(FORTUNE Magazine) – There isn't a manufacturing manager alive who doesn't wake occasionally at 4 A.M. and wonder what else could be done to cut unnecessary links or trim costs in the company's supply chain. Not becoming a bit obsessed about it is tough, given the amount of attention supply-chain management garners these days at the "C" level--plus the media coverage, the consultants' harping, and the clamoring of software vendors who claim to have the cure for whatever's rusting or about to crack.

An old-timer might contend that this is nothing new, that supply chain is just a buzz phrase for the "pipeline" or, more broadly, the "business." Whatever the name, there's plenty to worry about in an era of unprecedented revamping and tweaking to yield more profit. While the primary job belongs to just-before-dawn insomniacs, help is available. Consultants really can provide good advice. And some impressive potential partners are standing by to take over entire sections of the chain, letting companies focus on what they do best, which these days can come down to no more than product concept and marketing.

On the following pages, FORTUNE visits four companies, starting with Caterpillar. To figure out how to distribute a line of construction machines aimed at what is, for Cat, a new market populated by very impatient customers, the company asked a university's operations research professors to find answers using arcane algorithms and high-speed computers. The calculating resulted in a speedy, inventory-saving delivery system.

Two other companies make a business of taking over parts of other companies' operations. One is UPS, whose rapidly expanding logistics group in Atlanta was hired to help shorten the time between the roll-off of a new Ford at the factory and the drive-off by a customer at a dealer near you. The other is Solectron, now the largest infotech contract manufacturer, which is willing to do everything from product design to warranty service but has, so far, carefully blocked off activities that it says are the customers' jobs.

Finally, as a reminder that bringing in consultants and outsourcing parts aren't the only ways to win, there's an outlier: Mothers Work, a Philadelphia company almost fully integrated from raw material to retail. Here, a husband-and-wife team dominate a niche, maternity clothes, with a system for overnight stock replenishment of some 700 company stores that keeps inventories down and profits up. And that, of course, is what supply-chain management is supposed to be about, whether the product is an earthmoving machine or a pair of jeans with lots of room to adjust the waist.

CATERPILLAR Letting Computers Map Distribution

Talk about ambitious goals. Caterpillar, already the world's largest builder of construction equipment, says that by "mid-decade" it intends to get to $30 billion in annual revenues, up about 50% from where it is now. Given the dominance of its big yellow machines in its market--and no indication of spurts to come in heavy construction--it's not surprising that the company says it will need new products and new markets to pull that off. Nevertheless, a completely new line of equipment launched at the end of 1998 seems, at first, an odd move. It's not that the new machines don't bore, dig, shovel, and move dirt. They do. But these are "compacts." Parked alongside a regular Cat bulldozer, grader, or backhoe, they look like toys.

They also posed daunting supply-chain questions. The customers for compacts are not like buyers of big Cat machines, who may know a year in advance that they will be needing, say, another highway grader. Users of the compacts are in a hurry. They tend to go elsewhere if they can't get the machine they want within days of walking into a store or rental yard. This raised a classic distribution challenge: how to avoid lost sales without incurring the cost of carrying extra inventory. Helped by a university professor's computer modeling, Caterpillar devised a clever solution that avoids both pitfalls.

With its new machines, Cat is pushing into what it says is a $4 billion building-contractor equipment market growing about 10% a year. The company's plan is to get about $1 billion of that by 2006. However, this is unfamiliar territory for Caterpillar. Most of the customers are, for the company, small-timers--homebuilding contractors, landscapers, and the like. There's well-entrenched competition: Ingersoll-Rand's Bobcat construction machines probably have at least half the market. And with price tags as low as $25,000, Big Yellow is trying to sell high volumes of standardized, complete compacts for less than the cost of just one replacement tire for its customized, low-volume $3 million mining trucks.

The new compact line includes three different machines, each available in several models with slight variations in dimensions as well as in engine and hydraulic power. Two, for which Europe is the best market, are made in Caterpillar's plant in Leicester, England: a rubber-tracked excavator, which looks like a helicopter bubble with attached ditch digger, and a scaled-down front-wheel loader. Loaders get their name because, in their most common configuration, they are equipped with hydraulically operated buckets to pick up and move material.

The third machine in the compact line is a smaller loader assembled in a Sanford, N.C., plant that Caterpillar bought, gutted, and rebuilt. The machine uses engines from Britain, which it shares with the rest of the line, and welded-up parts and subassemblies from a new Caterpillar plant in Torreon, Mexico. And it's a "skid steer" like the Bobcat at which it's aimed. Unlike Cat's British loader, which has steerable front wheels, skid-steer vehicles change course by braking the two wheels on one side while powering the other pair. Thanks to Cat's control system, its skid-steer loader is easier to steer and harder to stall than it might seem.

Worldwide sales of the skid-steer loader are expected to be higher than those of the two new British machines combined. Now being assembled at a rate of 40 a day, it's already the highest-volume piece of equipment Caterpillar makes. Next year the production rate goes up another 20%.

While all three machines can be fitted with different accessories or tools, the skid steer is extraordinarily versatile. As substitutes for its basic bucket, Cat offers 52 interchangeable attachments, or what it calls "work tools." Besides several different types of buckets, the collection includes augers, brooms, pallet forks, hammers, trenchers, and a landscaper's delight that clears a home construction site of rocks and debris while aerating and leveling the soil, leaving it ready for grass seed. In time, the average contractor will likely need several different tools. To capture customers even before they need a new loader, Cat has designed all its skid-steer tools to fit on the rival Bobcat. Conversely, if a contractor needs a new machine, he can buy a Cat and still use his old Bobcat tools.

