(FORTUNE Magazine) – A MAESTRO OF THE PLANT FLOOR Anand Sharma of TBM Consulting Group

Anand Sharma, 55, a personable manufacturing consultant who runs TBM Consulting Group in Durham, N.C., has a reputation for finding out what a factory is doing wrong by simply walking through it with the plant manager. Sharma usually asks the manager about the factory's "rhythm." More often than not, the manager expresses puzzlement. But Sharma, like a seasoned orchestra conductor, may already have noted off-tempo components on the plant floor--a machine with a hardly perceptible squeak here, workers laboring at an uneven pace there, too much inventory piled up.

Sharma trusts his senses to point to evidence of bad processes. In addition to looking for obvious signs--is the plant well lit and clean?--he checks to see if operators at one part of the line are working at a very fast pace while others elsewhere are working slowly or stopping. He observes whether the progress of a part being made can be tracked from beginning to end by line of sight. Says Sharma: "Where other people see complexity, I look at how simple things can be."

The assessment and the walkabout help Sharma and the 72 manufacturing experts who work with him select the best site for their first improvement project in a plant. The TBM experts then come in to eliminate the root causes of problems on the production lines. In the process they may restructure the whole plant operation. But they don't just tear up things and leave. Unlike consulting firms whose employees depart after a quick fix, TBM often has its experts stay at a plant for years, because it believes that improvement of operations never stops.

In the decade since it was founded, TBM (short for time-based management) has worked with more than 500 manufacturers around the world, generally helping them lift productivity 15% to 20% a year. Sharma's clients include big companies such as Freightliner, Kaiser Aluminum, Mercedes-Benz, and Polaroid, as well as such lesser-known names as Batesville Casket Co. in Indiana, Cold Spring Granite in Minnesota, and Huffy Sports in Sussex, Wis.

Sharma's rare skills took time to hone. While growing up in a part of India that was ceded to Pakistan after the country's partition, he felt driven by what he calls "a compulsion to make things work right." The son of a professor of Sanskrit who died while he was still a boy, Sharma and his two older brothers were brought up by their mother, who had started a sewing school for women. Sharma loved to build radios and bicycles, and knew he wanted to become a mechanical engineer. He picked Roorkree University in the foothills of the Himalayas because at the time it was India's only institution of higher education that offered hands-on training with production machines. The university ran a shop that made and sold electric motors.

An imaginative, straight-A student, Sharma did a comprehensive study of how to set up and run a manufacturing company that caught the attention of the general manager of the Hindustan Machine Tool Co. Offered a job there upon graduation, Sharma showed up for work in a three-piece suit. The general manager sent him back to his hotel to put on work clothes. When he returned, he was taken to the tool room, handed a piece of steel, and told to file it. After Sharma finished, the general manager informed the aspiring executive that he had made seven mistakes--not holding the file right, straining his body, and so on. The boss asked Sharma how he expected to teach people to do their jobs.

Thus began an odyssey that turned Sharma into an expert machinist and tool operator. The general manager dispatched him for six months to a training center run by experts from Switzerland, where Sharma toiled on the morning shift from 4 A.M. to 2 P.M. He recalls working on an especially complex part for two weeks, only to be told by a Swiss taskmaster to redo it because its dimensions were off by two-millionths of an inch.

Back at Hindustan Machine, Sharma underwent two more years of training in various departments and then was assigned to head a tool room. He was soon challenged by a machine operator twice his age, who questioned whether Sharma could make a better part than he. Sharma did, and the questions stopped. Soon Sharma became bored with all the grooming to become a plant manager. But today he says, "I was unbelievably fortunate to have gone through that kind of training."

More training came in the U.S. when the young engineer was dispatched to Franklin, Pa., to gain experience at a Chicago Pneumatics plant. The blunt-talking Sharma questioned the plant's old-fashioned product design and sloppy manufacturing methods, and was given a chance to make improvements. Eventually he left and took a succession of manufacturing jobs at U.S. Steel, Zurn Industries, and a division of American Standard, where he rose to director of manufacturing and later to vice president of strategic planning, R&D, and operations.

Along the way, Sharma got an MBA at Boston University. He kept up with manufacturing trends and attended quality-control seminars offered by the founders of the quality movement, W. Edwards Deming and Joseph Juran. By the early 1970s Sharma was designing better and faster production lines at American Standard's various divisions. "I was always trying to reach for the next level," he says.

He still had a lot to learn. The revelation of how much better manufacturing can be came in 1979 when Sharma met Toyota manufacturing guru Shigeo Shingo and was invited to see some plants in Japan. What stunned Sharma was the ability of Toyota's workers and others to replace dies on presses in minutes instead of the hours, or even days, that it took in American plants. The Japanese did this, moreover, with a very simple technology that employed compressed air to lift the thousand-pound dies as if they were feathers.

When his division of American Standard was put up for sale, Sharma moved to a manufacturing-consulting firm in Connecticut as an executive vice president. In 18 months he turned the money-losing firm into a profitable one. Unable to get along with the owner, Sharma and three colleagues started TBM Consulting in 1991, operating out of Sharma's home.

Today Sharma applies what he learned from the famous Toyota Production System (TPS) and adds a large dollop of Americanization. TPS is based on a Japanese update of Henry Ford's vision of integrated production. Ford was practicing just-in-time supply of raw materials and parts at the legendary River Rouge plant long before the Japanese popularized the term. TPS evolved during the transition from mass production to mass customization. Unlike the old "push" systems designed to build to inventory, TPS aims to build to customer demand in the shortest possible time and with minimum resources. Its Westernized version is now widely known as lean manufacturing. Sharma goes beyond TPS by combining both lean production and quality elements from Six Sigma into what he calls LeanSigma.

