What NUMMI Taught Me About Making a Lean Transformation Last

by Jim Fulkerson

In 1982, General Motors closed an assembly plant in Fremont, California, that almost everyone was glad to see go. Two years later, the same building reopened with the same workers and started building some of the highest-quality cars in North America. Nothing about the people had changed. Almost everything about the system had.

That's the famous part of the NUMMI story, and it teaches a lesson worth keeping. When an operation is failing, look at the system before you blame the people. But there's a second lesson hiding inside NUMMI. You can only really see it decades after the plant opened, long after the cameras and the case writers moved on. It's the more important one, and it's the one I want to focus on here, because I lived inside it. The first lesson is about what makes a transformation work. The second is about what makes it last.

The plant GM wanted to forget

By the late 1970s, the GM Fremont plant had a reputation as one of the worst auto plants in America. The numbers backed it up. Absenteeism reportedly ran around 20 percent. On a bad Monday, or the day after payday, enough people just didn't show up that the company struggled to staff the line at all. Quality was abysmal. Cars came off the line with defects that had to be reworked, and a few were built so badly that GM's own audits ranked Fremont near the bottom of the whole corporation.

The labor relations were worse than the quality. The plant carried a backlog of thousands of unresolved grievances, by some accounts more than 5,000, plus a history of wildcat strikes. Trust between management and the United Auto Workers had collapsed into open hostility. Whatever culture the place had rewarded beating the system, because the system gave workers no reason to do anything else. So when GM shut the doors in 1982 and laid everyone off, the easy conclusion was that these particular people were why the plant had failed.

An unlikely partnership

Around the same time, Toyota and GM each wanted something the other had. Toyota was getting ready to build cars in North America, and before it committed to a plant of its own it wanted two things. Experience with the North American supply base. And a real read on what it was like to work with the UAW. GM wanted the other end of the deal. They wanted to learn how Toyota built small cars so efficiently, with such consistent quality.

In 1984 the two companies launched NUMMI, New United Motor Manufacturing, as a 50/50 joint venture in the shuttered plant. Toyota ran the operation and brought the production system. GM put in the facility and would sell a version of the car. The first one off the line was a Chevrolet Nova, basically a rebadged Corolla. Over the years the plant built others too: the Geo Prizm, the Corolla itself, the Tacoma. The Pontiac Vibe came later.

What made NUMMI a real experiment, and not just a new factory, was a staffing decision. Toyota agreed to rehire the old workforce. Roughly 85 percent of the original UAW workers came back, including the union leadership GM had spent years fighting. The "worst workforce in America" got handed the same building and a brand-new system. On purpose.

What changed, and what it created

The turnaround was fast. Within about a year, NUMMI was building cars at quality levels close to 

Toyota's plants in Japan, and absenteeism had dropped from around 20 percent to the low single digits. That change didn't come from tougher supervision. It came from a different assumption about who owns the work.

Every worker could pull the andon cord and stop the line the moment they saw a problem. Not just could. Was expected to. So quality got built in at the source instead of inspected in at the end. Teams owned their own standardized work and improved it themselves, which meant the people doing the job were the same people fixing it. Dozens of narrow job classifications collapsed into a handful, built around small teams. And Toyota committed to no layoffs. When sales slumped in the late 1980s, they kept that promise, using the downturn for training and improvement instead of sending people home.

We usually file all of that under "system." Andon, standardized work, classifications, the no-layoff promise. And it is system. But look at what it was really producing. Give people job security, real authority on the line, and ownership of their own work, and you don't just get better cars. You get a different way of seeing the job. That seeded a culture. And the culture turned out to be the most durable thing NUMMI ever built.

What I saw from the inside

I worked at NUMMI from January 2002 until July 2004. That's close to two decades after the plant reopened. By then the system had aged the way every system does. The original group Toyota trained back in the mid-1980s, the ones who'd been sent to Japan and built the place from the studs, had mostly moved on. New people had come in. And like almost every company I've seen, the turnover in leadership ran a lot higher than it did on the floor. Production workers stayed for years. Managers cycled through.

Here's the part that's stuck with me. By the time I got there, you couldn't assume the person leading you knew the Toyota Production System cold. Leadership had churned enough that TPS fluency at the top was spotty. On paper that should've been the beginning of the end. A lean system run by people who didn't fully understand it. And yet the plant still ran as a real lean operation. The andon cords still got pulled. Work still got improved on the floor, by the floor. The discipline held.

It held because the culture was engrained in the workforce, not in the org chart. The veterans carried "how we do things here" and taught it to every new hire, whether or not anyone told them to. The culture had gone self-reproducing. It didn't need the original trainers anymore. And here's the strange part, it didn't even need the current managers to be experts in the system. The floor held the system up from underneath.

Why this is the real lesson

Put GM Fremont and NUMMI side by side and you get the first lesson. The system matters more than the particular people you've got. But NUMMI's longevity proves something deeper. The culture a system creates outlasts the system's mechanics, and it outlasts the day-to-day competence of whoever happens to be running it.

This is where most transformations quietly fail. When a change effort is fresh, everything looks great. Everyone's just been trained, leadership is fired up, the tools are new, the discipline is high. It works beautifully on day one. The real test comes years later. After the champions have left. After leadership has turned over two or three times. After the training has faded into folklore. A system that lives only in procedures, in the heads of the people who built it, is fragile. It decays the day those people walk out. A system that's grown a real culture is self-healing. It renews itself every time a veteran shows a new hire the right way to do the work.

So flip the order. Most rollouts start with the tools. Kanban boards, 5S audits, value-stream maps, the visible stuff. Then they bolt on the system. Then they cross their fingers that a culture shows up on its own. NUMMI says do it backwards. Start with culture. That means respect for the people doing the work, real security, and the plain expectation that everybody improves the job. Build the system on top of that. Bring in the tools last. Culture first is the only order that survives the people who built it.

NUMMI closed in 2010, after GM pulled out in its bankruptcy and Toyota decided not to carry it alone. It closed for business reasons. Not because the culture failed. The floor was still running lean right to the end. The building didn't sit empty long. It's Tesla's main U.S. assembly plant now, still building cars on the same ground.

The lesson I carried out of Fremont is the one I'd give any leader starting a transformation today. Don't just ask whether your new system will work this quarter. Ask whether it'll still be working after everyone who built it is gone. You can copy tools. You can install a system. But only culture answers that second question. So start there.