The tool strategy, and indeed the whole compact line, fits neatly with another new Caterpillar venture: construction-equipment rental yards owned by Cat dealers here and abroad. Since the rental-yard program started in 1998, 61 U.S. dealers out of a total of 70 have opened a combined total of 280 rental facilities, stocking them not only with the compact machines but also with some intermediate-sized, Cat-built construction machines and other contractor paraphernalia such as lights and telescopic lifts from other manufacturers. So far about 35% of the compacts have been delivered to dealers for use in their rental yards; the rest have been sold to users.

The trick has been to set up rapid-response distribution. In an early survey, dealers judged that about nine out of ten potential purchasers or renters of work tools would buy a competing make if they couldn't get what they wanted the same day they saw it in the store or rental facility. The dealers also judged that a huge percentage of potential users of the compact machines on which the tools fit wouldn't wait much longer. (Customers in the Northeast are the most impatient.) To give customers what they want almost as soon as they ask for it, Cat at one point thought it might have to consign equipment to dealers and bear the inventory costs itself or give the dealers financial help to carry more inventory than they might otherwise want.

As with any product, a key factor is "time in channel." The amount of inventory needed in the system varies with the delay between order and delivery. The faster the replenishment of sold items, the less inventory needed. But calculating how to set up a rapid response system for 15 models of compact machines and 102 tools was a tough problem. First, there are 207 Caterpillar dealers in 90 countries to which deliveries have to be made. Next, the machines are manufactured on both sides of the Atlantic, and the work tools are sourced from Cat plants and third parties in the U.S., Britain, Finland, Hungary, and Mexico. The Torreon plant, for example, welds up all the less complicated skid-steer tools and then ships them up to Sanford to be completed and painted.

A solution that quickly came to mind was to distribute the new line through Caterpillar's parts distribution centers, or PDCs. No slouch at fast parts service, Cat has 22 PDCs around the world, charged with getting a part to a user of the big yellow machines in less than 48 hours, no matter where. Fail that and the part's free. Indeed, Cat stocks and ships parts for other companies, including DaimlerChrysler and Honeywell Aerospace.

The PDC managers figured that handling compact machines and tools would be a breeze, but there was a snag. The PDCs typically tack on a fee based on a flat percentage of the selling price. However, the building-construction-products division, which is responsible for the compacts as well as for an intermediate line of machines, operates on narrower margins than the rest of Caterpillar. For products like these, it said, it couldn't afford the PDC sales fees.

Still, recalls Robert Briggs, now general manager of the Sanford plant and then a member of the team developing the compact business, "we knew we could have a real mess on our hands if we always had the wrong machine or work tool in the wrong place at the wrong time." Looking for help, the building-construction-products division turned to Pittsburgh's Carnegie-Mellon University and a team led by Sridhar Tayur, a 35-year-old Indian engineer with a doctorate from Cornell who's a professor in the university's Graduate School of Industrial Management.

Tayur is an expert in operations research and has used that discipline of arcane mathematics and data-dense computer programs to create supply-chain models. He has also had plenty of nonacademic experience working for GE, Intel, and McKinsey. Caterpillar asked him to help devise a distribution method for the new line and determine appropriate inventory levels for the U.S., the most important market.

Cat supplied basic facts, including required factory lead times, transportation costs, numbers of tools or machines per truckload, and, of course, sales forecasts, including anticipated fluctuation in demand. It also laid out the range of options: direct delivery from all the factories to all the dealers, use of the PDCs, the construction of new, dedicated tool-distribution centers, separate channels for machines and tools, or a mix of any and all.

Using the standard techniques of his discipline, Tayur simplified the problem in part by decomposition, that is, by breaking it down into subproblems. He also used a method called aggregation, creating a "typical" dealer for each of 17 U.S. sales regions rather than plotting routes to all dealers. This not only reduced the number of calculations needed but also, by averaging the sales forecasts of a group of dealers--offsetting the optimists with the pessimists--probably resulted in an even more accurate estimate of demand.

Before the study started, Caterpillar had concluded that it would ship the machines themselves direct to dealers from the two factories. That left the questions of the best routing for work tools and the appropriate inventory levels for both machines and tools. Early on, however, Tayur got Cat to agree to consider a two-level solution for distributing work tools: one level with the lowest possible costs, the other for maximum speed. Until inventories dropped below a given minimum, orders would move through the slower, lower-cost route. If inventories fell below that fixed point, an additional order would be channeled through a faster route to bring them back to the minimum.

As if there weren't enough complications, Cat also wanted to anticipate change. What, it asked, would be the right channel design and the right level of inventories for 2000, the first full year of production, and 2004, by which time it figured it would be operating more or less normally? By then, for example, Torreon will be finishing and painting work tools, with no need to relay them through Sanford.

Among the techniques the Carnegie-Mellon group used to attack this complex problem was so-called infinitesimal perturbation analysis, for which no complete explanation is possible for the faint-hearted or mathematically disadvantaged. Very simply put, it's sophisticated trial and error. The technique involves first the creation of an algorithm and computer program that provides a solution to a problem but not the optimal answer, partly because it's based on a mix of fact and educated guesses.

