Precision OEM CNC Machining: 7 Checks Before You Release the PO

• 1
  Key takeaways
• 2
  Read the print at both levels before you talk to anyone
• 3
  Match the machine to the geometry, not the part envelope
• 4
  Verify the quality system, then ask for the records
• 5
  Ask how finishing and material traceability interact with tolerance
• 6
  Test engineering support and data handling before the first order
• 7
  Compare total cost, not unit price
• 8
  Protect the schedule with lead-time transparency
• 9
  Step by step: auditing a precision OEM CNC machining supplier
• 10
  Which machine class fits which part

A drawing carries two kinds of precision. Feature precision is the number next to a dimension: Ø25.00 ±0.02 mm, a flatness of 0.01 mm, a surface finish of Ra 0.8–1.6 μm. Process precision is how repeatably the shop hits that number across the whole order, on the fifth part and the five hundredth. Most disputes come from the gap between the two.

Before you request quotes, mark which dimensions actually control function. On a housing, it is often the bore and the mounting face. On a manifold, it is the port spacing and the sealing surface. Everything else can move a little. Sending a drawing where every dimension is tight forces the shop to quote slow, and it hides the two features that matter.

Then check the tolerance against the size of the feature. A ±0.005 mm tolerance on a 10 mm bore is a grinding or fine-boring job. The same callout on a 300 mm length is a different problem, because thermal expansion of aluminium is roughly 23 μm per meter per °C. A 3 °C shop swing over that length eats most of your budget before the tool touches the part.

If the drawing uses true position or profile of a surface, the datum scheme has to be manufacturable. Datum targets that sit on a rough cast surface, or a primary datum that only exists after a secondary operation, will force the shop to invent a setup. Ask how they plan to hold the part for each operation. A good answer names the fixture, not the machine.

Maximum part size is the easiest spec to compare and the least useful on its own. A shop with a 4,000 mm machine can hold a long extrusion, but that says nothing about a part with five angled faces and two intersecting bores. The question is whether the equipment portfolio covers your geometry class.

For simple prismatic parts with holes and pockets on one or two faces, a three-axis mill is the right tool and the cheapest one. When a part needs four or more faces machined without re-fixturing, a four-axis or five-axis center removes setups and the position error that comes with each one. On a part with a true position of 0.02 mm across two perpendicular faces, setup count is usually the dominant error source.

Deep pockets and thin walls change the answer again. A pocket deeper than four times its cutter diameter needs a long-reach tool, which deflects. Shops that run high-speed toolpaths with reduced radial engagement handle this better than shops that take full-width cuts. Ask what radial stepover they plan to use. Under 10 percent of cutter diameter is a reasonable answer for a deep, tight-tolerance pocket.

Rotary table work is its own category. A Ø400 mm rotary table with a five-axis center lets the shop machine a valve body or an impeller in one setup. If your part has features on a curved or conical surface, that is the setup you want, and it is worth asking directly whether they have it in-house or plan to subcontract.

A certificate on a wall tells you an auditor visited. It does not tell you what happens on a Tuesday afternoon when a tool breaks mid-run. ISO 9001:2015 covers the general quality system. IATF 16949:2016 adds automotive requirements like PPAP and traceability. ISO 13485:2016 is the medical device standard. ISO 27001:2022 covers information security, which matters if your drawings are prototypes for an unreleased product.

The useful follow-up is to ask for the inspection plan for a part like yours. Look for three stages: incoming material check, in-process monitoring, and final inspection. A shop that only inspects at the end is sorting good parts from bad. A shop that probes in-process is correcting before the part goes out of tolerance. The second is cheaper for both sides.

Ask what measurement equipment covers your tolerance. A ±0.005 mm callout needs a CMM or a high-accuracy optical system, in a temperature-controlled room, with a calibration record. Calipers and micrometers read to 0.01 mm at best and depend on operator feel. If the tightest dimension on your print is 0.01 mm, and the shop's only answer is a micrometer, the tolerance is not really being verified.

Request a first article inspection report on the first production part. It should list every dimension on the print with the measured value, not just a pass or fail. On a complex part, that report is the fastest way to find a misread drawing before 500 pieces are made.

