Your number is going to look bad, and some of that is fine
Most shop owners measure utilization for the first time, get something in the 50s, and assume they have found a disaster.
Sometimes they have. Often they have just discovered what a job shop is.
A job shop takes custom orders. Different parts, different quantities, different routings, arriving on a schedule nobody controls. That business model has a utilization ceiling built into it, and it sits well below what a shop running the same part all week can reach. If you are benchmarking against a number you read in an article about automotive production, you are comparing two different businesses.
The useful split is not high versus low. It is structural versus recoverable.
First, make sure the number is real
Before you split the causes, the measurement has to be honest. A utilization number built on the wrong denominator sends you after the wrong problem, and the most common version of that is a shop counting setup as production and concluding it has nothing to fix.
The formula is straightforward:
Machine Utilization (%) = (Actual Production Time / Available Time) x 100
- Available Time is the total time the machine could be running. If you run one 8-hour shift, that is 480 minutes. Subtract planned maintenance or scheduled downtime if you want to be precise.
- Actual Production Time is the time the machine is actively running a job. Not setup. Not waiting. Actually cutting, forming, printing, or whatever that machine does.
Example: Your lathe is available for 8 hours (480 minutes). Today it ran jobs for 5 hours and 36 minutes (336 minutes). Setup took 1 hour. It sat idle for 1 hour and 24 minutes between jobs.
Utilization = 336 / 480 = 70%
That means 30% of your lathe’s available time was not producing parts.
What counts and what doesn’t
This is where shops get it wrong. Be honest about what counts as production time:
Counts as production time:
- Actively running a job (spindle turning, laser cutting, press forming)
- Automated cycles where the machine runs without an operator
Does not count:
- Setup and changeover time
- Waiting for material
- Waiting for an operator
- Waiting for quality inspection or approval
- Unplanned downtime (breakdowns, tooling failures)
- Idle time between jobs
Some shops inflate their numbers by counting setup as “production.” Don’t do that. The whole point of tracking utilization is to see reality, not a comfortable fiction.
Note what this number cannot tell you. Utilization says the machine was running. It says nothing about whether it was running well, at speed, making good parts. That is a different metric with a different collection cost, and the two get confused constantly: utilization vs OEE sets out which one is worth your time first.
The goal is not 100%
The goal is not 100%. A shop running at 100% utilization has zero flexibility. No room for a rush order. No time for preventive maintenance. No buffer when something goes wrong. And something always goes wrong.
Compare against yourself
The most useful benchmark is your own number from last quarter. Shop-to-shop comparison is close to meaningless, because the part mix, the run lengths, the machine types, and the honesty of the measurement all differ. Your own trend line controls for all of that automatically.
Track it monthly per machine, not weekly and not shop-wide. Weekly is too noisy in a job shop, where one big job distorts everything. Shop-wide averages hide the machine that is actually constraining you underneath the ones that are fine.
The structural causes
These come from the kind of work you take. You do not fix them with better scheduling, and trying to will cost you customers.
Setup dominates when the mix is high
This is the big one. If your average job runs 40 minutes and your average setup takes 45, then even a perfectly scheduled machine spends more than half its day not producing. Nothing about that is a scheduling failure. It is arithmetic.
A production shop amortises one setup across 4,000 parts. You amortise one setup across 12. Same machine, same operator, completely different utilization, and the difference is the order book rather than the discipline.
Order arrival is not smooth
You cannot run a machine on work you have not been given yet. Job shops get feast and famine, sometimes inside the same week, and the famine days pull the monthly average down no matter how well you sequence the feast days.
The operator is a real constraint, not just the machine
Utilization measures machines, but a lot of shops are actually operator-limited. If one person is the only one who can run the big mill, that mill’s utilization is capped by that person’s availability, their holidays, and whatever else they are needed for. Measuring the machine and ignoring the person hides the actual constraint. This is where scheduling around operator skills matters more than machine capacity.
Some machines exist for capability, not volume
Most shops have one machine that runs 20% of the time and has to be there anyway, because when a job needs it, no other machine will do. That machine drags your shop average down and should. Averaging it in with your workhorses produces a number that describes nothing.
Track that machine separately, or accept that the average is meaningless.
The recoverable causes
These are the ones worth your attention, because they cost you hours you already paid for.
Idle gaps between jobs
The machine finishes at 10:40. The next job starts at 11:25. Nothing was wrong. Nobody knew the machine was about to free up, the material was still in the rack, and the program had not been pulled.
Those 45 minutes cost exactly as much as a breakdown and get noticed far less, because nothing broke. In a shop with five machines, three gaps a day is close to eleven hours a week.
Scheduling conflicts that cascade
Two jobs are pointed at the same machine at the same time. One of them waits. If that waiting job was feeding a downstream operation, that machine now waits too, and the delay spreads outward.
One conflict does not cost you one machine’s time. It costs you a chain of it. Catching double-bookings before they reach the floor is the highest-return fix on this list because of that multiplier.
Material that arrives after the job is scheduled
A job scheduled for Tuesday on material that lands Thursday does not produce a Tuesday. It produces an idle machine and a reshuffle. If material availability is not visible at the moment you schedule, you are scheduling fiction.
Setup that is bigger than it needs to be
Setup being structural does not mean the size of it is fixed. Batching jobs that share a fixture, pre-staging tooling and material before the current job ends, and standardising workholding all take real minutes out of a number you will pay every single job. There is usually more recoverable time in setup and changeover than anywhere else in a high-mix shop.
Nobody knows what is running
If your team asks you what is on machine 3 instead of looking, the schedule lives in your head, which means it stops when you do. A shared scheduling board removes you from the critical path of every transition.
How to tell which is which
Run this for one week on one machine. For every stretch the machine was not producing, write down the reason in one of two columns:
| Structural | Recoverable |
|---|---|
| Setup for a different part | Waiting for material that was in the building |
| No work available that machine could do | Waiting because another job was on the machine |
| Only operator was unavailable | Waiting because nobody staged the next job |
| Machine only fits a narrow job type | Waiting for someone to notice it was free |
At the end of the week, total both columns. The recoverable column is your actual opportunity, and it is usually a good deal larger than people expect before they write it down.
The structural column is not a to-do list. It is a description of your business.
What to do with the answer
If recoverable time is small and your utilization is still low, you do not have a scheduling problem. You have a sales problem, and buying software will not fix it.
If recoverable time is large, that is the cheapest capacity you will ever get, because you are already paying for it. Every hour you take out of that column is an hour of machine time that costs you nothing extra.
That second case is what Machestra is built for. Capacity is finite on the board, so a machine holds one job at a time and an operator is in one place at a time, and the double-booking that would have cascaded into three idle machines gets blocked when you try to make it. Materials attach to the job, so a job scheduled on steel you do not have is visible before Tuesday rather than on Tuesday. Everyone sees the same board, so the machine coming free at 10:40 is not news to the person who could have staged the next job at 10:20.
None of that touches the structural column. It is not supposed to.
Start free with 3 machines. No credit card, no time limit.