One of those guesses might be an estimate of the number of buckets needed on hand at a given dealer. The team then slightly changes a guess, possibly increasing a number, and lets the computer try again, recalculating the possible paths for a single tool from its source to each of the typical dealers. If that seems to move things in the right direction--that is, by speeding up transit time or cutting costs or both--the assumptions are nudged upward a bit more. If not, they are nudged down. The computer whirs again, and so on. It's a bit like artillery bracketing a target.

"Even after you get your answer," says Alan Scheller-Wolf, an assistant professor on the Carnegie-Mellon team, "you have to ask, 'What if my input data are wrong?' " Does this approach work if, say, Caterpillar's estimate of demand is faulty? To find out, the proposed arrangement was tested to see how it would work if the forecast were 20% wrong in either direction. Cat figured that intense competition ruled out a larger-than-20% underestimate of demand, and that a greater-than-20% overestimate of demand would require rethinking the whole program. In between, it was finally determined, shifts in demand didn't change the answers.

The data crunching, as Scheller-Wolf puts it, "is not something you do on your laptop." Used instead were Sun Microsystems SPARCstation 20 workstations. One of those could route one tool from one factory to one dealer in about 40 seconds. But each time an assumption or estimate was changed, plotting optimum inventories and a route for everything to everybody for both 2000 and 2004 required 9.8 hours on one workstation, or close to five hours on two working in parallel.

When all the analysis was concluded, Caterpillar had its inventory targets for machines and tools as well as a distribution plan for the tools. The tools would be routed from where they were manufactured through two Caterpillar distribution centers--at Morton, Ill., and Grimhausen, Belgium--and then on to a limited number of PDCs. These include four in the U.S.: Miami, Denver, York, Pa., and one near Morton. When that looked like the answer, Cat persuaded the PDC managers to drop their usual percentage fees. Instead, they would charge based on so-called activity-based accounting, or the actual expense of handling a tool, which turned out to be less.

Caterpillar's Briggs says the Carnegie-Mellon solutions are not what Cat would have come up with on its own. A couple of special tool-distribution centers, which the company had planned to build, were found unnecessary. Just as important, the response time in the system was sufficiently fast that the inventories that the dealers would have to carry were not high enough to require a subsidy from Caterpillar. Moreover, there was no need for a two-level distribution system: The rare emergency has been covered by using air freight.

"Carnegie gave us the highest response, lowest cost, lowest inventory," Briggs says. Since then, the building-construction-products division has been paring away further at logistics costs. For instance, it has been figuring out how to put more tools on a truck and how to get "free" ocean passage for tools by carefully tucking them into container loads of machines.

By late summer of this year about 15,000 compact Cats and some 30,000 work tools had been sold. Revenues in the first year, 1999, were $150 million. Cat forecasts better than $300 million this year. Some of that has to reflect the filling of the supply pipeline. But getting up to the $1 billion target in a few years no longer seems overambitious. Says Don Ings, head of the division, "CAT's strengths are its distribution system and its dealers. Getting 70 U.S. dealers fully knowledgeable and ready for a brand-new product is a big task. But once you do get them ready, it's like a big truck going downhill."

UPS LOGISTICS Delivering Solutions, Not Packages

After Ford and a unit of UPS announced a logistics deal last February, recalls Denny Barts, an executive who runs the new UPS program for the automaker, "Friends asked whether I was going to have people in brown shirts delivering Fords to customers." Barts quickly pointed out that he works for UPS Logistics Group, not the parent, and that the group's business is not primarily delivering things but providing experience, information systems, and services to help cure other companies' shipping and customer-service headaches.

For Ford, UPS Logistics promised to cut the number of days between the moment a car or truck rolls out of an assembly plant and the moment it comes off the "haul-away," or car-carrier truck, at a dealer. Right now that takes close to two weeks on average, and sometimes your just-ordered "performance red" Mustang can get mislaid--actually lost on some lot somewhere for days. The longer it's out there and the more it passes from truck to rail to truck, the greater the chance it will arrive with a busted side-view mirror or a scratch on the trunk lid.

UPS formed the logistics group in 1995 by bringing together a number of small operations it had started up or acquired earlier in the '90s. Their common mission was to make money by helping companies cover the gap between the end of the production line and the dealership or the ultimate consumer. In its first year, the group's revenues were about $350 million. This year Logistics will contribute something like $1 billion in revenues to its $27-billion-a-year parent. If nothing else, the demand indicates that there's a lot involved in making this segment of the supply chain work smoothly, and that quite a few companies would just as soon turn over the job to somebody else.

The UPS Logistics group trades on the parent's reputation for reliability and international coverage. Boasts Dan DiMaggio, the group's CEO: "It doesn't matter if it's India, Israel, or Iceland, we've got somebody just about everywhere doing something." DiMaggio's mission isn't to fill the brown trucks, though sometimes they are part of the solution. Much of what the group moves travels via common carriers--mostly trucking lines--because the load exceeds UPS size and weight limits, or because the parcel system is too expensive for deliveries that don't have to be there in a day or so.

