The encyclopedia · Engineering & Operations · Operational decision · 1999–2005
GM put queueing models on its assembly lines to find bottlenecks, saving over $2B.
GM modelled each line with queueing theory to find the bottleneck, then fixed it — saving over $2B across 30 plants.
General Motors
the move
General Motors build cars on long lines, and the throughput of such a line is limited by its most constrained station and the buffers on either side of it, not by the average rate of the whole line. For years the obvious response to a plant that couldn't meet its target was to add capacity, shifts or machinery broadly — an expensive and often wasteful answer.
GM built algorithms that estimate throughput performance with queueing theory and simulation, identify bottlenecks, and decide how much buffer to allocate. Paired with real-time data collection on the plant floor, the process let managers see exactly which station was holding up the line and how much buffer would relieve it.
The approach was rolled out globally and repeatedly. GM said it saved over $2 billion through improved productivity at 30 assembly plants in 10 countries, and a 25% improvement in productivity over six years. It was cited by executives as a key enabler of the company's Global Manufacturing System.
The project won the 2005 INFORMS Franz Edelman Award. Its value was not a single clever model but a repeatable method for turning a plant-floor problem into a measurable, targeted investment decision.
why it works
- Line throughput is set by the bottleneck, so fixing it gives the largest gain per dollar.
- Queueing models quantify how much a given buffer will actually raise output.
- Real-time data makes the model trustable and repeatable across plants.
- Targeting one station avoids spending on machinery that was never the constraint.
what transfers
When a complex system is under-performing, spend your money on the one constraint that holds everyone up, not evenly across the whole system.
what came after
The throughput method became part of GM's manufacturing playbook and was applied across its global assembly network. It demonstrated to heavy industry that predictive, analytical models could direct capital where it genuinely paid off, and it was held up as a model of operations research delivering measurable returns at scale.
references
spotted an error? The archive wants to know.