Find high-quality CNC machining online

• 1
  How to find high-quality CNC machining online: what a quote really tells you
• 2
  Tolerance is a capability, not a promise
• 3
  Why the machine list matters more than the website design
• 4
  Inspection is the part of the process you cannot see
• 5
  Lead time is a process signal, not a sales promise
• 6
  Signals that a supplier is a broker, not a shop
• 7
  Which sourcing route fits your part

A CNC quote is not a price. It is a claim about geometry, tolerance, and process. Two suppliers can return the same number for a bracket and mean completely different things: one will mill it from 6061-T6 bar in two setups on a three-axis machine, the other will fixture it once on a five-axis center and hit the bearing bores in the same operation. The price looks similar. The parts do not behave the same way.

So when you try to find high-quality CNC machining online, the first useful question is not "how much" but "how will you hold this feature". A supplier who answers that in one sentence is usually a real shop. A supplier who replies with a discount code is usually a broker.

The gap shows up later. A bore that reads 20.00 mm on the CMM at 20 °C can still fail in assembly if the mating part was machined on a different machine with a different thermal history. Tolerances are not a number on a drawing. They are the result of a setup, a toolpath, a measuring method, and a temperature.

This page gives you six checks you can run before you commit. None of them require a factory visit. All of them can be answered by email or by a DFM report.

• 1
  Setup countFewer setups usually means tighter true position.
• 2
  Measuring methodAsk which instrument and at what temperature.
• 3
  Process routeBar, plate, or casting changes the whole plan.

Every shop can write "±0.01 mm" on a website. The useful question is which machine will run your part and whether that machine can hold the number across a full batch, not just on the first article. GreatLight works to ±0.005 mm (±0.0002 in) on qualified features, but that number only applies to features we have discussed and measured.

Geometry decides most of it. A 40 mm deep Ø8 mm bore with a 6:1 depth-to-diameter ratio behaves very differently from a Ø8 mm through hole in a 5 mm plate. The long bore needs a stub drill, a boring head, or a helical path with a small stepover, and the tool deflection grows with the square of the length. A supplier who quotes both at the same rate has not read the drawing.

Surface finish is a second capability layer. Ra 1.6–3.2 μm is a normal as-machined finish. Ra 0.8–1.6 μm usually needs a finishing pass with a sharp insert and a stable setup. Ra 0.2–0.8 μm often means a separate operation, sometimes on a different machine. Asking which of these three bands your part needs is a fast way to separate a real answer from a generic one.

The last layer is material. Aluminum 6061-T6 cuts clean and stays stable. Titanium TC4 (Ti-6Al-4V) work-hardens if you dwell, so the toolpath needs constant feed and plenty of coolant. Inconel pushes tool wear up fast and shortens the window between a good part and a scrapped one. A shop that has run these materials will tell you which features worry them before you ask.

• 1
  Ask for a first-article reportIt shows the measurement method, not just the result.
• 2
  Watch depth-to-diameterOver 6:1, expect more setups or a different tool.
• 3
  Match finish to functionSealing faces need more than a cosmetic Ra.

A marketing page can claim anything. A machine list is harder to fake and much more useful. GreatLight runs 127 high-precision CNC machines, including 16 simultaneous 5-axis machining centers, 12 four-axis mills, 27 three-axis machines, and 16 mill-turn centers. That mix is what lets a shop route your part to the right spindle instead of forcing it onto whatever is free.

Three-axis machines cover flat plates, pockets, and simple profiles. They are fast and cheap per part. The limit appears when a feature sits on a face that cannot be reached without re-fixturing. Every extra setup adds a datum shift. On a part with several related bores, those shifts stack up and eat the tolerance budget.

Four-axis and mill-turn centers handle cylindrical work with cross features. A shaft with a keyway and two radial holes can often run in one mill-turn cycle, which removes the concentricity error that comes from moving the part between a lathe and a mill.