Headquartered in Atlanta, UPS Logistics now has facilities around the world. Some of the group's growth has come from acquisitions. This year alone, to fill out its network internationally, it has bought companies that expand its coverage of Australia, Asia, Latin America, and parts of Europe. For some customers, UPS Logistics does no more than provide computer-based systems they use for, say, routing and truck loading. For others it operates dedicated fleets delivering everything from perishables to young pigs, both of which require trucks with special equipment so that nothing meets an untimely end. For Papa John's Pizza, based in Louisville, the group delivers supplies to the chain's stores, following schedules based partly on how fast pizza dough rises during transit. In Dallas, UPS Logistics takes deliveries of imported electronic components such as capacitors, stocks them, and arranges to ship them on to assemblers throughout the U.S. In Singapore it collects and stores integrated circuits from National Semiconductor's nearby plant and then sees that they are relayed on, filling the company's orders from customers everywhere.

The group's largest facility is in Louisville, hard by the airport that is the major hub for UPS's package delivery service. Since 1996, UPS Logistics has built five huge warehouses with a total of 1.4 million square feet down one side of a private street. Now going up on the other side are another five buildings that will have a similar amount of space under roof.

A lot of this space is used for so-called service parts logistics. The UPS group maintains parts warehouses for clients and then, as needed, gets the parts delivered to the client's customers or technicians within hours, or if there's less urgency, within days. One Louisville building, empty at the start of the year but up and running by March, is dedicated to parts service for Compaq Computer and its acquired brands, Tandem and Digital Equipment. It's the first time that parts for all Compaq companies have been gathered in one spot.

UPS Logistics takes care of the full order-completion process for several companies, receiving and stocking the inventory and then picking, packing, and shipping to the customer. For Nike.com, the Swoosh's e-business, Louisville maintains the inventory of shoes and clothing, ships it to customers as directed by Nike--usually in the brown trucks--and then mans the phones as "Nike" in the event the shoes don't fit and have to be returned for credit or exchange.

For Nike, UPS handles only Internet orders. For others, like Basketball Marketing of Paoli, Pa., it's the sole warehouse and order-fulfillment department. Louisville receives the company's shipments of athletic clothes and shoes from its mostly Asian suppliers, then fills orders from sporting-goods retailers. Best known for its brand name, And 1, the company was started seven years ago by two pals who had played basketball together in high school before going on to Stanford and Wharton. As they saw it, the offerings of companies like Nike weren't cool enough or funky enough for the millions of young Americans hyped on street and playground basketball. "And 1" is the call when a player is fouled while making a basket and thus gets points for the basket and one foul shot.

In 1998, when annual sales were on the way to $50 million, And 1's unaffiliated warehouse company said it would store the stuff but could no longer be bothered shipping it to stores. Christina Houlahan, the company's chief operating officer, sat down to make a deal with UPS Logistics from what she admits was not the best negotiating position: Trucks carrying her entire inventory were circling the UPS site in Louisville. UPS said "No problem." They would receive it, stock it, and ship it out whenever And 1 put the ball in play. This year, revenues are running at a $150 million clip, with all the merchandise shipped out of the UPS warehouse. Houlahan says the folks at UPS Logistics were "critical partners" in that tripling of sales. "Most of the time they do phenomenal things for us. They're really cool."

For several high-tech companies, the group's Louisville operations go beyond just handling parts. They include repairing equipment in an "end of runway" shop that is part of the warehouse complex. For parent UPS, the group maintains those electronic boxes called DIADs--for "delivery information acquisition device"--that drivers hand you so that you can scribble your signature to acknowledge receipt of a package. For Getronics, it stocks and fixes equipment that the firm's technicians use to keep point-of-sale equipment running for McDonald's. For you, if your Sony or Toshiba laptop goes on the fritz, the group picks it up the day you call, flies it to Louisville, figures out what's wrong, replaces the faulty part, and gets it back to you the next day.

For something larger than a laptop, such as a Lexmark printer, the routine's a bit different. If you call Lexmark's tech service and the technician concludes that the problem is not you but the printer, you'll be sent a replacement along with labels to use in UPS-ing the defective one to Louisville. There, UPS Logistics diagnoses the problem. If the necessary repair is relatively simple, like replacing a subassembly, it's done on the spot, and the unit is readied to send on to another customer. If the problem is more serious, the printer goes back to Lexmark's plant in Lexington, Ky. In time, says Terry Licciardi, director of Lexmark's product-service operation, he wants UPS to be able to handle all repairs.

No product that the group handles has more value per unit than Ford's cars and trucks. The sticker price on a well-equipped Lincoln Navigator four-by-four tops $50,000. And no delivery system is quite as complicated. Ford has 19 U.S. and Canadian assembly plants turning out Ford, Mercury, and Lincoln automobiles and trucks that must get to 5,000 North American dealers. First they are hauled by truck or rail to intermodal "consolidation yards," where they are sorted by their destinations, to which they usually move by train. That's why you sometimes notice that vans, cars, and trucks are mixed together when you spot a trainload of them even though you assume that they can't all be from the same factory. When the trains or trucks arrive at 42 "destination ramps" around the continent, haul-away rigs are waiting--or are supposed to be waiting--to take the vehicles on to dealers.

In the past, says Frank Taylor, Ford's logistics VP, the delivery network was "a black hole" in which visibility was lost. Ford knew what went in one end and what was supposed to come out the other, but it didn't know where the vehicles were in between and had only a hope and a promise as to when they would get delivered. The railroads and trucking companies have computer systems that can track the shipments. But there has been no way to tie these together with Ford's systems or, more important, to get information to a dealer who wants to know which cars with what options he'll have on hand to sell next Thursday. "We have had enough problems keeping track of a railcar, let alone an individual vehicle," admits Taylor.

