Guide
How Do You Check That a CNC Supplier Can Really Make 15,000 Parts a Year?
Whether a CNC supplier can really make 15,000 parts a year is a question of arithmetic you can check, not of reputation: it comes down to how many spindle hours the part needs, how many hours the supplier can honestly offer, and whether a timed run shows the part leaving the machine at the rate the quote assumes.
By Saola Manufacturing engineering team · Published · Updated
On this page
Why does “no problem” not prove that a supplier has the capacity?
Because a yes costs the supplier nothing at quoting stage, and a machine list shows what could be made, not what is free for your part. Capacity is a number of hours, and you can ask for the hours.
Every shop quoting your part will say it can handle the volume, and most are sincere. The risk is not dishonesty but unexamined confidence: a quote is built on the cycle time of one machine, with no check of who else is already using it. A machine list shows ownership, not availability, and a line such as “eight CNC lathes” says nothing about how many hours are already sold.
The way out is to replace the yes with three numbers the supplier can show and you can recompute: the spindle hours your part needs, the scheduled hours available on the named machines, and the output rate the shop has actually achieved on similar work. If the supplier cannot produce these quickly, that is itself information.
How do you convert 15,000 parts a year into spindle hours?
Multiply cycle time by annual volume for the raw machine hours, add setup time for each lot, then divide by the OEE the shop has actually demonstrated. Divide that total by the scheduled hours one machine offers to get the number of machines.
OEE, overall equipment effectiveness, is availability multiplied by performance multiplied by quality (OEE.com, 2026). It is the share of planned production time that is truly productive. Dividing raw hours by OEE therefore turns an ideal figure into the hours of machine time you must actually buy. The table shows the method with illustrative inputs; replace them with the quoted cycle time and the supplier’s own figures.
Note what the method does not include. It assumes the part runs on one machine type, in one setup, and that the cycle time is the cycle time at the spindle, not the time on the quote sheet. Parts with a second operation, a different machine for each side or a long wait for a fixture need a line for each step, and the slowest step sets the capacity of the whole route.
| Step | Example input | Result |
|---|---|---|
| Cycle time per part (spindle in cut plus load and unload) | 4 min | 15,000 x 4 min = 1,000 h raw machine time |
| Setups per year (lots) and setup time each | 6 lots, 3 h each | 18 h added |
| Demonstrated OEE on similar parts | 60% | (1,000 h + 18 h) / 0.60 = about 1,700 h of scheduled time |
| Scheduled hours per machine per year, one shift | 8 h x 250 days = 2,000 h | About 0.85 of one machine, with nothing else on it |
| Existing load on the named machine | Example: 70% already committed | Free hours: about 600 h, so the part does not fit on that machine |
| Result | Needs a second machine, a second shift or lower cycle time | Ask which one the supplier plans |
Which three facts make the supplier’s number true?
The machines assigned to your part by name, the load already committed to them, and the number of shifts they run. Without all three, a promised capacity cannot be checked.
Ask for these in writing, per operation. First, machines assigned: make, model and spindle or axis count, because a part quoted on a multi-axis machine cannot be moved to a simpler one without changing the cycle. Second, existing load: the percentage of scheduled hours already committed to other customers over the next year, and whether the shop will keep a reserve for your part. Third, shifts: how many per day and per week, and whether the cycle assumes unattended running, which depends on the fixture, the tool life and the inspection that stops a run. Public buyers apply the same test: a federal prospective contractor must be able to meet the delivery schedule “taking into consideration all existing commercial and governmental business commitments” (FAR 9.104-1(b)).
Then add the constraints outside the spindle: the inspection capacity needed to check the sampling plan you specify, outside processes such as heat treatment or finishing that are arranged through partners, and material supply. A shop with free machines and no inspection time still cannot ship. The related guides on reviewing a CNC quote and supplier scorecard metrics cover how to read the remaining parts of a quote and how to track delivery after award.
Which OEE figure should you believe?
Only a figure the supplier can show from its own records on a similar part, with the loss categories it counts written next to it. OEE is not standardized across industries, so a number without its definition is not comparable.
