Capacity Utilization = Actual Output / Available Capacity × 100%. Available capacity is usually taken as effective capacity, theoretical minus planned losses. A machine that produced 800 units against 1,000 units of available capacity has 80% utilization.
Pushing utilization toward 100% at every station looks efficient locally and creates chaos systemically. High utilization at a non-bottleneck produces inventory the bottleneck can't consume, the definition of overproduction. High utilization at the bottleneck without a matching increase in output means the bottleneck was already the constraint and the resource was already fully loaded. The metric rewards busyness, not throughput.
Instead of "how utilized is this resource?" ask "is this resource keeping up with pull from downstream?" A resource that produces exactly what the next step needs, when it needs it, is doing its job, even if its utilization looks moderate. A resource running at 95% while downstream sits waiting is not.
Capacity utilization is worth measuring, but it must never become the target. Optimizing it in isolation produces the Efficient Islands failure mode. The system-level targets, on-time delivery, lead time, flow efficiency, should always dominate the local ones.