UPS has designed and, shoulder to shoulder with Ford employees, has begun managing a system that can track each vehicle by its VIN, its unique vehicle identification number. Given the unit cost, says DiMaggio, "the tolerance for error is zero, and I'm not kidding." To start the job, his logistics experts first built a large "time and transit matrix"--that is, they plotted out, step by step, the route between any factory and any dealer. Then, working with Ford at the factories and with the railroaders and truckers at the consolidation and destination ramps, they started to put in place a system that not only can locate any vehicle anytime but also is expected to cut delivery times up to 40% and typically by six days.

After starting up in March on the West Coast, the system is now being rolled out across the country by a team of 120 people, 93 from UPS Logistics and the rest from Ford. Taylor says the joint team has already cut the average transit time for all Ford vehicles in transit--including those shipped the old way--by 10%. If nothing else, haul-away truckers now have advance notice of how many trucks they'll need at a destination ramp on any given day.

The system is being extended to dealers, who will be able to track incoming deliveries so that they can, without any use of "probably" or "maybe," tell you that if you'd like your new Taurus station wagon in blue with a CD player, one like that will arrive next Tuesday. Sometime around the end of 2001, individual customers will also be able to track a car or truck that they've ordered from a Ford or Lincoln-Mercury dealer the same way you can track a UPS shipment today. Just go to the Website and enter the VIN. Even more important, you'll be able to count on its arrival in a week or so. For Ford, the biggest gains will be the reduction of inventory in the system. As Taylor says, "You can't sell it if it's on a railcar or a haul-away."

SOLECTRON All Along the Chain Except the Ends

They don't come any bigger in this business. For its fiscal year ended Aug. 25, Solectron Corp., the Milpitas, Calif., contract electronics manufacturer, reported revenues of $14 billion, up a walloping 46% from the prior year and five times its 1996 results. Only a couple of years back, Solectron seemed boxed into the No. 2 position in its industry behind SCI Corp. Now it's more than half again as big.

With 115 factories at 44 sites in 16 countries, Solectron churns out printed circuitboard assemblies like so many potato chips. More and more it's also the name behind the name on computer and telecommunications equipment marketed by such Grade A companies as Cisco Systems, Hewlett-Packard, IBM, Mitsubishi, NCR, Nortel, and Sun Microsystems. But this is growth on a Saturn rocket path. Can it continue? What happens if the U.S. economy or just the infotech sector hits a bump? And why does Solectron need anybody else, given all it now knows about how to design, build, ship, and service stuff like laptop computers? If you already operate along most of the supply chain, at what point do you decide to compete with your customers?

Like most of its peers, Solectron has expanded in large part through acquisitions, often closed with bundles of common stock. Some are simply capacity expansions. In its just-ended fiscal year, the company bought Bluegum, Australia's largest contract electronics manufacturer, or to use the term the industry prefers these days, electronic manufacturing services company, or EMS. The term emphasizes that the business is not all manufacturing. Solectron also spent $2 billion to purchase SMART Modular Technologies, which builds memory modules, cards, and subassemblies using purchased memory chips.

In much of its acquisition activity, Solectron has been buying sales by making deals with computer and telecom equipment makers. For these original equipment manufacturers, or OEMs, the deal Solectron offers is this: We pay you a tidy sum for your factories, and you give us a multiyear contract to build what you used to make for yourself. In fiscal 2000, Solectron took over an IBM plant in Brazil, acquired Ericsson factories in France and Sweden, bought Alcatel facilities in Texas and Puerto Rico, and in a huge transaction, picked up Nortel manufacturing assets in Canada, France, Northern Ireland, North Carolina, Mexico, Turkey, and Wales. For the fat Nortel package, it paid $900 million--and got back a promise of what could total $10 billion in orders over four years. Altogether, acquisitions accounted for 17% of its '99 increase in sales over '98, and 41% of the revenue jump from '99 to 2000.

Solectron is hardly alone in buying OEM assets. Among the EMS leaders, Jabil hangs back a bit, but the others--SCI, Flextronics, and Celestica--have been acquiring with abandon. You might think there will soon be nothing left to pick up, but that's not the case. Technology Forecasters in Alameda, Calif., estimates that total EMS industry revenues this year will be $101 billion, which it figures is only a smidgen over 13% of the available market--the combined cost of goods sold by all the world's electronics makers.

"We've hardly begun," says Susan Wang, Solectron's CFO and one not given to overstatement. Wang figures that because the market is so untapped--and because "it's so easy for people to agree that outsourcing is the lower-cost model"--the company could keep growing even in a recession. "OEMs could say that as long as they are laying off people, as long as they are going to cut, 'We're going to cut deeper and outsource this activity.' " Moreover, in good times or bad, the Ciscos of tomorrow are much more likely to let others do the manufacturing.

Even in the worst imaginable broad market decline, Wang believes, Solectron could get by. It uses a lot of temporary help that can be laid off. Production is divided among relatively small factory buildings that can be shut down. There would still be rent or its equivalent to pay and equipment to be depreciated, but that's only about 5% of Solectron's costs. "I realize that 5% is significant," Wang says, "but we've done the modeling and are convinced that we could sustain a 25% reduction in our business and still be profitable--if we do the right things."