The advent of e-manufacturing nevertheless raises a question. Now that some companies' salespeople are beginning to send orders electronically from the field directly to production machines, is Sharma's old-fashioned emphasis on manufacturing excellence an anachronism?

Not at all. Most manufacturing companies, experts say, still haven't fully mastered modern "pull" production technology, or making products to customer demand, as contrasted with the conventional "push" production, in which stocks of unsold goods can accumulate. "We still have a lot of bloated and screwed-up processes out there, left over from the neglect of the '60s and the misapplication of the operations-research approach," says noted manufacturing consultant Patricia E. Moody, co-author with Sharma of a forthcoming book, The Perfect Engine. Sharma argues that manufacturing is the key to success of e-business, as some failing dot-com companies have discovered to their chagrin. He adds that "there are also a lot of bad plants in Japan," where, contrary to received wisdom, the Toyota system is not widely used.

At one level TPS is built on the concept of kaizen, Japanese for "continuous improvement." TBM experts adjust rigid Japanese methods to freer American ways when they establish kaizen methodology in a plant. In Sharma's approach, for instance, production-line workers have a lot more say than Japanese workers about changes on the manufacturing floor. To assure himself of their input, Sharma refuses to work with companies that propose to lay off workers after his system is introduced; that destroys morale, he believes. Any superfluous line workers are assigned other jobs, with some becoming trainers. "We unleash the power of the people," he says.

TPS's initial kaizen study teams are constituted equally of production-line workers, managers and supervisors, and office workers. The teams set up model lines and practice the changes before they are introduced on the floor. Sometimes TBM totally reorganizes production, as it did starting in 1998 at the Maytag plant in Cleveland, Tenn., that makes gas and electric ranges. With no added workers, production of one product line zoomed by 100%. Workers' suggestions are readily accepted and incorporated into Cleveland's new system, which is deemed always open to improvement. That one plant has cut its annual production costs by $7 million and reduced its inventory by $10 million. Says Tom Briatico, vice president and general manager of the Cleveland operation: "Anand Sharma and TBM have skillfully trained us in assembly-line layouts, quick die changes, and, most important, how to manage our operations for daily improvement."

Naturally, Sharma doesn't always succeed. He puts his failure rate at 5% to 10%. But he attributes this to lack of participation by higher executives in his efforts to introduce LeanSigma. That was the case, he says, at Chrysler in the mid-1990s and at two General Motors plants in 1998. Such negative experiences have taught Sharma to pick clients cautiously. He and his associates now carefully study a manufacturing company and its plants and reserve the right to withdraw within a week of the start of a project if they don't like top managers' attitudes. Out of every two companies that approach it, TBM chooses one.

The companies most receptive to change, Sharma finds, are in the $50-million-a-year to $5-billion-a-year revenue range. Results such as production increases may quickly become visible. But it takes a long-range commitment, Sharma believes, to maintain constant improvement in financial performance. In his view, the task of improvement never ends. To keep client companies on their toes as well as to expose new managers to the LeanSigma system, TBM employs a unique method in which, in addition to the ongoing kaizen campaigns, it conducts "kaizen breakthroughs" at client companies.

It's not unusual for TBM to assemble 50 manufacturing managers from client companies around the world for two days of classroom instruction, usually at a hotel near a big manufacturer. They are then bused to a plant. One such group, with participants from 30 companies, was transported last year from Chattanooga to Maytag's Cleveland operation, which still works with TBM. The trainees were split into four teams wearing different-color hats--red, blue, brown, and yellow. Armed with big yellow stopwatches and clipboards with timing charts, the teams spread out through the sprawling plant, spending 22 days to come up with further improvements in the production rhythm. Maytag was happy to get the visitors' suggestions.

A typical accolade for what TBM achieves comes from Pat Lancaster, CEO of Lantech Technology, a Louisville maker of shrink-wrap machines. When TBM began working with Lantech in 1992, the company was losing money. Four years later, sales per employee had increased 80%, and the company enjoyed a healthy 10.5% return on sales. Though the CEO won't disclose the privately held company's financial assets since then, he allows that "the results in terms of profit and sales growth have been awesome." He adds, "We wouldn't be the company we are today without TBM. There isn't a place here that you could go to and not see the impact of multiple kaizen. The biggest role that Anand played was to tell us the next step."

THE QUEEN OF ELEGANT SOFTWARE Pamela Meyer Lopker of QAD

In an era when software makers have saddled manufacturers with companywide, ponderous, tremendously expensive--and often ineffective--systems, one supplier stands out. This company is QAD (pronounced "quad") of Carpinteria, Calif., founded in 1979 by mathematician Pamela Meyer Lopker, now 46.

QAD (which stands for quality, application, delivery) is not as well known as such enterprise resource planning (ERP) vendors as Germany's lumbering SAP or California's Oracle Corp. But QAD has created some of the best ERP software for manufacturing plants. Bruce Bond, a computer-industry watcher at the Gartner Group in Stamford, Conn., has called QAD's software "ideal" for individual factories. Other analysts praise the company's software for e-commerce, an area in which QAD has leapfrogged the giants.

A big reason is Lopker's uncompromising dedication to using the latest software technology and making QAD programs simple and easy to use. If the men who run the big ERP companies are known in the trade as the "spaghetti kings"--because of their initial use of an old spaghettilike software code--Lopker could well be called the queen of elegant software. Says she: "Software needs to be intuitive, easy to implement, and highly reliable. It needs to provide immediate value at a low cost."

This philosophy has helped QAD grow at a compound annual rate of 36%. With revenues of $239 million in fiscal 2000 and 1,600 employees, QAD is a now good-sized company by software industry standards. QAD has installed its software at 5,000 manufacturing plants worldwide, creating one of the largest client bases in the industry.

