Sterman

Case studies used as examples for the role of systems thinking in implementing improvement programs

September 27, 2001

How does system dynamics modeling help companies implement successful process improvement initiatives?

John Sterman, the Standish Professor of Management and Director of the MIT System Dynamics group, discussed the dynamics of organizational change as a part of the on-line seminar series that LFM-SDM hosts each month for LFM-SDM alumni. On September 28, 2001, Sterman answered this and other questions in his presentation, entitled "System Dynamics in Action: Creating and Sustaining Process Improvement."

"Many process improvement techniques have been developed over the past 30 years, and a number of companies have been successful in deploying them," Sterman said. "But nearly every company has a history of failed improvement programs. Employees often talk about the latest management fad or ‘flavor of the month program.’ Why do these programs so often fail despite their potential for significant improvement? More to the point, what can you do about it? What role might system dynamics and systems thinking tools play to help catalyze successful change in organizations?"

To illustrate this, Sterman provided the audience with two case studies of companies that used system dynamics modeling to design and implement process improvement programs, stuck with them, and came out ahead. Both DuPont and British Petroleum (BP) found that they were spending more on maintenance than top industry performers, but had less reliable equipment, a puzzling paradox. The issue wasn’t merely unreliable equipment. The root of the problem they faced was in the mental models that fostered a culture of reactive maintenance, where maintenance was seen as a cost to be minimized. Most maintenance resources were devoted to repair of broken-down equipment, with little time or incentive for proactive work, leading to still more breakdowns, and ultimately higher costs. Worse, maintenance mechanics felt that there was nothing they could do to break out of the cycle. And there had been a long history of failed attempts to implement programs to improve maintenance. Each time, they had little impact, or caused costs to rise, and were soon abandoned.

A small group at DuPont dedicated to maintenance process improvement came across the concept of system dynamics and decided to incorporate it into a modeling process. After bringing in an experienced modeler to act as a process coach and facilitator, the group developed a simple dynamic model of the maintenance system. All equipment was divided into one of three categories: operable, broken down, or taken down for preventive maintenance. As simple as it was, this technique was very unconventional.

"This [method] contrasts sharply with many traditional maintenance improvement tools, which often are extremely focused on detail complexity, such as tracking the repair history of each individual piece of equipment, and not on the dynamic complexity of the maintenance system-the interplay of resources, effort, skills, and equipment that determine breakdowns and repair rates" said Sterman. "The key was to identify the feedback processes that determine the flows of equipment amongst these three categories."

This led the DuPont group to another important insight – the company needed to stop thinking of maintenance as a cost center, and start thinking about the physics of what’s going on with the equipment. Because defects are usually latent, a machine can still run, at least for a while. Sterman explained that this can lead to a slippery slope. Even when proactive maintenance is necessary, plant managers often say that they can’t afford to take down a pump because it would have a negative effect on production schedules and profit margins. With less proactive maintenance, breakdowns increase, reducing resources available for proactive maintenance still further. Worse, as uptime falls, plant managers are less and less willing to take operable machines off line to do preventive and predictive maintenance work. In the short run, uptime remains higher, but soon defects accumulate, leading to more breakdowns and even less proactive maintenance work.

"The system can tip into an unfavorable reactive maintenance regime where, despite your best efforts, you end up with high breakdown rates, low uptime, and increased cost. Your maintenance staff spend nearly all their time in reactive repair work, with no time to do the right thing," Sterman said. "What we’ve got here is a system that can progressively slide downhill until you’re trapped into a reactive maintenance regime with dysfunctional budgets, behavior, decisions, rules, attitudes, skills, and, eventually, culture. Everyone knows that proactive maintenance is better than reactive maintenance (an ounce of prevention…) but find themselves trapped in a system that makes it nearly impossible to do the right thing."

The simulation model clearly showed how to escape this trap. A sustained effort to focus on proactive maintenance could reduce breakdowns, freeing up resources for further proactive effort. The vicious cycles that previously dragged the organization down would then become virtuous cycles, cumulatively and progressively improving reliability and, ultimately, lowering costs. However, implementing these policies necessarily led to a worse-before-better tradeoff: in the short run, increased proactive maintenance raises costs and lowers uptime as operable equipment is taken off line to carry out additional proactive maintenance work.

With the simulation results in hand, the challenge now was implementation. Simply telling people to change their deeply entrenched behaviors doesn’t work. The modeling team needed a way to create an environment in which the front-line workers could experience these dynamics for themselves.