Finishing is not cosmetic when tolerances are tight. Anodizing builds roughly 5 to 25 μm depending on the process, and hardcoat anodizing builds more. Electroless nickel adds a more uniform layer, typically 10 to 25 μm. That thickness lands on every surface, including bores and threads. A Ø6.00 +0.02/0 bore that gets hardcoat anodized after machining may come back undersized.

The fix is to state pre-finish and post-finish dimensions on the drawing, or to mask the critical features. Ask the shop how they handle this. If the answer is a shrug, plan on a rework loop. If the answer names a masking plan and a pre-plate dimension, that is a shop that has done this before.

Surface finish callouts deserve the same treatment. As-machined aluminium typically lands at Ra 1.6–3.2 μm. A Ra 0.8–1.6 μm callout usually needs a finer finishing pass or a light polish. Ra 0.2–0.8 μm is a lapping or fine-polish territory and drives cost fast. Only call it out where a seal or a sliding surface needs it.

Material traceability is a separate check. For aerospace, medical and automotive parts, ask for mill test reports that tie the heat number to the material lot. For stainless and titanium, that report is the only proof that the 316L is actually 316L and not a lookalike. For plastics, ask for the resin grade and the supplier, because POM and PEEK behave very differently across brands.

The fastest way to judge a supplier is to send a real drawing and watch what comes back. A quote that is only a price and a lead time is a warning sign. A useful reply includes a DFM note: a wall thickness that will chatter, a radius that needs a smaller tool than planned, a tolerance that will need a second operation. That note is worth more than a small price difference.

Response time matters at the prototype stage. A 12-hour turnaround on quotation and DFM analysis keeps a design cycle moving. If the first reply takes a week, the production replies will not get faster. Ask what happens when the drawing changes mid-project, because on OEM work it usually does.

Data handling is the quiet risk. Unreleased product drawings are the most valuable file a company owns. Ask where files are stored, who can open them, and whether the shop will sign an NDA before the RFQ. A shop that holds ISO 27001:2022 has a documented information security system, which is a better answer than a verbal promise.

You can also test the engineering bench with a single question: what would you change on this part to make it cheaper without losing function? A supplier that answers with a specific change has engineers who read the drawing. One that answers with a discount has salespeople.

Unit price is the number at the bottom of the quote, and it is the least reliable way to compare two suppliers. Add the cost of a rework loop, a late shipment that stops an assembly line, and an engineering change that has to be re-quoted. Those costs land on you, not the supplier.

The biggest hidden cost is a tolerance the shop cannot hold. If a shop quotes a low price and then delivers parts at the edge of the tolerance band, your incoming inspection becomes the sorting operation. That is expensive, and it gets worse as volume rises. A slightly higher unit price from a shop that holds the center of the band is usually cheaper overall.

Batch size changes the math. Low-volume, high-mix work needs fast setups and flexible scheduling. High-volume work needs consistency and process control. A shop optimized for one will struggle with the other. Ask what share of their work is high-mix low-volume versus long-run production, and check that it matches your program.

Shipping and finishing are line items people forget to compare. A domestic finisher plus a separate assembly step adds handling and lead time. If one supplier can machine, finish, and inspect under one roof, the parts move fewer times and fewer things get lost. Fewer handoffs also means one party is accountable when something is out of spec.

Lead time is not one number. It is a chain: quote and DFM, material purchase, programming and fixturing, machining, finishing, inspection, and shipping. Ask which link is longest for your part. On a titanium part with a specialty alloy, material purchase may dominate. On a complex five-axis part, programming and the first-article inspection often dominate.

For prototype work, the chain compresses. If the shop stocks your material grade, production can start within 24 hours of approval, and parts can ship in 3 to 5 days. That only holds if the material is on the shelf. Ask whether the listed alloys are actually stocked or ordered per job.

For production, ask about the late-delivery rate and what causes it. A shop that tracks its own on-time performance has data to share. A shop that does not track it cannot improve it. Historical performance below 2 percent late is a reasonable target for a stable program with released drawings.

Finally, agree on what happens when something slips. A supplier that flags a problem two weeks early gives you options: re-sequence the build, air-freight a partial batch, or accept a revised date. A supplier that flags it on the ship date gives you none. Put the notification window in writing.