Accuracy guide

How Accurate Are CNC Machines?

Short answer: a well-maintained 3-axis mill holds ±0.01 mm on a good day, while a temperature-controlled 5-axis cell can hold ±0.005 mm. This page shows where those numbers come from, which factors move them, and how to check a machine before you release a production run.

Get Instant Quote Talk to an Engineer

±0.005 mm capability16 five-axis centers100% inspectionISO 9001 / IATF 16949

On this page

1
Key takeaways
2
What accuracy actually means on a CNC machine
3
How accurate are CNC machines by axis count
4
What pulls real accuracy away from the spec sheet
5
Accuracy benchmarks by part and industry
6
Standards and reports you can ask for
7
How to verify CNC machine accuracy before a production run
8
Accuracy by machine type and part requirement

Quick answer

Key takeaways

Accuracy is not one numberPositioning accuracy, repeatability, and thermal drift are three separate specs. A machine can be accurate but not repeatable.

Axis count changes the resultEach extra rotary axis adds a stack of errors. A 5-axis part is harder to hold than the same part cut in three setups.

The part decides the toleranceThin walls, deep pockets, and hard alloys push real accuracy away from the machine's spec sheet.

Verify before you commitBallbar, laser interferometer, and a first-article CMM report tell you what the machine actually does today.

±0.005 mm is a shop capabilityAt GreatLight that figure covers the full process: machine, fixture, tooling, temperature, and inspection.

Definitions

What accuracy actually means on a CNC machine

How accurate are CNC machines depends on which number you are asking about. Positioning accuracy is how close the tool tip lands to the commanded coordinate. Repeatability is how close it lands on the same spot the second, hundredth, and thousandth time. They are tested separately and they fail separately.

A machine with ±0.01 mm positioning error but ±0.002 mm repeatability will still produce a consistent batch. A machine with ±0.005 mm positioning error but ±0.015 mm repeatability will not, because no two parts come out the same. For production, repeatability usually matters more than the headline accuracy figure.

A third number is thermal stability. Iron grows about 11 µm per meter per °C. A 3 °C swing in the shop moves a 1,000 mm axis by roughly 33 µm. That is larger than the tolerance band on many aerospace and medical parts, which is why temperature control is part of the accuracy spec, not a comfort feature.

1
Positioning accuracyDistance between commanded and actual position under test conditions.
2
RepeatabilitySpread of results when the same move is made many times.
3
Thermal driftError that grows as the machine, spindle, and coolant warm up.
4
Process capabilityWhat the whole shop can hold, including fixture and tool wear.

Machine types

How accurate are CNC machines by axis count

A 3-axis vertical mill is the simplest kinematic chain. Three linear axes, one spindle, no rotary stack. Entry-level machines land around ±0.01 mm to ±0.02 mm. A maintained machine on a temperature-controlled floor can reach ±0.005 mm on light finishing passes with sharp tooling.

A 4-axis mill adds one rotary table. Each rotary move introduces runout, backlash, and angular positioning error, which multiply by the distance from the table center. A feature 150 mm from center with 0.01° angular error shifts by about 26 µm. Keep work close to the rotary center and the penalty shrinks fast.

A 5-axis machine adds a second rotary axis and a much longer error stack. Trunnion tilt, pivot distance, and tool-tip offset all contribute. Modern simultaneous 5-axis centers with linear scales and thermal compensation hold ±0.005 mm to ±0.01 mm, but only when the post-processor and tool offsets are correct. A wrong pivot length can put you 0.1 mm off before the first cut.

The same part cut on 3-axis in five setups can lose more accuracy than a 5-axis cut in one setup, because each refixturing adds a datum shift. Setup count is often the biggest error source on the floor.

1
3-axisSimplest error stack. Best accuracy per dollar for prismatic parts.
2
4-axisAdds angular error scaled by distance from table center.
3
5-axisBest for contoured parts, but requires correct pivot and tool offsets.
4
Mill-turnFewer setups, but turret and spindle alignment must be checked often.

Error sources

What pulls real accuracy away from the spec sheet

Geometry errors come first. Squareness between axes, straightness of each guideway, and spindle axis alignment set the floor. These errors are measured with a ballbar or laser interferometer and corrected in the control. They drift after a crash or a hard season of heavy roughing.