Five-axis machining solves reach and setup count at the same time. A part with features on five faces can run in one or two setups instead of five. The trade-off is programming time and a higher hourly rate. It pays off on complex parts, not on simple brackets. If a supplier quotes five-axis for a flat plate, ask why.

• 1
  Size envelopeUp to 4,000 mm maximum processing size.
• 2
  Rotary workØ400 mm rotary table for cylindrical features.
• 3
  Right machine, right jobFive-axis on a flat plate is wasted money.

Machining creates the shape. Inspection proves the shape. A shop that inspects only the first and last part of a batch is gambling on the middle. GreatLight inspects 100% of parts before shipment, with a raw material check, in-process monitoring, and a final inspection, and reports are available on request.

The measurement method matters as much as the pass rate. A caliper reads to 0.02 mm on a good day. A micrometer is better on outside diameters. A CMM or a vision system is what you need for true position, profile, and angular features. If a supplier cannot say which instrument will check a given callout, the tolerance is not really controlled.

Thermal effects are the quiet failure mode. Aluminum expands about 23 μm per meter per °C. A 300 mm part measured at 26 °C instead of 20 °C reads roughly 0.04 mm long. That is eight times a ±0.005 mm tolerance. Real shops either control the room or state the measurement temperature in the report.

Traceability closes the loop. Material certificates, heat lot numbers, and inspection records let you trace a problem back to a batch instead of scrapping a whole shipment. For medical and automotive work, this is not optional. ISO 9001:2015, IATF 16949:2016, ISO 13485:2016, and ISO 27001:2022 cover quality, automotive, medical, and information security respectively.

• 1
  Ask for the instrumentCMM for position, micrometer for diameters.
• 2
  State the temperature20 °C is the reference for most drawings.
• 3
  Keep the recordsMaterial certs make a recall traceable.

Anyone can promise a date. A believable lead time comes from a shop that can name the bottleneck. If the answer is "we will start tomorrow", the next question is which machine and which operator. If the answer is "the five-axis queue is three days", that is a real plan.

GreatLight quotes and returns a free DFM analysis within 12 hours, can start production within 24 hours, and ships parts in 3–5 days for standard work. Historical late-delivery probability is below 2%. Those numbers reflect how the shop schedules, not a guarantee for every part. A new geometry with tight tolerances may need a first article before the full run.

Watch the ratio between quote speed and technical depth. A quote that arrives in ten minutes with no questions usually means the supplier did not open the model. A quote that arrives in a few hours with two DFM notes means someone read the drawing and thought about fixturing.

Volume changes the plan too. There is no minimum order quantity here, so a single prototype and a 10,000-part run both go through the same shop. But the economics shift: prototypes are priced for setup, production runs are priced for cycle time. A supplier who quotes both the same way is not pricing either one properly.

• 1
  Name the bottleneckA real answer mentions a machine or a queue.
• 2
  DFM before priceTechnical notes show the model was opened.
• 3
  Prototype vs. runSetup cost dominates one; cycle time dominates the other.

Brokers are not dishonest by default. They resell capacity and sometimes add value. The problem is that they cannot answer process questions, and process questions are exactly what you need answered before a purchase order.

A broker cannot tell you which machine will run the part. They cannot explain why a feature needs a second setup or why a tolerance is tight. They forward your drawing and forward the answer. Every round trip adds a day and loses detail.

Other signals are subtler. Photos that appear on several websites. A machine list with no model numbers. A tolerance claim with no measurement method. A price that is far below the rest of the market with no explanation of how the cycle time gets there.

The fix is simple. Ask two technical questions about your own part and see who answers them. Ask what the first operation is and how the datum is set. Ask how a specific callout will be measured. A real shop answers in its own words. A broker forwards the email.

• 1
  No model numbersMachine lists without models are hard to verify.
• 2
  Forwarded answersReplies that repeat your question back at you.
• 3
  Price with no routeLow numbers need a cycle time behind them.