Most interesting, if things got bad, Solectron's U.S. plants--its big sites are in Milpitas and Austin, Texas--wouldn't necessarily be the first to suffer. Says Wang: "The cost structure is not the lowest, but these plants are very effective, so they're actually quite competitive" with the company's plants abroad. As she points out, time can be everything, and the U.S. plants have the skill and experience to deliver quickly. "If our customers miss the market window, they lose market share bigtime. That hurts a lot more than paying us a few more pennies for assembly."

One thing feeding optimism is that Solectron has been broadening its base. While printed circuitboard assemblies still account for about 70% of revenues, the company has been expanding "box build," or assembly of complete products such as personal computers or telecom equipment. At the same time it is adding services to manufacturing, something that has been going on for a couple of years but tends to get overlooked in all the noise about the big deals in which assets and orders are traded. In June, Solectron opened a "new-product introduction center" in Scotland that will, among other activities, provide "early prototyping and new-product design." Last year it acquired a laptop repair business. Early this year it bought another company doing much the same thing for wireless handsets. And part of what it got from Nortel was electronic computer-aided design capabilities in five U.S. and Canadian locations as well as a product repair operation in Wales.

The goal obviously is to prototype it, build it, ship it, fix it, and upgrade it. That doesn't leave a lot of room for activity on either end by customers. But small as the remaining room is, Solectron hasn't entered it. Koichi Nishimura, its near-legendary 62-year-old CEO, says it won't. Solectron, says Ko, as he is known to one and all, is "a supply-chain facilitator" that will leave innovative design on one end and demand creation on the other to its customers. "There are very important boundaries here," Ko says. "One of the things we're not going to do is compete with our customers." But then he pauses for a beat and adds, "At least for now."

However, if there is anything certain in infotech, it is that nothing is certain. Last spring and summer all the players were rudely reminded of that when inventories evaporated for some components, including tantalum capacitors used in cell phones and, yes, even memory chips used everywhere. All of a sudden the grand concept of the interlinked supply chain, with all the players looking out for everybody else in order to make stuff better, faster, and cheaper for the consumer somehow was forgotten in a frenzy of finger pointing.

With shortages rippling through the business and little relief in sight for some components, Solectron has found itself in the middle, criticized from both ends. Says Kevin Burns, a senior VP and Solectron's chief materials officer: "It's unbelievable. The component suppliers point at us and say, 'You guys are a big part of the reason. You brought a lot of price competition, and you just drove us down to where we couldn't invest in our business.'" Looking the other way, Burns says the company's customers tell him it's his job to make sure there are enough parts and pieces available to build what they need when they need it. But he complains that he's not getting enough help from many of Solectron's customers, who haven't been sufficiently open about product plans and future requirements.

Burns points out that there are a couple of ways to make sure components are available when a customer needs them. One, he can sign long-term contracts, locking in a price. But if that turns out to be a penny or so higher than the spot market at the time they are used, customers will complain bitterly. Solectron has also been talking with several suppliers about advancing them funds in one form or another so they can increase capacity. However, as he says a bit mournfully, "How does Solectron get paid for locking in suppliers that way?"

At least the subject is being discussed, Burns says. "The shortages have been a wake-up call about how to optimize the supply chain and make smart decisions on price, continuity of supply, inventory, availability, and other things." The question is whether enough people have heard the alarm. The suppliers have a point. They've been battered on prices for the past several years and haven't had the cash even if they had the will to add capacity. Then the cell phone business went crazy, prices shot up, and now the suppliers can expand. But what's the probability of a glut a year or so from now?

The bigger problem, Burns says, is on the other end. "If you get the senior executives in a customer organization in a rational discussion, they say it's a no-brainer--the important thing is continuity of supply." But as Burns adds, that "gets lost in translation" further down in the organization, where the buyer is being hammered on by his bosses to cut this year's costs by 5%.

There is an old answer to such supply conflicts, one that's now out of favor. It's called vertical integration. While Ko wobbles a bit on the current division of jobs in infotech between OEMs and contract manufacturing companies like Solectron, he is much more emphatic when he says he is not about to become a component manufacturer or chipmaker. Of course, today's arrangements between suppliers and the EMS companies as well as between the manufacturing entities and the OEMs can survive if the whole industry pays something more than lip service to the idea of working together for the benefit of the consumer. But don't bet on it.

MOTHERS WORK Giving the Lady What She Wants

It started as a mail-order catalog in the early '80s when Rebecca Matthias couldn't find maternity clothes that would look right when she went to the office. Gingham dresses and T-shirts with cutesy sayings didn't do it. Once launched, the business outgrew first her closet, then the apartment, then her parents' house, then a loft in an old building. Her husband, Dan, a computer specialist, left his job and joined her. In 1993, Mothers Work went public. For the year that just ended on Sept. 30, the company's $366 million in revenues were up about 22% from the previous year. Mothers Work has better than one-third and possibly as much as one-half of the U.S. maternity clothes market. No competitor comes close.

Today, side by side at a single desk surrounded by samples in a large room on the second floor of a block-sized plant in downtown Philadelphia, Rebecca and Dan Matthias, COO and CEO, respectively, oversee a supply chain with a high degree of vertical integration. It starts with designers working on the same floor and includes a vast central warehouse in the same building serving 726 company-run retail outlets. All of Mothers Work's lines of maternity apparel are sold through these stores, each with point-of-sale cash registers tied back to the headquarters computer so that what's sold yesterday can be replaced tomorrow when the UPS truck arrives.

