John Deere Runs on Chaos

How one company factory combines two trendsetting ideas: mass customization and complexity theory.

Mass customization: Good idea, but how does it work? Complexity theory: Interesting concept, but how do you apply it? You can benchmark both ideas in one visit to one place: a 625,000-square — foot facility in Moline, Illinois that houses the John Deere Seeding Group. There Deere & Co., the world’s largest maker of farm machinery, produces seed planters in 45 different models and with a total of 1.7 million options – in effect, customizing each planter. To make that happen, Deere deploys methods borrowed from bio-mathematics, scheduling production so carefully that every one of those 1.7 million permutations can be built on the same assembly line – thereby putting complexity theory to its ultimate test.


Deere in the Headlights

The story behind Deere’s adoption of these two cutting-edge concepts begins in the late 1970s. After decades of making a wide range of tractors for commercial farmers, Deere launched a new strategy: To cut costs, it would scale back options. Before long, Deere was hearing back from its customers: Less was not more. So Deere switched strategies. It rediscovered the importance of variety and soon began promising farmers even greater choice than before. At the same time, cost-conscious executives insisted that all of this flexibility come without the cost of building a new factory. A program called Vision XXI was launched to reconcile those goals.

This challenge and its solution propelled Deere to the forefront of industrial-process design. Unable to use traditional “hand scheduling” to achieve the new level of variety, Deere’s Bill Fulkerson, a 55-year-old math professor turned staff analyst, used the Net to post a plea for automation ideas. One reply came from computer scientists at a company that used biological algorithms to build scheduling software for the U.S. Department of Defense. Adapting algorithms originally developed to help analyze cell division, these computer scientists (who were soon to form their own company, Optimax Systems Corp., based in Cambridge, Massachusetts) designed software to create “perfect” assembly schedules.

The new software, called Optiflex, was installed in January 1994. Today it runs in six plants and gives Deere both the variety and the efficiency that the Vision XXI project aspired to. Here are some of the principles that make the system work.

The Whole Is Greater Than the Sum of Its Parts

The move from simplicity to complexity required a completely new mind-set. For example, the company’s traditional manufacturing approach had allowed assemblers to see only one piece of the entire product. “Before, someone might build the same part over and over again but never know what it did or where it went on the planter,” says Janis Atkins, 42, a completed-goods supervisor, who used to schedule the planter assembly line with pencil and paper.

Vision XXI turned that approach on its head. The new design instituted a process flow, eliminating the temptation to let inventory pile up. The company also moved machines closer together, reducing the amount of handling required for each part. Then workers were reorganized into 12 modules, each responsible for building particular subassemblies and attaching them to planter frames.

Giving Information Is Giving Control

To help managers and workers cope with their increasingly complex tasks, Deere floods them with information on everything from assembly schedules to quality control. As a result, module leaders have the information they need to control their own budgets – including regular reports on their units’ staffing, overtime, and maintenance costs. On the operational level, this information lets module leaders and team members plan ahead. Because the Optiflex system knows the configuration of every planter that will be built over the next 15 days – and anticipates the parts that each planter will need – module leaders can schedule work for subassemblies with near-perfect efficiency. “We know a week ahead of time when a new kind of planter frame is coming,” explains Brad Dykeman, a 33-year Deere veteran who runs a robot welder. “We know what needs to be done, and it gives us a chance to adjust.”


If You Own the Process, You Own the Result

Given the complexity and flexibility of the Deere assembly line, day-to-day decision making cannot come from the top. As much as possible, the company distributes authority to where information and incentives already reside. Quality is the area most affected by the new system. Because their pay depends both on the number of completed planters and on the quality of those machines, workers have a vested interest in making the process run smoothly – in their own module and elsewhere on the line. For example, if Nelsoandra Cole, a 29-year-old assembler, finds something wrong on a planter as it rolls past her module, she deals with it immediately. “I find the person responsible for the component, and I have it fixed right then and there,” Cole says. “And if that person can’t get to it in time, I find someone who can.”

Paul Roberts ( is a frequent contributor to Fast Company. You can find Deere & Co. on the Web (