Lopker's pursuit of perfection goes back to her childhood. Her father, a U.S. Navy aircraft-electronics engineer, kept Pam, as well as an older sister and a younger brother, abreast of technological developments. Their dad would often gather the kids to take apart mechanical devices like a swimming pool sweeper or an air-conditioning pump. Pam gained the distinction of being the first girl to take auto shop at her Cupertino, Calif., high school. In one project she dismantled and rebuilt an old Austin-Healy she had bought, earning the first A ever given in the course.

Lopker majored in math at the University of California at Santa Barbara and took computer-programming courses in her senior year. After graduating in 1977, she worked for two software companies in Santa Barbara. While helping a college friend find software to run a factory that made sandals and leather goods, she discovered a lack of good software for manufacturers, particularly for midsized companies with $50 million to $250 million in annual revenues.

The friend, Karl Lopker, whom she soon married, had started his company before his graduation from college with a degree in electrical engineering. Karl, who is now 49, also had a strong mechanical bent. He had grown up in Los Angeles, helping his engineer father build machines in his shop, including equipment that painted white lines in parking lots. Karl says he started making sandals because he could not find a pair to fit his extra-wide feet. His company had grown, and now he needed software to tie together manufacturing with distribution, finance, and other activities.

Pam Lopker started QAD in 1979 on a shoestring. Karl sold his interest in his company and joined QAD two years later. He's now CEO; Pam is president and chairman, the technical brain behind QAD who's responsible for software development.

To get started, QAD hired two programmers and sold enhanced Hewlett-Packard software that ran on HP computers. In 1985 Pam decided that QAD's customers, mostly small manufacturers, didn't want to be tied to software that ran on only one company's machines. In a pioneering move, the Lopkers decided to create manufacturing software based on the then-emerging Unix and DOS operating systems. To build such next-generation products properly, QAD took a bold step: It stopped selling software for a time while its founders immersed themselves in an intensive study of manufacturing.

By 1987, with Pam leading the software development, QAD created the first business application to run on Unix, which it called MFG/PRO (for Manufacturing Professional). All along, Pam's guiding principle had been to combine complex manufacturing principles with software that's easy for the everyday user to handle. "A software system is not much different from an electronic or mechanical device on the shop floor," she says. "You can build a Rube Goldberg device or a very elegant, simple machine." Putting these principles into practice, and using the latest computer languages, QAD created software that can be directed with much simpler instructions than the programs of the big ERP vendors call for. It takes three lines of code in MFG/PRO, for instance, to issue an instruction that can demand up to 100 lines in older languages.

In an updated form, MFG/PRO is now in use at 20 of the FORTUNE 100 companies and in hundreds of smaller corporations. Users include automotive suppliers Delphi and Johnson Controls, consumer-products giants Colgate-Palmolive and Unilever, electronics producers Philips and Sun Microsystems, food and beverage companies Coca-Cola and Quaker Oats, and medical and biotech companies Merck and Genzyme.

Unlike the big ERP vendors, which typically try to sell a company a single system to run all of its manufacturing plants, QAD from the start emphasized a "distributed" architecture in which each plant's software is implemented separately. This was a smart approach, because not all plants run the same way. Some build to order; others operate as order-filling job shops; still others do repetitive manufacturing. The big ERP vendors, led by SAP with its German philosophy of top-to-bottom control, tried to impose a uniform system on diverse plants. QAD, as Pam Lopker puts it, "went the loosey-goosey, freewheeling, American cowboy way," allowing each plant to execute the manufacturing system that suits it best.

There's another big difference. While QAD concentrated on manufacturing, the bigger ERP vendors tried to create software that would work not only for manufacturing companies but also for banks, insurance companies, and government institutions. "My premise," says Lopker, "is that you can't be all things to all people because it will get too complicated and too difficult to implement."

The speed with which QAD software could be implemented proved startling--typically weeks instead of months or even years, and more than four times faster than the industry average. In one telling example, software engineers at a Philips manufacturing plant in Australia, after conducting a six-month runoff between QAD and SAP software, started wearing T-shirts inscribed SAP = SLOW AND PAINFUL and QAD = QUICK AND DEADLY. Lopker estimates that one out of ten customers comes to her company after "ditching" one of the big ERP vendors.

QAD's flagship product, MFG/PRO, works to coordinate basic manufacturing-control programs of the sort described in the profile of Myron Zimmerman of VenturCom, which appears later in these pages. Designed for deployment at the plant-operations level, MFG/PRO allows a manufacturer to improve asset utilization in every node of the factory. The software not only schedules and supervises production but also handles distribution and support functions and links up with corporate financial programs.

Because 65% of QAD's clients are overseas, the company spotted the trend toward globalization early and began providing software suitable for use around the world. Accordingly, MFG/PRO incorporates different countries' methods for such functions as invoicing, accounting, and amortization. Lopker also recognized the need for supply-chain software long before "supply chain" became a buzzword. Years ago QAD was already automating the transactions between plants and their suppliers and customers.

To exploit the emerging e-commerce market, Lopker has led a big effort to develop an advanced software system that she calls eQ. QAD is selling eQ to both MFG/PRO users and other companies; the idea is to help them cash in on the power of the Internet to build global supply-chain applications. The eQ software uses a model to decide how an incoming order should be handled, tailored to the preferences of a company's trading partners. It can process transactions across multi-enterprise supply chains by running in an automated, unmanned mode.

The eQ system has won many tributes. In a recent report the Boston firm Advanced Manufacturing Research said, "QAD has always scored the highest of marks on speed and ease of implementation, and the real power of eQ is that end users, not IT people, have the flexibility to add and alter relationships and policies." Another eQ plus: In partnership with QAD, IBM will soon be marketing eQ software and opening up new markets for both companies.