A general reference notes that the way OEE components are defined is not the same across all industries, and that a single headline “world class” target is often misleading (Wikipedia, 2026). The same reference points out that standard OEE is measured against scheduled time, while the broader TEEP measure uses calendar time, so make sure the hours you divide by and the OEE you use are on the same basis.
Performance is the loss most often overstated. One OEE software vendor reports a gap of three to eight points between declared and measured performance at the sites it has measured (TEEPTRAK, 2026); treat that as one vendor’s observation, but let it justify asking for measured data. Useful evidence is a machine-monitoring export or a shift log for one month on a part of similar size and complexity, showing planned time, stops and good output.
How do you design a run-at-rate trial?
Run the real part, on the named machine, with production tooling and the normal operators, for a defined window, and divide good parts by the theoretical output at the quoted cycle time. A pass at the agreed threshold turns the quoted rate from a claim into a demonstrated rate.
Run at rate is the automotive name for this check. One explanation describes it as verifying that a line can meet its required output using production tooling, staffing and standard processes, with a full shift or a defined multi-hour window rather than a short burst, and says the pass threshold is set by the customer or program team (FlowFuse, 2026). It states the measure as net good output divided by theoretical output, so downtime, changeovers, scrap and rework all reduce it without being counted separately.
For a CNC part, build the trial in four steps. First, set the target: required annual volume divided by the supplier’s scheduled hours gives the hourly rate that the schedule needs. Second, fix the conditions: production fixture, production program, gauges the inspection plan will use, and no extra staff that will not be there in series. Third, run for the agreed window and record good parts, rejects, stops and reasons. Fourth, extrapolate honestly: the demonstrated rate multiplied by the scheduled hours you have confirmed is the supportable annual capacity, and the gap to your forecast is the risk.
Repeat the trial when the part moves to a new machine, a new fixture or a second shift. It is also a good point at which to run the first-article check described in the guide on first article inspection with an overseas supplier, since a part that cannot hold tolerance at rate has not proven the rate.
How does a managed network change the capacity question?
It adds a second question: which plant makes the part and which step runs elsewhere. The arithmetic is the same, but it has to be done per route, with the lead CNC plant and any partner steps counted separately.
Saola places turned and milled work at its lead CNC plant and manages partner plants for processes such as casting, stamping and finishing. The same sum applies to such a route: spindle hours per part family against the machines and shifts the CNC plant can name, then the booking steps of any process arranged through a partner, and the slowest of those sets the answer.
For a buyer, the practical consequence is to ask who holds each step and to request the capacity facts for each one. A single quote that covers machining plus a finished surface depends on both. Parts that fit CNC turning or CNC milling can usually be reasoned about from machine hours alone; add the booking step where a partner process is involved. Send the drawing and the annual volume through the quote form and state whether you want a run-at-rate trial before award.
Frequently asked
Is a supplier’s machine count enough to assess capacity?
No. Machine count shows what the shop owns. Capacity depends on which machines are assigned to your part, how much of their time is already sold and how many shifts they run.
What OEE should I assume when I do not have supplier data?
Do not assume one. Ask for measured OEE on a similar part with the loss categories listed. Where it is missing, run the trial and use the demonstrated rate in its place.
How long should a run-at-rate trial last?
Long enough to include normal stops, tool changes and inspection. One source describes a full shift or a defined multi-hour window rather than a short burst (FlowFuse, 2026), and the customer or program team sets the threshold.
Does a run-at-rate trial replace a first article inspection?
No. The trial proves the rate; the first article inspection proves that the part meets the drawing. Run both on the production setup.
Should the capacity commitment go into the contract?
Yes. State the rate, the machines or routes it relies on and the volume the supplier will reserve, so a later load increase is the supplier’s problem to solve, not yours.
Sources
- Run at rate in PPAP and AIAG: what it proves, FlowFuse, 2026
- How to calculate OEE in industrial production, TEEPTRAK, 2026
- Overall equipment effectiveness, Wikipedia, 2026
- Calculating OEE, OEE.com (Vorne), 2026
- FAR 9.104-1, General standards (of contractor responsibility), Federal Acquisition Regulation, FAC 2026-01, 2026