Between the ends of the supply chain are a nearby company-owned textile warehouse, an expansive fabric-cutting room in the headquarters building, and sewing shops owned by others but carefully monitored by Mothers Work. The sewing shops are all over: just down the street, an hour outside town by truck, or weeks--even months--away in some 20 countries in Latin America, Asia, and the Middle East. The company's margins, unusually attractive for the apparel and retail businesses, attest that the arrangement is functioning smoothly. Operating income in the first nine months of fiscal 2000 was 7.6% of sales.

The key to how Mothers Work works is, as Dan repeats (and repeats): "Give the lady what she wants when she wants it." There's a relatively narrow window for when she does. About four million babies are born in the U.S. every year, which means that some three million American women are pregnant on any given day. Some are early in the first trimester, when clothes still fit, and others are in the last weeks, when it's getting late to buy something unless there's some other special event coming up. First-time mothers are the best customers. "Repeat offenders," says Rebecca--who herself falls into that category, with three children--"don't buy that much."

So figure a continually changing group of 1.5 million or so consumers who aren't about to wait until next season for a new skirt, dress, pair of jeans, or T-shirt. You've got to have the garment on hand when she walks in the door. The secret is making sure it is out on the store table or hanging on the rack without overloading the inventory and falling into the margin-destroying routine of markdowns on top of markdowns to get rid of stuff the store thought would sell but didn't.

The newly pregnant woman can be a great customer. She is, says Rebecca, "like a woman whose house has just burned down. She needs to replace her jeans, her underwear, her social gown, the clothes she wears to the grocery. And she needs to buy for more than one season." Mothers Work has about 12,000 stock-keeping units, or SKUs, individual items on the inventory list; each size and color of a given garment is an SKU. Says Dan: "It's kind of like a hardware store: low volume, high assortment." Moreover, Mothers Work has all the women's apparel complications: multiple seasons and changing fashions, with surprise losers and winners.

The company still publishes a catalog, but mostly as a way of pulling customers into the retail stores. Placed in ob-gyn offices, the catalog has a $10-off coupon for store purchases. The stores themselves are divided into four lines. A bit over half of them are Motherhood Maternity stores, selling a middle-market line in fashion and price. The big competitors are the maternity departments in Sears, Penney's, and Target. Then there is the Mimi Maternity chain, a youthful "with it" line for the buyer who normally shops at the likes of Bloomingdale's; plus discount shops in factory-outlet malls; and at the top of the market, A Pea in the Pod, a high-fashion chain with big-name designers and celebrity customers. Price tags on dresses run down from $450 at A Pea in the Pod to $16.90 at Motherhood Maternity. Nearly 30 malls have two or more company stores dicing up the market by price and style; and the company operates 131 leased spaces in department stores such as Macy's and Babies "R" Us that carry fashions from Motherhood, Mimi, and Pea.

The Motherhood and A Pea in the Pod lines were acquired in 1995 and then greatly expanded. As that happened, the Matthiases began to segment the business by price point, lifting the level at Pea and lowering prices sharply at Motherhood, to go after the Sears-type customer or, in Rebecca's words, "to compete with the real big guys out there where there's a big market." With average prices down about 50%, the average Motherhood store's revenues almost doubled. That opened the way to cut costs through high-volume purchasing. "The approach," explains Dan, "is lower the price, get the volume, and then get the margin."

For most U.S. apparel companies, the way to get the margin is to produce overseas. Yet that means they must order well in advance, push out piles of product to the retail tables and racks, and hope the line will sell. They have no chance to correct mistakes, catch a change in taste, or capitalize on a winner. By the time an order is placed at the factory in India or China and then cut, sewn, shipped, and cleared through U.S. customs, the next season, and possibly the next latest thing in fashion, will already have arrived.

At Mothers Work, says Dan, not having what the lady wants when she wants it--and maybe worse, having what she doesn't want--"drives people around here crazy." To make sure their stores have what customers want and still avoid markdowns, Mothers Work operates not on push but on pull, or replenishment. Stores receive a basic inventory. When something sells, headquarters sends out a replacement.

To accomplish this involves a three-level supply chain: domestic production mostly within a short distance of the Philadelphia headquarters and warehouse, plus production in the Caribbean area and Mexico, and last but not least, distant sewing plants that are mostly in Asia. Much of the fabric sewn by the local and Caribbean suppliers is first purchased and cut by Mothers Work in Philadelphia.

About 60% of the high-end clothing, meaning most of what's sold at A Pea in the Pod and much of the Mimi line, is made in the U.S. For one thing, it's fashion-sensitive and relatively low in volume. Besides, the fabrics are frequently more expensive and the designs harder to sew, so mistakes can be costly. That calls for close supervision by Mothers Work and experienced hands at the sewing machines. On the low end, T-shirts are almost all made in Asia, with jeans mostly sewn in Mexico: Style hardly changes and volume is high. For items between the two extremes, orders are placed simultaneously at three levels. Domestic production gets to the stores quickly to start the style or the season. Mexican or Caribbean work follows in a few weeks. The big shipments arrive from more distant factories in a couple of months.