Tooling is the second source. A 12 mm end mill deflects under cutting force. In aluminum with a 40 mm overhang, radial deflection can reach 20–40 µm. Reduce overhang, use a stub cutter, or take a lighter finishing pass. Tool runout in the holder adds another 5–15 µm if the holder is worn or the collet is dirty.

Fixturing and workholding matter more than most people expect. A part clamped on three points will move when the clamps release. Thin walls spring back after the cut. Vacuum fixtures and soft jaws reduce this, but the real fix is to plan the clamping sequence so the final finishing pass runs on a relaxed part.

Material behavior closes the list. Titanium and Inconel work-harden and push the tool away. Plastics and thin aluminum walls move with temperature and clamp pressure. The machine may be perfect and the part still comes out 0.03 mm out of tolerance.

1
GeometrySquareness, straightness, and spindle alignment. Check after any crash.
2
Tool deflectionGrows with overhang and cutting force. Keep overhang under 4× diameter.
3
WorkholdingClamp-induced distortion is a common cause of out-of-round bores.
4
Thermal growthWarm up the spindle for 20–30 minutes before finishing tight features.

Benchmarks

Accuracy benchmarks by part and industry

General industrial parts and consumer enclosures usually need ±0.05 mm to ±0.1 mm. Any maintained 3-axis machine hits that without special measures. Focus on surface finish and burr control instead.

Automotive and EV parts sit around ±0.02 mm to ±0.05 mm on mating features, with looser tolerances elsewhere. Fixture repeatability and in-process gauging matter more than the machine's headline number, because volumes are high and the first part sets the trend.

Aerospace structural and engine parts tighten to ±0.01 mm, often with profile tolerances on contoured surfaces. That pushes the shop toward 5-axis, temperature control, and CMM verification on a sampling plan.

Medical devices and implants go to ±0.005 mm or tighter on critical features, with surface finish often specified between Ra 0.2 μm and Ra 0.8 μm. Cleanliness, traceability, and inspection records carry as much weight as the tolerance itself. At GreatLight, that combination is covered by ISO 13485 processes and 100% inspection before shipment.

1
Industrial±0.05 to ±0.1 mm. Standard 3-axis work.
2
Automotive / EV±0.02 to ±0.05 mm on mating faces.
3
Aerospace±0.01 mm plus profile callouts on contoured surfaces.
4
Medical±0.005 mm and finer finishes on critical features.

Standards

Standards and reports you can ask for

ISO 230 defines how machine tool accuracy is tested. ISO 230-2 covers linear axis positioning accuracy and repeatability. ISO 230-3 covers thermal effects. ISO 230-4 covers circular tests, which is where ballbar results fit. Ask a supplier which parts of ISO 230 they actually run and how often.

ISO 10791 applies to machining centers and adds test pieces, including the cone frustum used to check 5-axis interpolation. A shop that runs a cone frustum test can show you real 5-axis accuracy instead of a spec sheet.

Inspection reports matter as much as machine specs. A first-article inspection report, a CMM program, and material certificates give you a paper trail. For medical and aerospace work, that trail is often a contract requirement.

At GreatLight, inspection covers raw material check, in-process monitoring, and final inspection before shipment, with reports available on request. The shop runs ISO 9001:2015, IATF 16949:2016, ISO 13485:2016, and ISO 27001:2022 systems across three plants covering 7,600 m².

1
ISO 230-2Linear positioning accuracy and repeatability.
2
ISO 230-3Thermal distortion of the machine structure.
3
ISO 230-4Circular tests for interpolation and squareness.
4
ISO 10791Machining center tests, including the cone frustum.

Verification

How to verify CNC machine accuracy before a production run

Run these checks in order. Skipping step 1 makes every later result suspect.

