Machine utilization rate is the share of available time a machine spends actually running, calculated as run time ÷ available time × 100. A machine that runs 6.5 of 8 scheduled hours has an 81 percent utilization rate. It answers one narrow question, was the machine on?, and nothing about whether it ran well.
That narrowness is the whole point and the whole danger. Utilization is easy to measure and easy to misread as productivity. A line can post 95 percent utilization while making scrap at reduced speed, and a plant can push a non-constraint machine to full utilization and only succeed in burying the floor in work-in-process. This post gives the formula and the time bases, separates utilization from OEE, walks the calculation step by step, and explains why chasing utilization on the wrong machine costs you money.
What is machine utilization rate and how do you calculate it?
Machine utilization rate is run time divided by available time, expressed as a percentage. Run time is the clock time the machine was actively producing; available time is the time it was scheduled or expected to be available. The formula is: utilization = (run time ÷ available time) × 100.
The answer depends entirely on which "available time" you choose as the base, and this is where most utilization arguments start:
- Against scheduled time run time ÷ scheduled (staffed) hours. This is the common shop-floor version and is close to the availability idea, since it asks how much of the planned shift the machine ran.
- Against calendar time run time ÷ all 24/7 hours. This is the strictest base and matches the idea behind TEEP exposing unstaffed and no-demand time as unused capacity.
- Against loading time run time ÷ planned production time. At this base, utilization is essentially the availability factor of OEE.
Because the same machine yields three different utilization numbers depending on the denominator, you cannot compare utilization across plants unless the base is identical. State the base every time you quote a figure, or the number means nothing.
By the numbers. At the macro level, the Federal Reserve's G.17 release put U.S. total industry capacity utilization at 76.3 percent in December 2025, about 3.2 points below its 1972–2024 average, and manufacturing lower still near 75.8 percent (Federal Reserve, Industrial Production and Capacity Utilization G.17). That national figure measures output against sustainable capacity, not machine on-time, but it frames the point: even at the industry level, a quarter of capacity sits idle, and plant-level machine utilization is what rolls up into it, see capacity utilization for the plant view.
Why is machine utilization not the same as OEE?
Because utilization only confirms the machine was running, while OEE confirms it ran well. A machine can post high utilization and low OEE at the same time: running slowly, producing scrap, or operating below rate all leave run time high while dragging performance and quality, and therefore OEE, down. Utilization is a one-dimensional "on or off" measure; OEE is three-dimensional.
Put concretely: a filler that runs the full shift but at 70 percent of rated speed and throws 5 percent scrap shows near-100 percent utilization and maybe 66 percent OEE. Utilization applauds the busy machine; OEE asks what the busy machine actually produced. If you optimize for utilization alone, you reward motion, not output. That is why utilization belongs beside OEE on the board, not instead of it.
| Metric | Numerator | Denominator | Answers |
|---|---|---|---|
| Machine utilization | Run time | Scheduled time | Was the machine on? |
| Availability | Run time | Loading time | Was it up when planned? |
| OEE | Fully productive time | Loading time | Did it run fast, up, and clean? |
| TEEP | Fully productive time | Calendar time | How much of the asset do we use at all? |
How do you calculate and improve machine utilization step by step?
Treat utilization as a diagnostic you compute the same way every time, then act on only where it matters. The routine below keeps the number honest and points the improvement at the right machine.
- Pick and record the time base. Decide whether available time is calendar, scheduled, or loading time, and write it down. Every later comparison depends on this choice being fixed.
- Capture run time at the source. Read actual running minutes from the machine, not from a shift estimate. Reconstructed run time hides short stops and inflates utilization.
- Divide and label. Compute run time ÷ available time × 100 and label it with the base, for example "82% of scheduled time."
- Ask whether the machine is a constraint. Before improving it, check whether this machine sets the pace of the whole line. Utilization only translates to more sellable output on the constraint.
- Improve utilization only on the constraint. On the bottleneck, attack the reasons run time is low, changeovers, waits, breakdowns, because every recovered minute becomes throughput. Elsewhere, high utilization is optional and often harmful.
Why does chasing utilization on a non-constraint build WIP?
Because a non-constraint machine that runs flat out produces parts faster than the constraint can consume them, and the surplus has nowhere to go but into a queue. Throughput of the whole line is set by the slowest necessary step, so pushing an upstream machine to 100 percent utilization does not add a single sellable unit, it just piles work-in-process in front of the bottleneck.
Little's Law makes the cost exact: at fixed throughput, lead time = WIP ÷ throughput, so the extra WIP you create by over-running a non-constraint stretches lead time in direct proportion, ties up cash, and hides quality problems inside a bigger queue. See Little's Law in manufacturing for the arithmetic and Theory of Constraints for why the constraint governs. The practical rule from constraint thinking is blunt: a non-constraint should be idle whenever running it only builds inventory. Idle time on a non-bottleneck is not waste; forced utilization there is.
What is a good machine utilization rate?
There is no single target, because the right number depends on the time base and on whether the machine is a constraint. Against scheduled time, well-run discrete operations often sit in the mid-to-high 80s and continuous process lines higher, but a healthy plant will deliberately run some non-constraint machines well below that. A low utilization number on a non-bottleneck is not a problem to fix; it is capacity you are wisely choosing not to over-run.
Judge utilization by context, not by a universal threshold. Ask three questions: Is this the constraint? Is the base consistent with how I measure other machines? And is the machine's output good, or just busy? For the machines that set your pace, low utilization traces straight to downtime and changeover time, which is where machine downtime tracking and cycle time analysis do the real work. Utilization tells you a machine is quiet; those metrics tell you why.
How does machine utilization relate to asset utilization and availability?
Machine utilization, availability, and asset utilization are three views of the same run time measured against wider and wider bands of time. Availability divides run time by loading time, so it isolates unplanned downtime during hours the machine was expected to run. Machine utilization usually divides by scheduled time, so it also carries the cost of running fewer shifts than you could. Asset utilization stretches the band widest, comparing actual good output to the theoretical maximum across all calendar time, which folds in demand, staffing, and capacity decisions a single machine never controls.
The three rarely agree, and the spread between them is diagnostic. When availability is high but machine utilization is low, the machine is reliable but under-scheduled, a demand or planning question, not a maintenance one. When utilization is high but OEE is low, the machine is busy but not effective. Reading the gaps tells you which lever to pull, and pulling the wrong one is how plants spend maintenance money on a machine whose real problem is an empty order book.
How does live machine data turn utilization into throughput?
Utilization is only as trustworthy as the run-time number underneath it, and end-of-shift estimates systematically overstate run time by missing short stops. Direct machine monitoring timestamps every state change, so run time is measured, not remembered, and the utilization rate on each machine is drawn from the machine itself. Layer that with reason codes and you learn not just that a machine was idle but why.
From there the improvement is targeted: raise run time on the constraint, let non-constraints breathe, and watch line throughput climb without new capital. That is the story in the CLS case study where connected-floor data replaced clipboard estimates; run the utilization and OEE math for your own line with the OEE calculator and roll the numbers up into plant-level views with throughput and capacity utilization.