Episode Details

The first two essays in this series established two facts. First, the NUMMI experiment proved that the same workers, in the same building, under a different operating system, went from GM’s worst plant to GM’s best. Second, Frederick Taylor’s published writings showed that the dominant Western manufacturing model was explicitly designed to remove intelligence from the people doing the work.

This week I want to show you what happens when you flip the switch — when you actually build a system that asks frontline workers to think. The data comes from the most sustained experiment in deployed frontline intelligence in industrial history: the Toyota suggestion system.

The Numbers

Toyota receives more than one million improvement suggestions from its workforce every year.

Let that register. Not one million complaints. Not one million grievances. One million ideas — discrete proposals for how to do the work better, submitted voluntarily by the people doing the work, processed through a formal system, and implemented at a rate that has historically hovered around 90 percent or higher.

That implementation rate is the key. Toyota doesn’t just collect suggestions. It acts on them. The system is designed so that the vast majority of ideas — particularly small, localized improvements submitted by individual workers or small teams — are evaluated and implemented quickly, often within the work group itself. The worker sees the result of their thinking reflected in the actual work process, usually within days.

The math works out to roughly one suggestion per worker per week. Not per year. Per week. This means the average Toyota production worker is generating, articulating, submitting, and seeing implemented approximately 50 improvement ideas every year. Over a career, that’s thousands of contributions to the operating system — each one a small proof that the worker’s intelligence matters.

The Contrast

Now consider what the same metric looked like at General Motors during the decades when Toyota was building this system.

GM had a suggestion program. Most large Western manufacturers did. But GM’s system required that each suggestion demonstrate a minimum calculated return on investment — reportedly around $20 per suggestion — just to justify the administrative cost of processing it. The system was designed to filter out “small” ideas, on the theory that only ideas with measurable financial impact were worth management’s time.

Think about what that filter communicates to the worker. It says: your idea has to be worth at least $20 to us before we’ll even read it. If your suggestion is about ergonomics, or sequencing, or a better way to position a part that saves two seconds per cycle, or a safety observation that doesn’t have a dollar sign attached — don’t bother. The system doesn’t want it.

The result was predictable. Participation rates at GM and most Western manufacturers were a fraction of Toyota’s. Not because the workers had fewer ideas. Because the system told them their ideas weren’t wanted.

What the Suggestions Actually Contain

The common misconception about Toyota’s suggestion system is that it produces a torrent of trivial ideas — a million sticky notes saying “put the trash can closer to the workstation.” That misconception reveals the bias. It assumes that because the ideas come from “operator-level people,” they must be small.

Some of them are small. Many of the best ones are. A two-second cycle time reduction, applied across three shifts running 500 units per shift, 250 days per year, compounds into thousands of hours of recovered capacity annually. Toyota understood something that Taylor’s system was designed to prevent workers from demonstrating: the people doing the work see things that no one else can see, because no one else is in that position, at that angle, with that frequency, for that duration.

But the suggestions aren’t