The short answer
Utilization tells you whether the machine was running. OEE tells you whether it was running well.
Both are useful. They answer different questions, and shops get into trouble when they treat them as interchangeable, or when they reach for OEE before they have the data to calculate it honestly.
Here is the difference, with the same machine measured both ways.
What each one measures
Utilization is one number over one number:
Machine Utilization (%) = (Actual Production Time / Available Time) x 100
That is it. Was the spindle turning or wasn’t it. If you want the full walkthrough, see the full walkthrough on the utilization page.
OEE is three numbers multiplied together:
OEE = Availability x Performance x Quality
- Availability = Run Time / Planned Production Time. Did the machine run when it was supposed to?
- Performance = (Ideal Cycle Time x Total Parts) / Run Time. When it ran, did it run at the speed it should?
- Quality = Good Parts / Total Parts. Of what it made, how much was sellable?
Availability is nearly the same idea as utilization. Performance and Quality are the two things utilization is blind to.
The same lathe, measured both ways
Take the same lathe again. One 8-hour shift, 480 minutes. It ran jobs for 336 minutes.
Utilization: 336 / 480 = 70%
Now the same shift with the extra data OEE needs. Say 30 of those 480 minutes were scheduled breaks and planned maintenance, so planned production time is 450 minutes. The lathe produced 200 parts. The ideal cycle time for that part is 1.5 minutes. Eight parts came out scrap.
| Factor | Calculation | Result |
|---|---|---|
| Availability | 336 run / 450 planned | 74.7% |
| Performance | (1.5 x 200) = 300 ideal / 336 run | 89.3% |
| Quality | 192 good / 200 total | 96.0% |
| OEE | 0.747 x 0.893 x 0.960 | 64.0% |
Same machine. Same day. Same operator. 70% utilization, 64% OEE.
The gap between those two numbers is the point. Utilization said the lathe was busy for 70% of the shift and left it there. OEE said that of the parts it made in that time, it ran about 11% slower than it should have, and 4% of the output was scrap. Utilization counted a scrap part exactly the same as a good one.
Why OEE is almost always the lower number
Because it is a product, not an average. Three factors below 100% multiply down fast.
Look at what 90% on each factor gives you:
0.90 x 0.90 x 0.90 = 72.9%
Three individually respectable numbers produce an OEE in the low 70s. This is why the world-class OEE benchmark of 85% is genuinely hard, and why an OEE in the 60s is not the disaster it looks like next to a utilization figure.
| OEE | What it usually means |
|---|---|
| Below 40% | Common in shops measuring for the first time. Substantial recoverable capacity. |
| 40-60% | Typical for a job shop with varied parts and frequent changeovers. |
| 60-85% | Good. Most well-run discrete manufacturers live here. |
| Above 85% | World class, and worth double-checking your ideal cycle times before you believe it. |
If your OEE comes out above 85% on the first try, the most likely explanation is that your ideal cycle time is set to whatever the machine actually did, which makes the Performance factor meaningless.
What data each one costs you
This is the part that decides which metric a small shop should use, and it gets skipped in most explanations.
Utilization needs: job start time, job end time, machine. That’s it. If you schedule work anywhere other than your head, you already have this.
OEE needs all of that, plus:
- An ideal cycle time for every part number, not a guess and not last month’s average
- A part count per job, recorded per machine
- A scrap count per job, recorded honestly by the operator who made the scrap
- A clean split between planned and unplanned downtime
That last list is where OEE programs die in small shops. The formula is easy. Sourcing four new data streams off a floor that isn’t currently recording any of them is not.
Which one to track first
Track utilization first. Almost always.
Utilization is the metric you can start on Monday with data you already have, and for most small shops it will surface the big problem immediately, because the big problem is usually idle time rather than speed. A lathe at 45% utilization does not have a cycle time problem. It has a scheduling problem, and no amount of OEE precision will tell you anything the 45% didn’t already.
Move to OEE when one of these is true:
- Utilization is already healthy (65% or better) and jobs still run late
- You are quoting from cycle times and the quoted times keep missing
- Scrap is significant enough that you are counting it anyway
- A customer or a certification requires OEE reporting
- You have identified a genuine bottleneck machine and want to squeeze it specifically
That last one is the best reason. OEE is expensive to collect, so collect it on the one machine that constrains the whole shop rather than on all twelve.
The trap in the middle
There is a failure mode worth naming: shops that adopt OEE as a target and start optimizing the number instead of the shop.
Because OEE multiplies, the easiest way to raise it is to run long batches of one part on the machine. Fewer changeovers lifts Availability, a settled process lifts Performance, and fewer first-off rejects lifts Quality. The OEE dashboard goes green.
Meanwhile you have built inventory nobody ordered and pushed the jobs customers are actually waiting on to the back of the queue. This is the classic high-mix problem, and it is why high-mix, low-volume shops should be careful about treating OEE as a goal rather than a diagnostic.
A metric is a thermometer. Optimizing the thermometer doesn’t make the room warmer.
Where setup time lands
One question comes up constantly: does setup count?
In utilization, no. Setup is not production time, and counting it inflates your number. In OEE, setup lands in Availability as downtime, so it hurts you there too.
Neither metric gives you credit for setup, which is correct, but neither one shows you the size of the problem either. If setup is eating your shop, look at it directly rather than through either metric.
Getting the first number without a project
You do not need a system to start. Pick your busiest machine, write down when each job starts and stops for one week, and calculate utilization at the end of it.
If you would rather not keep a log sheet, Machestra gives you the utilization side as a by-product of scheduling: every job already carries its machine and its start and finish times, so the number falls out of the board you were already keeping. It does not calculate OEE, because it does not collect cycle times or scrap counts, and a tool that guessed at those would be giving you a confident wrong answer. For real OEE you want machine monitoring wired to the controllers, which is what tools like JITbase do.
Start with the metric you can measure honestly. An accurate 70% beats an invented 64%.
Start free with 3 machines. No credit card, no time limit.