The Lopkers are banking on eQ to make QAD a leader in easy-to-use e-business software for midrange multinational manufacturers. "We're leveraging off our position in the plants using MFG/PRO," says Karl Lopker. They also hope that eQ will lift QAD stock out of its recent doldrums. Upgrading MFG/PRO and developing eQ has cost a lot. As a result, QAD has lost money in the past three years, although it is expecting to report a small profit for its latest quarter. The Lopkers, who own 55% of the stock, have seen the value of their holdings melt from more than $400 million when the stock reached its $23-a-share peak three years ago to about $45 million at the recent $2.50-a-share plateau.

Morphing from centimillionaires into mere multimillionaires doesn't seem to faze them. But they are concerned about the depressed stock price. "While money is not the main driver of what I'm in business for," says Pam, "your daily stock price is a measure of your success and ability to win the game." Nevertheless, she argues, there's a connection between providing customer satisfaction and earning long-term value for shareholders: "Some companies," she says, "do whatever they can in the short run to drive their stock prices up, which may include selling products that aren't ready for the customer or reducing costs by providing less than adequate support services. At QAD we are continuing to focus on long-term customer value, and we believe that is the best approach."

Both the Lopkers and Wall Street analysts expect QAD to return to profitability in the current fiscal year as eQ gains momentum. Though Value Line gives the stock only a 4 rating for timeliness on a scale of 1 to 5, analyst Rick S. Plummer says, "I like QAD's potential. The future could be quite promising."

THE JEWELERS OF MACHINING John Maroney and son John Cameron

Behind many successful U.S. exploits in space--from landing the first men on the moon to placing a roving robot vehicle on Mars in 1997--stands a small machining company in a Los Angeles suburb that produces parts so finely machined that some are thinner than a human hair. Maroney Co. of Northridge, Calif., combines the discipline of a science lab with the flair of an art studio to turn out some of the jewellike industrial parts shown on a later page.

This highly profitable company is run by the father-son duo of John Maroney and John Cameron, who uses his middle name as his last name so as not to be confused with his dad. The company employs just 17 people. But most are highly skilled machinists, and it boasts an assortment of production machines that much larger outfits lack. One of Maroney Co.'s great strengths is the skillful use of electrical discharge machines (EDMs), which employ an electrode or fine wire electrically heated to the brightness of a laser beam. EDMs can sculpt metal parts, both inside and outside, with a finesse and precision impossible with other types of machine tools.

While EDMs are widely employed to make dies and molds in industries producing everything from automobiles to watches, Maroney uses its EDMs and other machines on jobs that other shops can't tackle. An electrode shaped like a tiny golf club, for example, can enter a drilled hole in a small block of metal to scoop out a cavity of an unusual shape. An EDM wire machine might be used to drill tiny holes at seemingly impossible angles to meet the requirements of a scientific instrument.

The Maroneys are classic American geniuses. What led them to fine machining was an early love of much bigger machines, especially rough-road motorbikes and racing cars. John senior, 69, displayed a mechanical bent while growing up near Los Angeles, where he fixed clocks and other appliances for neighbors. In high school he spent most of his time in the machine shop. After dropping out he went to work as a machinist at a small shop in Burbank. Later, in 1955, he set up his own shop with a hand-operated lathe that his former boss helped him buy for $1,000. Maroney at first did jobs that other shops turned over to him. His luck changed when a salesman persuaded him to buy one of the first EDM machines.

The EDM process is unlike that of any other production machine. When its electrode or thin wire is heated by current, it gives off sparks that can carve a work piece in a desired manner by making tiny craters in the metal. The process takes place in the safe confines of deionized water or mineral oil, either of which can serve as an insulator.

The problem with early EDMs was that electrodes eroded as they worked on a piece--a process for which computer controls now compensate. In those days, adjusting for this erosion had to be done by hand. Moreover, it took the skill of a John Maroney to build electrodes out of copper, copper tungsten, or graphite in ways that anticipated their wear. His skill enabled his shop to outdo competitors.

That remarkable capability led TRW's aerospace division to select Maroney Co. as the sole supplier of the key sleeves and the fuel-metering mechanism for the descent engine of the Lunar Excursion Module (LEM) that landed astronauts on the moon. Maroney beat out 28 other EDM shops across the country for the contract, which put it into aerospace work in a big way. To improve his EDM skills, Maroney spent several months at Swiss EDM makers in 1968.

The company has since participated in many other NASA and military aerospace programs. NASA's Jet Propulsion Lab in nearby Pasadena, for instance, turned to Maroney to build 235 minute parts for the Mars Pathfinder spacecraft that landed on the red planet in 1997 and deployed the Sojourner rover. Among other items, the Maroney shop machined crucial actuator gears used to reel in Pathfinder's deflated airbags. Made of especially tough steel, the gears were produced on two different types of EDMs, the only machines that can shape such heat-hardened materials. The Maroney shop has also been supplying parts for the Space Shuttle and the new International Space Station.

As son John Cameron, 48, likes to say, "The Maroney Co. has been to the moon and to Mars." But his shop is also doing a lot of work for earthbound applications, ranging from heart valves for people to parts for the robot hands and heads used in Disney theme parks' "Pirates of the Caribbean" exhibits. Maroney also makes parts for Panavision's silent-running movie cameras and metal sleeves through which Pfizer measures out doses of insulin powder. Other clients include companies that make radar, lasers, electricity-producing gas turbines, auto suspensions, and oil-well sensing devices. Not all these Maroney parts are tiny, but most have difficult-to-carve shapes and are made of unique materials such as beryllium copper, tungsten, titanium, and nickel incomel, a nickel-iron alloy.

