CNC Capacity Planning: Reading a Supplier's Promise Date
A promise date is an arithmetic output, not a courtesy. When a shop says "four weeks," it has usually done a rough version of a calculation involving machines, hours, setups, and orders already in the queue. When a shop says "two weeks" on a job that needs 900 spindle hours, it has done no calculation at all.
Understanding how the arithmetic works lets you sanity-check a date in about two minutes, spot the supplier that is overpromising before you have paid a deposit, and decide when paying to reserve capacity is cheaper than paying for expediting later.
The arithmetic: gross hours, losses, sellable hours
Start with spindle hours. A shop with 12 machining centers running two 8-hour shifts, 26 days a month, has 12 × 16 × 26 = 4,992 gross machine hours per month, or about 1,150 per week.
Then subtract the losses nobody quotes you:
- Planned maintenance: 3-5% of gross.
- Unplanned downtime — tool breakage, alarms, chip conveyor jams: 4-7% on older machines, 2-3% on machines under five years old.
- Changeover and setup: a job running 500 pieces may need 1.5-2.5 hours of setup; a 20-piece job can need the same, and that setup consumes a machine slot like production does.
- Idle time waiting for material, programs, fixtures, or an inspector.
A realistic shop sells 75-85% of gross hours. The table below shows the conversion for a 12-machine shop at a mix of job sizes.
| Line | Basis | Hours per week |
|---|---|---|
| Gross spindle hours | 12 machines × 16 h × 6 days | 1,152 |
| Planned maintenance at 4% | Scheduled PM, calibration | −46 |
| Unplanned downtime at 5% | Alarms, tool failure, warm-up | −58 |
| Changeover, 6 setups per machine at 1.4 h | 84 setups weekly | −118 |
| Material and program waiting at 3% | Inbound stock, first-off proving | −35 |
| Sellable spindle hours | Realistic output | 895 |
| Committed to open orders at 85% | Backlog already sold | −761 |
| Free capacity | What a new order can use | 134 |
134 hours a week across 12 machines is 11 hours per machine. That is one decent-sized job — perhaps 3,000 small parts at 25 seconds of spindle time each — before the shop is at 100% and quoting lead times it cannot meet without overtime.
Utilization: the number that decides your delivery
Utilization is sellable hours divided by gross hours, and it is the single best predictor of whether a promise date holds. At 70%, a surprise rework lot costs you nothing. At 90%, a single machine breakdown pushes everything downstream by three days.
Most shops plan to 80-85% and treat anything above 90% as a warning sign rather than a success. That sounds inefficient until you look at the queue: at 85% utilization, average queue time is roughly five to six times the processing time. At 95%, it is twenty times. This is not a theory of the shop's making; it falls out of how waiting time grows as a system approaches saturation.
So when a supplier quotes two weeks on a part with eight hours of processing, ask what utilization they are running. The honest answer — 82%, with a slot in week two — is more useful than a confident date with no basis.
See also: Scrap Rate in Machining: Where Losses Come From and Who Pays and What Drives CNC Prototype Cost, Line by Line.
Queue, priority, and why your job waits behind a 20,000-piece run
Machine shops schedule by job, not by customer. A 20,000-piece order that occupies a machine for nine days will not be interrupted for a 50-piece job unless someone pays for the interruption. Your 50 pieces sit in the queue and are cut in the gaps.
This is why small orders often take longer than the arithmetic suggests. A 50-piece job needs 6 hours of spindle time and 2 hours of setup; in a busy shop it may wait five working days for a gap. That is a scheduling decision, not a capacity shortage.
Two things shorten the wait. Consolidating small orders into a single release with a common material and finish gives the shop a reason to run them together. And asking for a scheduled slot — a named week — rather than a lead time in days gives the planner something to hold.
There is a second queue too: the finishing queue. Anodizing and plating are often outsourced, and a shop that runs 300 parts a day may send them out in batches twice a week. A job that finishes machining on Monday can sit until Thursday waiting for a truck to the anodizer, then wait again on return. Always ask whether the quoted date includes outsourced finishing, because four days of a three-week lead time can disappear there.
When to book capacity ahead
If your forecast is real, book it. A capacity reservation is a simple agreement: you commit to a monthly spindle-hour volume or a monthly piece volume for three to six months, and the shop holds machine time and material.
What that costs varies. Common structures are a 10-20% premium over the standard price, a take-or-pay commitment where you pay for 80% of the reserved hours whether you use them or not, or a material prepayment that secures the schedule without a premium. In a tight market — and machine time tightens predictably in the four to six weeks before Chinese New Year and again in the late-summer peak — a reservation is cheaper than expediting.
Book when three conditions hold: your forecast has less than 20% variance, your parts share material and fixturing so the shop can plan a run, and your program is longer than the lead time you would otherwise wait.
Reading a capacity claim
A supplier who says "we have 40 machines" is telling you about assets, not availability. Ask three questions and the answer will tell you whether the claim is real.
What is your current utilization, and how is it measured — booked hours against sellable hours, or a gut feel? What is your average queue time for a job released today? And which machine will my part run on, and when is that machine free?
A shop that can answer all three in one email is planning. A shop that answers only the first is selling. Neither answer is dishonest, but only one of them is a promise you can build a production schedule on. Our own how we work page describes the scheduling sequence we use from PO to dispatch, including how we handle slots for repeat programs.
Frequently asked questions
How many hours a week can a 12-machine shop actually sell?
Around 895 out of 1,152 gross hours, once maintenance, downtime, changeover, and waiting are removed. That is 78% of the clock, which is a healthy number for a job shop running mixed batch sizes.
Why does 85% utilization cause long lead times?
Because queue time grows faster than utilization near saturation. At 85% booked, average waiting is five to six times processing time; at 95% it can be twenty times. The last 10% of utilization costs more delivery reliability than it earns in revenue.
Should I pay to reserve capacity?
If your forecast varies by less than 20% and your program runs longer than the standard lead time, yes. Reservations typically cost 10-20% or a take-or-pay commitment, which is usually less than repeated expedite fees and air freight.
Does a bigger machine count mean shorter lead times?
Not necessarily. Asset count says nothing about queue. Ask for current utilization, average queue time for a job released today, and which specific machine your part will run on — and when that machine is free.
Related reading
Send a drawing for a quote and DFM review.
STEP, IGES, X_T, PDF or DWG up to 50 MB. Tell us quantity and finish, and the quote comes back with tolerance and process notes, not just a price.