supply-chain suffering. On April 18 automotive supplier TRW Inc., Cleveland, said erratic scheduling of production at automotive manufacturers, including last-minute production cutbacks, forced the company to incur unanticipated costs due to excess inventory and labor. Sales in the automotive-parts division slipped 13% in the first quarter to $2.6 billion. Bullwhip Effect: When information about customer demand reaches a supplier after that cycle is already changing, it's called the bullwhip effect, a term coined by Big Steel. The giant integrated mills sit several links removed from the end customer. As such, they often learn too late that consumers are buying fewer Ford Tauruses or steel cans for soup. Laurence Cox, coordinating director of commercial and planning systems at National Steel Corp., Mishawaka, Ind., is in charge of fixing the integrated steelmaker's supply-chain troubles. He took on National Steel's order-entry system back in 1996. Every National Steel order has 150 to 250 parameters, and Cox found 81% had an error. A preponderance of incorrect orders can skew demand. Cox pulled together a team of 10 employees to fix the problem, setting down two rules: Every order must be accurate, and if there's concern that it's not, delay the order. Cox gave team members more responsibility. Each would maintain information about capacity on one product line, and would work directly with factory managers to obtain the latest manufacturing data. Only those 10 would be allowed to enter orders, and they would decide who -- when competition sprung up among salespeople for a hot product in limited quantity -- warranted it. The group also suggested ways to improve the process. One person recommended using an e-mail program to search out overnight problems in orders. When it found a glitch, the application would e-mail the employee responsible for fixing it. A year after Cox discovered the clog, complaints from the field-sales staff had dropped by 75%. Cox swung into action again last year, two years after the $2.9 billion steelmaker purchased a software package, made by i2 Technologies Inc., Dallas, to improve production scheduling and reduce inventory. By last December, National Steel had failed to achieve expected results. Cox found the culprit: incorrect information. Complex software refuses to crunch it. "i2 forces you at gunpoint to clean up your data," he says. Most of the bad data came from the order-status system, which covers the amount of steel ordered, made, and shipped. The auto industry is a large customer, and it demands defect-free steel. To ensure that its products meet expectations, National Steel makes as much as 25% more than the customer orders. Complications arise when the customer either rejects less than 25% of the order, or more than that amount, creating either a surplus or a shortage of finished materials. In either case, National Steel scrambles to find another customer to buy the extra stock, or to drum up new sources to fill the shortage. Worse, each plant records its order status differently. Cox assembled an i2 "rejuvenation team" of four. The group includes an information-systems specialist with experience in customer service, a veteran in steel production planning steeped in the quirks of the order-status system, and an engineer who worked on information technology at National Steel's mills. Each member served as an operator between departments to figure out where the information bottlenecked. Then they figured out how to clear the traffic. The group established standards for accounting for excess steel. It also found ways for the plants to provide customer-service staff with timely figures on excess steel or capacity. National Steel soon will learn how well these amendments work as it begins rolling out the i2 system this spring. Integration Woes: Another trick to making supply-chain software work is to make sure it integrates well with the other systems in the company. Beaverton, Oreg.-based Nike learned this lesson the hard way, as its common shares shed $2.61 billion in market value following revelations that difficulties of implementing a new demand- and inventory-management system led to product interruptions. Simply put, the company ended up ordering the wrong shoes -- as many as 5 million pairs of them. These troubles contributed to the company's drop in earnings by one-third to $97.4 million in the quarter ending in February, compared with the same period last year. The sneaker manufacturer is revealing little about the glitch. "We're not going into a lot of detail about what those issues and what specific technical aspects of our implementation were," says Leslye Mundy, a Nike spokesperson. But interviews with analysts, former employees, and systems integrators who worked on Nike's supply chain offer a glimpse at what happened. In the mid-1980s, as Nike neared the $1 billion sales mark, it built its original demand-management system. Close to 100 information specialists created the custom-made application to enable the company to run its futures system, whereby orders from retailers are placed six months ahead of delivery. Those orders in turn are sent to contract manufacturers in Indonesia, Vietnam, and China. As Nike grew, everything became more complicated. Its product line mushroomed. It contracted with new factories. By 2000 sales hit $8.9 billion, and the number of manufacturing steps alone to make a complex sneaker reached between 128 and 136, estimates John Shanley, a senior analyst at Wells Fargo Van Kasper, San Francisco. Over the years, programmers adjusted Nike's custom-made software to accommodate changes in the growing business. "It's been modified thousands and thousands of times. These little arcane changes had to create serious problems as Nike moved to a whole new system," says one former employee. Nike began implementing new supply-chain software made by i2 in March 1999. The package accounted for roughly $40 million of an overall Nike investment totaling $400 million for various other software including applications from Siebel Systems Inc. and SAP AG. As Nike installed the i2 product, it customized the code to accommodate the eccentricities of its original system. These modifications slowed the new application so that it took as long as two minutes to pull up a screen. Meanwhile, employee turnover hurt the project. The CIO involved in the decision to initiate the supply-chain renovation left the company. As Nike replaced the old system with the new one, it sent flawed data to contract manufacturers in Asia. Some received double orders for the same shoes, while Nike failed to order enough of other sneaker models. Observers believe both Nike and i2 took large steps too fast. Rigorous system testing would have revealed the glitches, they say. The sneaker maker went live with the software to thousands of suppliers and distributors simultaneously -- one sure way to stumble. Complicating matters, Nike chose to adapt the i2 software to fit its own special needs, instead of using the firm's standard settings for apparel companies. Executives at i2 declined to comment. Says spokesperson Beth Elkin, "We feel it's been thoroughly covered and it's behind us." For Nike the supply-chain headache remains at the triple-Excedrin level. "We've identified what those problems are and we're working to resolve them by the end of the calendar year," says Nike spokesperson Mundy. To mend its immediate supply-chain woes, the world's largest shoe company has sold excess sneakers to large customers at a discount, and put missed models into production. In fairness to i2 Technologies, its advanced-planning-and-scheduling and supply-chain solutions have been installed at numerous companies without a serious glitch, and i2 claims the software has saved its customers many billions of dollars. One thing is certain. Before this year is out, another major supply-chain headache will emerge to afflict a major manufacturer or manufacturing industry. Software can help companies avoid the problems, but not without a smart manager capable of ensuring the accuracy of data and willing to make fast decisions when supply goes awry. As Agilent's Ross puts it, "Software helps with the processes, but if you look at the companies that did well during the parts shortage last year, it's excellent judgment that drove their success, not blind luck."