These days Maroney is riding the crest of a wave toward greater miniaturization of consumer and industrial goods such as cell phones, medical devices, and parts for optical-fiber networks. The shop recently made the prototype of the world's smallest mass spectrometer, an instrument for identifying chemical compounds, for the Jet Propulsion Lab. It could serve as a handheld device in airports to detect drugs.

John Cameron, who is the elder Maroney son, became company president in 1996. His brother, Michael, 43, is the company lawyer. "Big John," as father Maroney is called, retains the title of CEO and still works at the company on his original hand-operated lathe, but he lets John run the company. Three of Big John's five golden retrievers, which accompany him everywhere, snooze near his lathe. He is an inveterate collector of cars. A high-domed building adjoining the company's spotless production facility is jam-packed with BMWs, a Hudson, a Porsche, Ford roadsters, and other older models, all of which Maroney hopes one day to fix. He owns 22 cars plus a big 1949 Ford truck, which he bought for $3,000 when he saw it parked on his street.

Making machines work right has always been Maroney's passion. When a household appliance such as a vacuum cleaner would break down, Maroney would refuse to go out and buy a replacement, not because he couldn't afford it but because he felt that "you'd be buying the same problem again." He repaired such machines himself. He still drives pickup trucks hopped up to run on racetracks, known in the trade as "spec" trucks. From time to time he and son John sponsor a racing team.

John Cameron picked up machining skills as a youngster working as a janitor in his dad's shop. In high school and junior college, Cameron spent a lot of time racing dirt bikes. He was so good that at 17 he was selected to represent the U.S. abroad. He spent time in Czechoslovakia and later in Paris studying art, rejoining his dad's company in the 1990s.

For a family-owned shop, the company invests an unusually large amount of money in varied production equipment. Forty-nine production machines and 12 measuring devices can be found on the Maroney floor, including eight EDMs, as well as milling and grinding machines, lathes, and many other types. Says Cameron: "We like to control everything here under our own roof, from concept to conclusion."

Attracted by Maroney Co.'s reputation, top Swiss and Japanese EDM makers vie to place their latest machines on its shop floor. Similarly, the shop attracts skilled operators eager to do challenging jobs. It attracts plenty of customers too, despite its premium prices. Says John Maroney matter-of-factly: "If there isn't some challenge in a job, chances are you can get it done for less money someplace else." Customers seem to have no problem with that. Maroney Co. has never employed a salesman, but it gets more work than it can handle.

HOW TO "BREED" SMOOTH PRODUCTION Bill Fulkerson and Dick McKinnon of Deere & Co.

A few years ago the idea would have been laughed right off the plant floor. Why not use the concept of artificial selection, similar to plant and animal breeding, to "breed" smooth production scheduling on assembly lines? The computer software that now makes this possible, known as genetic algorithms, is similar to a practice humans have followed for millenniums: mating plants and animals whose desirable characteristics can be enhanced by continued breeding. The result can be a beautiful rose, a winning racehorse--or a superefficient factory.

Two unassuming Deere & Co. engineer-mathematicians, William "Bill" Fulkerson and Richard "Dick" McKinnon, both 58, hold the distinction of having applied genetic algorithms for the first time in an industrial setting. In much the same way that a horse breeder mates swift animals to produce even faster-running offspring, computers at several Deere plants repeatedly crunch data on incoming orders and the availability of key machines to create an "improved strain" of a production schedule that minimizes bottlenecks.

This stunning feat builds upon the work of John Holland, a University of Michigan professor who turned the ideas underlying the artificial selection of plants and animals--a form of human intervention into Darwinian natural selection--into software algorithms. It is now known that at the level of the chromosomes, the carriers of hereditary genes, superior offspring result when parts of the chromosomes with desirable characteristics break up and "cross over" to link into new combinations of genes that reinforce these traits.

In a brilliant mental leap, Holland in the 1960s started encoding data into chromosomelike strings of binary numbers and other types of computer code. In a purely abstract exercise, he assigned higher values--the counterpart of desired characteristics in plant and animal breeding--to parts of the strings of numbers. Thus, in a piece of a binary code, 010110111, the last three digits were deemed the higher-value part. That string of code, the counterpart of a chromosome, would then be broken up just before the last three digits. The broken strings were next run through a computer where they were reconnected to other strings with similar characteristics and evaluated for their worth. If the computer detected a reinforcement of these "characteristics," it could instantly breed successive "generations" of the code string to fast-forward the evolutionary process.

Fulkerson and McKinnon, who happen to be good friends, are another duo of American uncut diamonds with modest beginnings. Both grew up in small communities and found math an easy subject even as youngsters. Both have always loved to solve puzzles.

Bill Fulkerson has always felt that "an individual can make a lot of difference--become a causal factor to precipitate change." Growing up on his father's isolated farm in Missouri, he went to a one-room country school. He later majored in math and minored in biology at Central Missouri State University. After earning a master's degree, he stayed on as an instructor in math for four years and then joined a U.S. Army research facility in Leavenworth, Kan. He did military operations research "in a rebuilt stable behind the penitentiary." Always independent-minded, Fulkerson says, "I was never a team player," although he sang tenor in a choir and played the trumpet as a teenager and still sings in a church ensemble.

At Deere, where he has worked since 1976, Fulkerson is called a "technical consultant to the director of information technology" and acts as project manager for system documentation. A humorous man, he prefers to describe himself as "a loose cannon and a futurist confused by the present-day world." His job reflects Deere's policy of allowing brilliant individuals to scout the frontiers of science and technology and bring new methods to the company.