Always keeping some domestic production, says Charlie Torok, the company's logistics vice president, "is our ace in the hole. That's where we correct our mistakes." If a product starts to sell faster than anticipated, local production can close the inventory gap. A shortfall in embroidered slacks that had become a surprise bestseller, for instance, was covered by Alice Chang, who owns a bright and airy 50-machine loft business in Philadelphia a few blocks from Mothers Work.

Last spring Mothers Work created a supply problem for itself when it ran a sale of two $11.90 T-shirts for $19.90. It worked a lot better than anticipated; inventory dropped precipitously. But even though the special price cut its margins, Mothers Work wasn't about to miss a sale. A plant outside Philadelphia turned out an extra 100,000 T-shirts in a month.

Mothers Work's system for replenishing stock in its retail outlets is triggered by those point-of-sale terminals. They capture information on individual customers (170,000 or more new names per month), create mailing lists organized by when the baby's due, send and receive digital photos that can tell the manager how to set up displays, and going the other way, let the home office know what the store looks like. Their key task, though, is to provide daily reporting on what's been sold so that it can be replaced as quickly as possible. Some stores receive as many as five replenishment shipments a week.

At the company warehouse in Philadelphia, the replenishment process gets under way each weekday evening. In the case of items like sweaters and blouses that are packed flat, the computer issues an internal order to top up the picking stations, instructing workers to move the required number of a given item from the inventory area to one of thousands of bins stacked in rows in the order fulfillment area. By early the next weekday morning, the computer double-checks the number of items in the bins, adding more as needed. Then it issues pick lists, or lists of items that need to be shipped to replenish the stock of a given store.

The warehouse's main order-picking section, with flat-packed items representing 6,200 SKUs, is a busy scene. Rectangular carriers, made of water-resistant synthetic fabric and each bearing a store's pick list of needed items, travel on overhead rails parallel to the bins. Pickers, almost all of them women and each responsible for a section of an aisle of bins, are waiting at their stations. As a carrier moves along, a computer-operated green light in front of each bin alerts the picker to add, say, a blue medium T-shirt from that bin and push a button when she does so. If the light stays on, she keeps picking blue medium T-shirts for that store and pushing the button until the light goes out. Once she "has got all the lights"--that is, picked all the items in her section for that order--the carrier moves on, another arrives, and the green bin lights glow once more.

At the end of the line, workers zip the carrier shut and lock the zipper tabs with a seal so that in-transit theft will be easy to spot when the carrier arrives at the store. There it is emptied and returned to Philadelphia for another order. The company has 24,000 carriers in its system. Monday is the busiest day. On one Monday this September, workers set a record: 140,000 items gathered and shipped in 11 hours. Dan claims that 99.5% of what's needed at the stores is available to ship when they need it.

Elsewhere in the Philadelphia building is a separate, automated system for picking garments, such as dresses and suits, that are stored on hangers and then shipped in flat boxes. The warehouse's $1.5 million "hanger sortation" system, built by a company in Bischoffheim, Germany, with a U.S. affiliate in New Castle, Pa., Gartner Storage & Sorting Systems, operates over a 22,000-square-foot area. First it gathers, say, all 20 medium-sized brown dresses needed by all the stores on a particular day. Then it sends them by overhead trolley to a "racetrack area," where it releases them individually to slide down onto one of 116 side positions where separate store orders are gathered.

Next, the sortation system returns to the inventory area for today's order of, for example, 15 small-sized blue suits. When all the picking is over, each store's order can be found on one of the side-position rails, which might hold just one store's orders or several orders separated from each other by a hanger-borne long strip of cloth. Installed some 15 months ago, the sortation system has cut the time required for picking hangered items in half and has reduced the number of people in the section from 41 to 16. On that record September Monday, this area picked 18,000 items. Mothers Work says the equipment paid for itself in just under a year.

Once a week, the Mothers Work computer reviews all store inventories to identify where specific items are moving or failing to move. In early fall, women in Dallas might still be buying cotton skirts, while in Boston they are switching to heavier material. On Sunday morning the computer e-mails store managers, instructing them to pack up slow movers and return them to Philadelphia, where they can be routed to other stores the day after they arrive.

Given Mothers Work's huge market share, the tough question is how to grow from here. In 1996 the Matthiases made a major error when they took over Episode, a chain of regular women's wear. Two years of losses later, they liquidated it at a cost of more than $20 million. Now they're trying to stay closer to the business they know. Speaking of the possibility of international expansion but wary about anything not tied to the pregnant American woman, Rebecca says, "We learned our lesson when we went astray. There's a lot more easy money to be made here before we tempt fate elsewhere."

Although it seems that every mall in America has a Motherhood store, Rebecca is convinced that there's room for more and says a total of 1,000 outlets is doable. In the past year the company has added clothes for nursing women and plus sizes for those who were large even before they got pregnant.

Like any business in the country, Mothers Work is also working the Internet. Proving that you can take the boy out of computers but not vice versa, Dan's eyes light up when he talks about sales through Motherswork.com. He's not eager to say how well the Internet sales are going except that this September they were about four times what they were in the same month last year. He figures the whole thing is a natural for Mothers Work, since "it's just like the catalog business. It's single-order fulfillment. We've always worked that way. So when we went into the Internet, it was like it just rolled into our system. It comes into our computers, automatically goes down to the pick list, and goes out."

FEEDBACK: psiekman@fortunemail.com. Stories from FORTUNE's Industrial Management & Technology section can be found at fortune.com/imt.