Who this checklist is for

If you're sourcing a CNC lathe—new or lightly used—and you don't have a dedicated machine tool engineer on staff, this is for you. I've been the person signing off on incoming equipment for a mid-size job shop since 2019. In that time I've rejected roughly 2 out of every 11 machines that arrived on our floor. Most of those rejections could have been caught earlier with the seven checks below.

They take about three hours to run through properly. Skip them and you'll spend that time arguing with a vendor instead.

1. Confirm the written spec actually matches what you need

Not what the sales sheet says the machine can do. What your parts require.

Pull three or four representative parts from your current work. Write down the tightest tolerance on each one. Then check that tolerance against the machine's stated positioning accuracy and repeatability—not its "best case" numbers from a brochure, but the figures that appear in the acceptance test section.

Here's what most first-time buyers don't realize: the positioning accuracy printed on a spec sheet is often measured under ideal lab conditions—constant temperature, no load, brand-new machine. Your shop floor is not that. Ask for the operating-condition spec, or at minimum the ISO 230-2 test results from a machine they've already shipped.

2. Check spindle runout, not just spindle speed

Max spindle RPM is the number everyone quotes. Spindle runout is the number that determines whether your finish passes inspection.

Ask for a runout measurement at the spindle nose, taken cold and after a 30-minute warm-up. If the vendor can't produce that number, that's not a red flag by itself—but insist on measuring it before you accept delivery. A dial indicator on a test bar is enough. Industry consensus puts acceptable radial runout for a general-purpose lathe spindle in the 0.005 mm range or better; precision work needs tighter. If they push back on you checking this yourself, walk.

3. Verify the control system, not the sticker

The control is the machine's brain, and it's also the thing parts are hardest to source ten years from now.

Three questions, in this order:

  1. Which control model, exactly? "Fanuc" is not an answer. "Fanuc 0i-TF Plus" is.
  2. Who supports it locally? If the answer is "the manufacturer's regional office," get the regional office's name and a phone number you can call.
  3. What's the spare parts situation? Boards, drives, encoders. Not consumables—electronics.

I went back and forth between two suppliers in Q2 of last year for this exact reason. One offered a lower base price with a control I recognized but couldn't get serviced within 300 km. The other was 8% more with a control supported by a shop two towns over. I took the second one. It cost us more on paper. It saved us a two-week downtime event in November.

4. Ask what "in stock" actually means

When a supplier says a machine is in stock, ask a follow-up: "Where, physically?"

The answer matters. "In our warehouse" is one thing. "Available to ship from the factory" is another. "In production for another customer, but we can redirect it" is a third—and it usually means a longer lead time than the first two.

The assumption is that in-stock machines cost more because they're in demand. The reality is often the opposite—in-stock machines at a good price are exceptions, not the rule, and the supplier quoting you a great in-stock deal may be counting on you not asking where the machine is.

5. Field-check the service and spare parts network

This is the step most buyers skip. It's also the one that costs the most when it goes wrong.

Ask the supplier for two things: the name of the nearest authorized service technician and how many spindles or ball screws they've replaced in the last 12 months for this model. A vendor who sells a lot of a particular machine will have spindles on the shelf. A vendor who sells one or two a year will not.

Industry rule of thumb: if the spare parts lead time exceeds two weeks for the spindle or ballscrew, you need a plan B for critical jobs. That plan B is either a second machine or a subcontractor relationship. Factor that cost in before ordering.

6. Call a real customer from the last 18 months

Not a reference list the supplier curates. Ask for three customers who bought this exact model in the last 18 months.

Then ask those customers three questions:

  • Did the machine pass acceptance on the first visit?
  • How many unplanned downtime hours in the first year?
  • If they had to buy again, would they buy from the same supplier?

Customer #2 is the one who tells you the truth. Customer #1 is usually a friend of the sales rep; customer #3 is often too far past the pain to remember it clearly.

7. Structure payment terms so you keep leverage

I'm not 100% sure what the right split is for every project—it depends on machine size and shipping distance—but the principle is consistent: hold at least 20% until acceptance testing passes on your floor.

Standard splits I see working well:

  • 30% deposit
  • 40% before shipment
  • 30% after on-site acceptance

Suppliers who push for 100% before shipping are usually doing it for cash flow reasons, not because they're dishonest. But once your money is gone, your negotiating position is gone too. Keep a slice back.

What people get wrong

Three mistakes I see repeatedly, even from experienced buyers:

Trusting the spec sheet. Spec sheets describe a machine under conditions you will never operate in. Your acceptance test describes your machine under your conditions. Write the acceptance criteria into the contract, not just the spec.

Skipping the warm-up measurement. A cold machine measures differently from a warm one. Any runout or accuracy number you get should specify which state the machine was in. "Runout: 0.003 mm" is meaningless without that.

Treating the deposit as the commitment. The deposit secures the order. The acceptance test secures the machine. They are not the same event, and suppliers who conflate them are telling you something about how the rest of the relationship will go.

Run through all seven and you'll take maybe two extra weeks on the front end. That's cheaper than one rejected machine on the back end.