Dick McKinnon followed a somewhat similar path to Deere. He grew up in Aurora, Ohio, near Cleveland. He, too, was fond of math. But under the influence of his father, a high-school and later college basketball coach, McKinnon spent most of his time in school at sports: baseball, track, and golf. Study came easy in high school, but he had to hit the books when he went to Northwestern University to major in math and minor in economics.

McKinnon came to Deere as a quality engineer upon graduation and has been involved in improving manufacturing processes and factory layouts. Although he has an office at the impressive Saarinen-designed steel-and-glass corporate headquarters in Moline, Ill., he can usually be found at one of Deere's plants, checking on production operations. "Dick never sits in an office," says Fulkerson. "He's totally integrated to factory operations." McKinnon has kept up his interest in sports by serving as statistician for the local basketball team, the Quad City Thunder of the Continental Basketball Association.

On one of his Deere assignments, McKinnon found himself facing a perplexing problem. He had just helped redesign the company's seed-planter factory in Moline. The plant boasted the best available machinery; had eager workers organized in teams, with their pay depending on the number of planters they built each day; and stood out as a clean, airy place to work. But scheduling a smooth flow of production turned out to be nearly impossible.

Prior to 1993 the plant had made only planter parts, which it shipped to dealers for assembly. The dealers usually hired high schoolers to put the planters together. But as the machines got more and more complex, warranty claims soared, and Deere decided to assemble the machines in-house. Production people had never faced the issue of the order in which to produce the planters, which come in a great variety of sizes and with different equipment.

Basically, a planter, which is pulled by a tractor, is shaped like an elongated steel frame. Mounted on the frame are "row units," each holding a container of seeds and fertilizer, either dry or liquid. The planters come with four-, six-, eight-, 12-, 16-, or 32-row units and various hydraulic and electronic controls. Deere offers 307 options, and millions of combinations are theoretically possible. Even a more realistic figure of 5,000 variations proved difficult for schedulers to cope with.

The Moline plant had converted to a system of building planters to fill farmers' orders. But manual scheduling turned out to be so complex that, after six months under the new system, the plant was perpetually 500 orders behind--orders that farmers could cancel. "The planter mix would be different each month," says McKinnon. "As soon as you would figure out a pattern that would make some production people happy, the mix would change. You learned that last month didn't apply to the next batch."

The problem was how to balance the flow of planters through the factory by taking into account a number of potential choke points, such as the limited number of welding stations. Extra-large planters, for instance, would take an hour to weld. Hence, they could not be bunched together on the line. "We didn't want 20 workers standing around waiting for two people to finish their work," says McKinnon.

Sequencing was further complicated by the company's decision to ship the planters to dealers as the machines came off the assembly lines. The planters would be put aboard waiting trucks, typically four machines per destination. To avoid turning the waiting trucks into inventory-storage machines, the four planters needed to come off the assembly lines within hours of each other. Although the production schedulers were trying their best, the system was simply too complicated to execute by hand. McKinnon decided that scheduling should be computerized.

Over lunch one day in 1993, he mentioned his problem to his friend Fulkerson. If anyone ever lived up to the French physiologist Claude Bernard's bon mot that "chance favors the prepared mind," it was Bill Fulkerson. It turned out that he had the answer.

A highly curious individual in a freewheeling job, Fulkerson already knew about genetic algorithms. He had studied evolutionary genetics and statistics in college, where he had considered becoming a geneticist. On top of that, he had recently spent six weeks learning about genetic algorithms from one of John Holland's former students, a professor at the University of Illinois.

Fulkerson knew that genetic algorithms could be applied to scheduling. "I was like a hammer..." begins Fulkerson. And McKinnon concludes, "And I gave him the nail." The two were pioneering at a time when other production specialists were focusing on improving order forecasting. That would be an almost impossible task, Fulkerson believed: "I was convinced that if we could improve production scheduling, we could eliminate over-reliance on forecasting."

He went on the Internet in search of a vendor that might offer the algorithm-based software he had in mind. He turned up Optimax Systems, a startup in Cambridge, Mass. Optimax had been formed by two computer scientists, Jeff Palmucci and Gyl Syswerda, a former student of John Holland's, and by consultant Jeffrey Herrmann. The three had worked for the consulting-research firm Bolt Beranek & Newman and had successfully installed a genetic-algorithm scheduling system at an Air Force lab in California. Now they were looking for commercial clients.

When the Optimax team flew to Moline and said it could solve the scheduling problems at the planter operation, "a majority of the people didn't believe them," McKinnon recalls. But when the team returned three months later with a prototype scheduling program, McKinnon says, "it did everything they said it would. People were amazed. The Optimax team demonstrated the prototype at 10 A.M. At 1 P.M., the question plant people were asking was, 'When can we have it?'"

Feeding an old order for 400 planters into the software system, McKinnon and his associates got a different sequence for planter assembly than they had been using. The genetic algorithms also showed how they could assemble 21 or 22 planters a day without overtime; until then, the plant had been able to build only 18 a day with overtime. "The system knew all the rules and kept everyone working at the same pace without building planters too close to each other," says McKinnon. Since the scheduling system was put into operation early in 1994, there have been no order cancellations or delays in executing orders.

The Deere people helped Optimax build the system by setting priorities. "We asked the people involved in production a lot of questions. 'Would you rather have this than this?'" says McKinnon. The questions involved details about planter types and components such as liquid fertilizer tanks, which take more time to make. "We just kept asking questions," he says, "and then assigned value points to the different segments based on the answers."

The program also split up planters into three categories--small, medium, and large--as it strove to meet multiple objectives. When scheduling was done by hand, says McKinnon, "all you could keep track of were big things. Now, with genetic algorithms, we could look at little things too. We could handle different models of planters with all the various options. All this is now done automatically and much faster." It used to take three days to schedule the next two weeks' production, McKinnon says, but a computer now does it in an hour or two, and "much better than any human." But the resulting schedule isn't frozen. If changes based on last-minute orders need to be made, they can be entered by hand.