Inspired by existing system dynamics role-play games (also known as "management flight simulators"), the team converted their model into a board game, played in a two-day "learning laboratory." The "Manufacturing Game" represents a typical plant, with chips representing equipment, mechanics, and parts. The players, representing operators and maintenance managers, have to meet demand and maximize profits while dealing with breakdowns and allocating mechanics between reactive repairs and proactive maintenance. While highly simplified, the game realistically captures the delays and tradeoffs involved. In the learning laboratory workshops, the game rapidly became a real plant, generating the same conflicts and emotional reactions observed in the field. But because the game allows people to see the long-run impacts of their decisions in an afternoon, and because people can try things they felt they never could in the real plants, it enabled them to see for themselves that it was possible to escape the reactive maintenance trap. They learned what it would feel like on that journey, particularly, the fact that things would get worse first, and only later improve. For many players, the game was the first time in their careers they experienced the possibility that reliability could be improved, and that they could make a difference.

The game proved to be popular, and soon plants throughout the system were showing results. Just as in the game, the first impact was a reduction in uptime and a rise in costs, but soon the benefits started to accrue. Mean-time-between-failure (MTBF) for key equipment rose more than 2.5 times faster among plants that played the game and implemented the program compared to those plants that didn’t. Costs fell 20 percent at participating plants while costs actually rose 5 percent at plants that didn’t participate. Product quality and delivery reliability improved.

Soon, other companies heard about the game and the learning laboratory. The leader of the DuPont effort took early retirement, licensed the game from the company, and became an entrepreneur. One of the first plants he worked with was British Petroleum’s Lima, Ohio oil refinery. Once a highly profitable plant, the Lima refinery had, after years of cost-cutting, slid into the same vicious cycle of reactive maintenance and low uptime. BP announced that they wanted to sell the plant due to its low performance. The game, learning lab, and other system dynamics tools were introduced to BP in 1994. Eventually more than 80 percent of the employees participated, and soon action teams were hard at work on improvement initiatives throughout the plant. As expected maintenance costs ballooned 30 percent in the early months. In 1996, BP, unable to find a buyer, announced that they would close the plant. But the benefits of the maintenance program began to accrue and receive recognition. By 1997, the MTBF on pumps had risen 480 percent, safety incidents and lost hours had fallen by a factor of four, and they had set 34 production records while reducing pollutant emissions. The program saved the plant $43 million annually at a cost of just $320,000 per year in additional maintenance efforts. In 1998 BP found a buyer, rescuing the region’s largest employer. The new owner, noting the high productivity and reliability of the plant, pledged to increase the number of jobs by expanding the chemicals operation. Today, the game and learning lab are used throughout BP and in a growing number of other firms.

"The BP experience illustrates the power of a shift in mental models," Sterman says. "The BP team reduced butane flare-off to zero, generating annual savings of $1.5 million a year and reducing pollution as well. The effort took two weeks and cost $5,000, a return on investment of 30,000 percent annually. What had stopped them from implementing this improvement long ago? Members of the team knew about the problem and how to solve it for eight years. They already had all the engineering know-how they needed to solve the problem, and most of the equipment and materials were already on site. The only barriers were the mental models through which employees came to believe that they were powerless, that the problem was imposed by external forces beyond their control, and that a few people could never make a difference.

"These entrenched mental models changed in four essential ways. The belief that the problem was out there had to change from ‘our equipment is lousy and there’s nothing we can do about it’ to ‘our equipment performs poorly as a result of our own past policies-if we change our behavior, the equipment will respond.’ The focus on defect correction through repairs had to shift to a focus on defect prevention and elimination. The focus on minimizing maintenance costs had to shift to maximizing overall organizational performance. And they had to realize that escaping the trap of reactive maintenance necessarily involved a worse-before-better tradeoff."

Sterman concluded, "The system dynamics model was essential, as it led to the initial insights into the dynamics of process improvement and the synergistic effects of high-leverage policies. The model also allowed the modeling team to develop the game and helped make it realistic. Ultimately implementation success required the them to embed their insights into a learning environment that involved the active participation of the people on the front lines, that enabled people to discover those insights for themselves, and that spoke not only to their heads but also to their hearts."

Follow up information: The maintenance case, along with other case studies of successful application of system dynamics, and modeling tools are discussed in Prof. Sterman’s new textbook, Business Dynamics: Systems Thinking and Modeling for a Complex World, published in 2000 by McGraw-Hill. Information on the game is available at www.mfg-game.com.