Processing tolerance key factors in CNC production
A processing tolerance is the band a cut dimension is allowed to land in, and what you hold depends on more than the machine spec sheet. This page is written for design engineers, process engineers and sourcing teams who need to judge where a tolerance band comes from and which parts can hold it. Read it and you can tell which factors are driving your stack before you release a drawing.

In this article
- 1
- 2
- 3
- 4
- 5
- 6
Key takeaways
What processing tolerance key factors actually control
A processing tolerance is the allowed variation between the nominal dimension on the drawing and the dimension you measure on the finished part. It is written as a total band, so ±0.05 mm means a 0.10 mm window. The tolerance key factors in CNC production are the individual error sources that consume that window: machine geometry, fixture stiffness, tool deflection, thermal expansion, material behaviour and measurement uncertainty. None of them act alone.
Engineers often treat tolerance as a property of the machine. On the shop floor it behaves more like a budget. A lathe rated to ±0.005 mm can hold ±0.02 mm on a long, slender shaft because the part bends under cutting force. The same lathe holds ±0.008 mm on a short, thick flange. The machine did not change. The stiffness of the feature did.
This is why tolerance review starts before CAM. Look at the feature length-to-diameter ratio, the wall thickness, the material and the datum scheme. Those four things narrow the realistic band faster than any machine list. If a drawing asks for ±0.01 mm on a 300 mm unsupported aluminium rib, the problem is the drawing, not the machine.
- 1Tolerance bandTotal allowed variation, not a plus/minus value alone.
- 2Error budgetSum of machine, fixture, tool, thermal and material errors.
- 3CapabilityWhat the process holds repeatedly, not once.
Machine geometry and workholding as processing tolerance key factors
Machine geometry sets the floor. Axis straightness, squareness between X and Y, spindle runout and ballscrew backlash all feed directly into the cut. A spindle with 5 µm runout cannot produce a Ø20 mm bore to ±0.005 mm, no matter how slow you feed. Thermal growth in the ballscrews adds a slower drift on top of that, typically a few microns over a long run.
Workholding is the factor engineers underestimate most. A three-jaw chuck closes on a round part and pushes it into a triangle. A vise lifts thin parts in the middle. A vacuum plate holds flat parts but cannot resist side load. Every one of those effects shows up as a dimensional error at the feature, often larger than the machine error itself.
The fix is usually boring soft jaws to the part diameter, supporting thin sections with a matched fixture, or moving to a 5-axis setup so more features are cut in one clamping. On a 5-axis machine the part stays in one frame, so positional error between features drops. GreatLight runs 16 simultaneous 5-axis centers with a Ø400 mm rotary table for exactly this reason.
Setup also matters at the datum level. If the drawing calls for a datum that is hard to reach after the first operation, the operator will pick a different face. That single choice can shift a dimension by 20–50 µm before a tool touches metal.
- 1Spindle runoutAbove 5 µm, tight bores become a gamble.
- 2Soft jawsBored to part diameter, they remove chuck distortion.
- 3Single setupFewer clampings mean less positional error.
Tool deflection, wear and material response
A cutting tool is a cantilever. Push it sideways and it bends. A Ø6 mm carbide end mill hanging 40 mm out of the holder deflects far more than the same tool held 20 mm out. That deflection shows up as a tapered wall, an oversize slot or a dimension that drifts as the tool wears. Reducing stick-out and taking lighter radial cuts is often worth more than any machine upgrade.
Tool wear is a slow change in effective diameter. On a Ø10 mm end mill, 10 µm of flank wear moves the cut wall by roughly the same amount. In a 500-part run that drift can walk the part out of tolerance near the end. Tool life management, not just tool choice, is one of the processing tolerance key factors on longer runs.
Material behaviour is the third piece. Aluminium 6061 cuts clean and holds tight bands on stiff features. Titanium Ti-6Al-4V spring-back, work-hardens and pushes the tool away, so the same feature needs a wider band or a finishing pass. Plastics like POM and PEEK move after machining as internal stress relaxes, so a dimension checked an hour after cutting may not match the one checked the next morning.
The practical rule is to match the band to the material. If the part is 6061-T6 with thick walls, ±0.01 mm is reasonable. If it is a thin titanium bracket, plan for ±0.05 mm and inspect after stress relief.
- 1Stick-outKeep it under 4× diameter where the geometry allows.
- 2Wear driftMeasure the first and last part, not just the first.
- 3Stress reliefPlastics and titanium may move after the cut.