One of the biggest advantages of genetic algorithms, McKinnon says, "is that now I get a warning when something will show up that I really don't want." In the past such problems as, say, too many fertilizer tanks scheduled would not be apparent until the strain began to be felt at the welding stations. "Now," McKinnon explains, "managers can see such problems 15 days in advance and can plan for them appropriately by adding extra shifts or new machines."

In the planter scheduling room, yellow, blue, and red blocks flash on a spreadsheetlike schedule on a computer screen as Larry DeClerck, the genetic-algorithm scheduler, proudly demonstrates the system. Blue stands for near-perfect, yellow is a step down, and red spells trouble. "This is so much better than what we had before," says DeClerck. "We were trying to build to order, but not very successfully."

From the planter factory, Fulkerson and McKinnon have spread the use of genetic algorithms throughout Deere's agricultural-machinery division, the company's largest. Generic algorithms now schedule the production of those cruisers of the prairie, Deere's huge, $200,000 harvesting combines, at a plant in Moline. They also schedule engine production in Dubuque, Iowa; tractor output at a plant in Mannheim, Germany; and other production elsewhere.

The breakthrough at Deere attracted production experts from dozens of companies, and they still keep coming. Optimax, which was acquired in 1997 by the software highflier i2, still dominates its field. "We've just scratched the surface," says Optimax's Jeff Palmucci, now an i2 Fellow. The algorithms have been incorporated into i2's supply-chain software, and work is under way at universities to use genetic algorithms for automatic design of software programs and integrated circuits, among other things.

The successful application of genetic algorithms at Deere has led to their use at GM, GE, Motorola, Caterpillar, and many other companies. These companies have cut inventories, sometimes by hundreds of millions of dollars. They have also shortened assembly cycles, boosted yields, and decreased defects. No one has calculated the exact dollar savings at Deere. "But the people on the plant floor and the suppliers have been affected in positive ways," says Fulkerson. "This means the system is doing the job."

A MICROSOFT ALLY IN THE FACTORY Myron Zimmerman of VenturCom

The manufacturing floor has been slower than the rest of the economy to reap the full benefits of modern computing. One big reason: Today's dominant computer operating system, Microsoft's Windows, until recently could not be entrusted with running production machines because it lacked key features to ensure reliability. A small, fast-growing software company in Cambridge, Mass., called VenturCom has remedied the deficiency, paving the way for Microsoft's further advance into the factory.

The force behind VenturCom since it was founded in 1980 has been Myron Zimmerman, 48, yet another farm-bred genius. "Without third-party software providers like VenturCom, Microsoft would never have been able to achieve the level of penetration on the plant floor that it has today," says Dan Miklovic, a vice president at the Gartner Group in Stamford, Conn., which follows the computer industry.

The road to Zimmerman's impressive achievement began in rural western Pennsylvania. He grew up in a family of strict Mennonites, though they were not of the Amish or "plain people" variety who shun machinery. His parents spoke German when they wanted to keep secrets from young Myron and his two brothers and two sisters. Raising cattle, corn, and hay were a sideline for Zimmerman's father, who was primarily a self-taught inventor and small manufacturer. He still runs Zimmerman Industries, an $8-million-a-year company he founded, for which he invented a widely used type of mobile cement mixer.

Myron worked both at his father's company and on the farm, which was big enough for him and his friends to launch model rockets that whooshed as high as 1,500 feet. He also built radios and go-carts and read a lot of science fiction and history. He excelled in math and science in high school and majored in physics and math at Juniata College in Huntingdon, Pa. As a freshman, Zimmerman says, he "fell in love with computers." He not only ran a computer lab but also helped the college chemistry department computerize its instruments.

He started writing software early on, deriving pleasure from "form and function coming together in exactly the right way" and from bringing about "harmony between a computer and the equipment you're trying to control." Graduating from college cum laude, Zimmerman applied to four universities. MIT reached out to offer him financial aid. He majored in atomic physics there, doing laser spectrometry on single atoms and participating in pioneering research for which one of his mentors shared a Nobel Prize in Physics last year. He went on to get a doctorate in atomic physics even as he continued his interest in computing.

When the Unix operating system came on the scene in the late 1970s, Zimmerman quickly mastered it and installed it in his and other MIT labs. He also began to do consulting while still a graduate student, putting in the new operating system at industrial companies. In 1980 Zimmerman got together with three MIT friends to form VenturCom (for Ventures in Computing). His roommate contributed $10,000 toward the purchase of a Digital Equipment minicomputer, which was installed in a spare bedroom in their apartment. Within a year the founders hired two employees.

VenturCom began developing software, including a specialized version of Unix called Venix. The company also wrote application software that could be used with Venix to control industrial jobs. It was put to use at an Alcoa aluminum-rolling mill and in a dozen other large projects at companies such as Digital Equipment, Boeing Aerospace, and Foxboro Corp. VenturCom grew to a staff of 40 in the late 1980s, with Zimmerman serving as president and chief technical officer. He retains the latter title today. The other three founders left, although all remain friends. A few years ago the company hired a professional CEO.

Zimmerman's concentration on software that controls industrial production equipment stemmed from his early work on computerizing instruments at Juniata College and MIT. About 90% of factory tasks, such as painting cars or handling automatic packaging, can be run in so-called "soft" real time, which means that a computer generally must issue a command every second. "Hard" real time, on the other hand, is essential in the remaining 10% of the tasks, which are more demanding, such as running a nuclear power plant or directing the movements of a robot. In those applications, commands must be issued every thousandth of a second, or even every millionth.

