Key takeaways
  • Capacity utilisation is actual output divided by maximum possible output over the same period.
  • High utilisation at a non-constraint produces inventory, not throughput. Only the constraint's utilisation converts to output.
  • As utilisation approaches 100%, queues and lead times rise sharply. This is a property of variability, not of poor management.
The trade-off nobody warns you about
Above roughly 85%, lead time rises steeply

Any process with variability sees waiting time increase non-linearly as utilisation approaches full. Running everything flat out is how a plant ends up with high utilisation, high inventory and long, unreliable lead times simultaneously.

Calculating it

Actual output divided by maximum possible output for the same period. The difficulty is defining maximum: theoretical capacity assumes no downtime, changeovers or breaks and is unachievable; practical capacity accounts for planned downtime and realistic performance and is the more useful denominator. State which you are using, because the two produce very different percentages.

Why utilisation is a poor universal target

  • Only the constraint converts utilisation into throughput. Everywhere else it converts into inventory.
  • Measuring every station on utilisation encourages overproduction, which is the waste that generates the others.
  • It rewards long runs and discourages changeovers, which lengthens lead time and reduces responsiveness.
  • It makes idle time look like failure, when planned idle time at non-constraints is what protects the constraint.

The queueing effect

Because arrivals and processing times vary, work queues even when average capacity is sufficient. As utilisation rises, the queue grows disproportionately: the difference between 80% and 95% utilisation is not a fifth more work, it is several times more waiting. Any operation promising short lead times needs deliberate spare capacity, and that spare capacity is a design decision rather than waste.

Where to target high utilisation

At the constraint, and only there. Protect it with a buffer of work so it never starves, do not schedule changeovers on it when they can be done elsewhere, and move inspection off it. Everywhere else, the correct target is enough capacity to keep the constraint fed with reliable lead time.

What to measure instead

Throughput, lead time, and adherence to schedule tell you whether the operation is working. Utilisation is a useful diagnostic for capacity planning and a poor performance target, because what it encourages is not what you want.