Thermal drift and inspection as processing tolerance key factors
Steel expands about 11 µm per metre per °C. A 300 mm steel part in a shop that swings from 18 °C in the morning to 26 °C in the afternoon changes length by roughly 26 µm. That is larger than many tolerance bands on the drawing. Coolant, spindle heat and even direct sunlight on the machine can shift the cut over a shift.
The countermeasure is simple: measure at a controlled temperature, or measure and correct. A part checked on a warm CMM reads larger than the same part checked cold. GreatLight inspects 100% of parts before shipment with raw material checks, in-process monitoring and final inspection, and reports are available on request. That closes the loop between what the machine did and what the drawing asked for.
Measurement uncertainty is the last factor and the easiest to forget. A caliper reads to 0.01 mm but carries maybe 0.02 mm of uncertainty in the hands of a careful operator. A micrometer is better on round features. A CMM is better on position and form. If the tolerance is ±0.01 mm, a caliper cannot prove the part is good. Use a tool with at least 4× better resolution than the band.
Capability data pulls it together. Run 30 parts, measure the critical dimension, and calculate the spread. If the process capability index is below 1.33, the band is not stable yet. The next step is to find which factor moved, not to tighten the drawing.
- 1Temperature controlMeasure at 20 °C where the band is tight.
- 2Gauge ruleResolution 4× better than the tolerance.
- 3Capability indexBelow 1.33 means the process is still moving.
Which tolerance band fits which feature
Use this as a starting point, then confirm with a capability run.
| Feature and material | Realistic band | Dominant factor |
|---|---|---|
| Short steel bore, Ø20 mm, L/D 1.5 | ±0.005–0.01 mm | Machine geometry |
| Long steel shaft, Ø20 mm, L/D 10 | ±0.02–0.05 mm | Tool and part deflection |
| Aluminium 6061 pocket, 3 mm wall | ±0.02–0.05 mm | Workholding and wall spring |
| Titanium Ti-6Al-4V bracket | ±0.05 mm | Material spring-back |
| POM or PEEK bushing | ±0.05 mm after stabilisation | Post-machining stress relief |
| 300 mm steel part, uncontrolled shop | Add 0.02–0.03 mm | Thermal drift |
| Position of holes, 5-axis single setup | ±0.01–0.015 mm | Machine and fixture |
When to hold the band and when to loosen it
If the feature is short, stiff and cut in one setup, hold ±0.01 mm or tighter and inspect with a CMM. If it is long, thin or made of titanium or plastic, loosen the band to ±0.05 mm and spend the money on a better fixture instead of a tighter machine.
Processing tolerance key factors: common questions
What is a realistic tolerance for CNC machining?
On a short, stiff feature in aluminium or steel, ±0.005 mm is achievable and GreatLight works to that band on suitable parts. On long or thin features the realistic band widens to ±0.02–0.05 mm because deflection and vibration grow with the length-to-diameter ratio.
The drawing should match the feature, not the machine catalogue. A capability run on 30 parts tells you what the process actually holds.
Does a tighter tolerance always cost more?
Not always, but usually. A tighter band means slower feeds, more finishing passes, extra inspection and sometimes a new fixture. If the feature is already stiff and cut in one setup, the extra cost is small.
If the feature is flexible, the cost jump comes from the fixture and the inspection time, not the cutting time.
How does temperature affect a measured dimension?
Steel grows about 11 µm per metre per °C. A 300 mm steel part can change by roughly 3 µm for every 1 °C of shop swing. Aluminium moves about twice as much.
Measure at a controlled 20 °C when the band is tight, or record the temperature and correct the reading.
Why does my part measure differently the next day?
Internal stress from the cut relaxes over hours, especially in plastics, thin aluminium and titanium. The part creeps after it leaves the machine.
If the dimension is critical, add a stress relief step or a stabilising pause before final inspection.
Can 5-axis machining improve tolerance?
Yes, on parts with features on several faces. Cutting them in one setup removes the positional error that comes from re-clamping. GreatLight runs 16 simultaneous 5-axis centers for that reason.
It does not fix tool deflection or thermal drift on a single feature. Those need a different answer.
What inspection should I ask for?
Ask for first article inspection, in-process checks on the critical dimension and a final report. GreatLight inspects 100% of parts before shipment and provides reports on request.
Match the gauge to the band. A caliper cannot prove a ±0.01 mm dimension.
Send your drawing and get a tolerance review
We quote and return a free DFM analysis within 12 hours, flag any band that will not hold, and start production within 24 hours once you approve.
12-hour quoteFree DFM analysis100% inspectionNo minimum order quantity