On your office desktop, no such speed is needed. You switch on your PC, wait a bit, and select applications by hitting appropriate buttons. Seconds pass. Not so on the factory floor. An industrial computer dedicated to a task starts working the second you turn it on. If a millisecond command is missed, a robot might shatter a car windshield or injure a worker.

From the start, Zimmerman's aim was not only to meet this requirement for speed but also to link the specialized real-time software with mainstream computer operating systems. The latter run such "housekeeping" chores as allocating space on storage disks, controlling displays on a screen, and facilitating connections to a network.

In the absence of such unifying systems, manufacturers traditionally used such stand-alone control devices as programmable logic controllers (PLCs), early and rather crude versions of industrial control computers, to run conveyors, paint booths, and other equipment. PLCs were not connected to larger factory networks. Unifying them, Zimmerman believed, would greatly reduce the cost of computing in factories by putting the control and connecting software in a single box. The greater "intelligence" of the operating system would link the production machines with the rest of the factory, allowing, say, maintenance people or accountants to see production data. An operating system would also enable manufacturers to integrate complex motion controls for machines. It would provide graphic displays and other convenient features too.

The two sets of software, a real-time control system and an operating system, would run on the same computer, which would switch back and forth between these functions. Since data processing happens so fast, the user would be none the wiser when 20% of a second was taken up by real-time tasks and the remaining 80% by more general processing. But real-time tasks would take precedence over the more generalized work of the operating system. Like a traffic cop letting police cars through red lights, it would allow real-time commands to be issued whenever necessary.

Soon after VenturCom had successfully incorporated its real-time software into Unix, there emerged a new contender in the fast-moving field of software. In the mid-1990s Microsoft introduced its big new operating system, Windows NT (for New Technology). It offered a far greater range of applications than Unix. Though highly reliable, Unix by then had fragmented into 36 different "flavors," or versions, and there was a dearth of off-the-shelf software to support Unix's expansion in the factory. Microsoft was emerging as the dominant force in software and attracting small collaborating companies by the thousands to further its reach with myriad off-the-shelf programs.

But a big shortcoming threatened to limit Microsoft's inroads into the world of manufacturing. Designed for the more forgiving office applications, NT lacked the capability to execute real-time tasks, especially the hard variety. A number of small companies stepped forward to remedy the problem with real-time extension software for NT. But VenturCom eventually outdid them all.

Zimmerman and his programmers had analyzed NT and decided that the huge program, with some five million lines of code, could be slimmed to the point where it could be installed on the factory floor and equipped with a hard real-time extension, which they called RTX. "People said we were crazy," says Zimmerman. "But after we showed that we could run NT with a small amount of memory, and then could scale up its functions, people went 'Wow!'"

VenturCom also had to deal with a major annoyance: the so-called blue screen of death that plagued NT users in its early days. When NT crashes, white text describing what has happened appears on a screen that has turned blue. Most of the time such failures were traced to improperly tested hardware and not to NT itself, but fears of the blue screen persisted on factory floors. RTX, used in conjunction with a slimmed-down NT, overcame these fears by ensuring that hard real-time controls keep running even if NT crashes. RTX then brings the manufacturing operation to a deliberate, slow halt so that it can be reset and restarted.

When VenturCom approached Microsoft in 1995 with an offer to help it to both market a smaller version of NT and embed RTX in it, the response was favorable. Eager to conquer the factory market, Microsoft knew that VenturCom had already successfully developed RTX for Unix. VenturCom stood to benefit too, since it would be cashing in on the growing trend toward using NT in other areas of factory operations, such as accounting and distribution. "The trend that we're capturing," says Zimmerman, "is bringing a whole array of software to one box, where we merge Windows and real-time functionality."

As small competitors have fallen by the wayside in the past year--either through acquisition by VenturCom or by leaving the field--VenturCom has emerged as the dominant RTX company working with Microsoft. VenturCom resells NT with RTX and has also licensed scheduling software to Microsoft that enabled the big company to offer a product called NT Embedded in 1999. This slimmed-down version of NT runs on computers in manufacturing, communications, and other fields.

Microsoft is on its way to becoming a dominant presence on the factory floor. According to Jerry Krasner, executive director of Electronics Market Forecasters, part of the trade journal publisher CMP Media Electronics Group, 93% of the software developers surveyed by his organization are now using some form of Windows for industrial automation. Bill Veghte, a VP at Microsoft, says VenturCom is playing "a critical role" in this trend.

So impressed has Microsoft been with VenturCom that it wound up investing close to $5 million in the company, for just under 10% of its stock. Intel and GE Equity have made investments of a similar size, and venture capitalists own smaller stakes in the still privately held VenturCom. The company is beginning to turn its small technology into a big business success. Revenues leaped to $19 million in fiscal 2000, from $11.5 million the year before, and the company employs 140 people today, up from 65 a year ago. The company expects to become profitable this year, after running at a small loss because its fast growth consumed lots of cash.

Always looking to the future, Zimmerman is now working to further improve interconnections inside manufacturing plants, such as those between production machines. Twice a week, he puts aside software and becomes an ice-hockey player. But he never forgets his main mission. When he gets bumped a bit too hard as he sweeps down the ice as a right wing for the Killer Bees, a Boston-area amateur club, Zimmerman will remind the younger opposing player: "Cool it, kid. We've got to be back at work tomorrow morning!"

FEEDBACK: gbylinsky@fortunemail.com

Stories from FORTUNE's Industrial Management & Technology section can be found at fortune.com/imt. Executives in manufacturing and research and others eligible to receive FORTUNE's Industrial Edition can subscribe by calling 888-394-5472.