1
Warm up the machineRun the spindle at 8,000–12,000 rpm for 20–30 minutes and exercise all axes through their full travel. Cold machines drift 10–30 µm in the first hour. Measure only after the spindle housing temperature stabilizes.
2
Check the environmentLog shop temperature and humidity for 24 hours. Hold the finishing area within ±1 °C if you are chasing ±0.005 mm. A machine next to a loading door or a heat-treat oven will never hold tight tolerances.
3
Run a ballbar testA 100–300 mm ballbar circle reveals squareness, backlash, and servo mismatch. Look for a circularity value under 10 µm for a tight-tolerance machine. Repeat at two feed rates to separate geometry from servo error.
4
Run a laser interferometer checkMeasure positioning error and repeatability on each linear axis over its full travel. Compare the results against the last calibration file. A jump of more than 5 µm on a single axis usually means a loose scale, a worn guide, or a bumped machine.
5
Cut a test part with known geometryMachine a stepped block, a circular pocket, and a set of bores. Measure with a calibrated CMM or micrometer. Bores are the fastest tell: out-of-round beyond 5 µm points to spindle or workholding issues, not to the control.
6
Confirm the fixture and offsetsRe-zero the work offset, check tool length offsets with a presetter, and verify the pivot length on 5-axis machines. A wrong pivot length of 0.05 mm shows up as 0.05 mm error on every tilted feature.
7
Run a first-article inspectionMeasure all critical dimensions on the first part and compare to the drawing. Document the results. If any dimension sits within 20% of the tolerance limit, adjust the process before releasing the run.

Reference

Accuracy by machine type and part requirement

Typical achievable values on maintained machines with correct fixturing and tooling.

Machine type	Typical positioning accuracy	Best for	Watch out for
3-axis mill	±0.01 to ±0.02 mm	Prismatic parts, plates, fixtures	Multiple setups add datum shift
4-axis mill	±0.01 to ±0.015 mm	Shafts, hubs, features on a rotary	Angular error grows with distance
5-axis mill	±0.005 to ±0.01 mm	Contoured surfaces, one-setup parts	Wrong pivot length offsets everything
Mill-turn center	±0.005 to ±0.015 mm	Turned parts with milled features	Turret and spindle alignment drift
CNC lathe	±0.005 to ±0.01 mm	Round parts, threads, bores	Bar stock and chuck pressure effects
Entry-level 3-axis	±0.02 to ±0.05 mm	Prototypes, non-critical brackets	No thermal compensation, no scales

FAQs

Frequently asked questions

How accurate are CNC machines in mm, not in marketing terms?

A maintained 3-axis mill holds ±0.01 mm to ±0.02 mm on light finishing cuts. A temperature-controlled 5-axis cell with linear scales holds ±0.005 mm to ±0.01 mm. Entry-level machines without scales and thermal compensation sit at ±0.02 mm to ±0.05 mm.

Those figures assume correct tooling, rigid fixturing, and a warm machine. A cold machine or a long tool overhang can double the error before the part is even measured.

Does a higher axis count always mean better accuracy?

No. Each rotary axis adds error sources. A 4-axis or 5-axis machine has more places to go wrong than a 3-axis machine.

The advantage is setup count. A 5-axis machine cuts a contoured part in one setup instead of five, and removing four refixturings often improves the final part accuracy even if the machine itself is not more accurate.

How often should a CNC machine be recalibrated?

Most shops run a full laser and ballbar check every 6 to 12 months, plus a check after any crash or spindle change. High-tolerance cells in medical or aerospace work are checked more often, sometimes quarterly.

Between calibrations, watch the first-article results. A slow drift in the same direction is an early warning that the machine needs attention.

Can you hold ±0.005 mm on a 4,000 mm long part?

Full-length ±0.005 mm on a 4,000 mm part is not realistic. Thermal growth alone moves that length by tens of microns over a normal shop day.

The practical approach is to hold tight tolerances on local features and use a looser, controlled tolerance for the overall length. Discuss this with the shop before the drawing is frozen.

What causes a machine to lose accuracy over time?

Guideway wear, ball screw backlash, spindle bearing wear, and drifting linear scales are the usual suspects. Crashes bend or shift components in ways that show up immediately in ballbar results.

Temperature changes in the shop and a dirty coolant system also contribute. A machine that ran fine in winter may drift in summer if the shop is not climate controlled.

Do I need a CMM report for every part?

No. For most production runs, a first-article inspection plus in-process checks covers the risk. A full CMM report on every part is expensive and slow.

Ask for full CMM reports on critical features, on the first article, and on any part where the process has shown drift. That gives you the data without the cost of 100% CMM inspection.

Send us your tolerance and we will confirm what we can hold

Upload your drawing and we return a quotation with free DFM analysis within 12 hours, plus a clear statement of achievable tolerances for your features.

12-hour quote100% inspection±0.005 mm capabilityNDA on request

Get Instant Quote Talk to an Engineer