Milling and turning are not two prices for the same job. They are two cost structures, and which is cheaper depends almost entirely on your part’s shape and quantity. This page sets out where the money goes in each, gives a worked calculation for the point at which one route overtakes the other, and is honest about when the cheaper-looking process is the wrong one to buy.
Key takeaways
- Cost follows geometry. Rotational parts belong on a lathe, prismatic parts on a mill; forcing either the other way adds setups, and setups are the expensive part.
- Setup is a fixed charge spread over the batch. At single-digit quantities it can dominate the unit cost; by a few hundred pieces it is close to irrelevant and cycle time takes over.
- The worked example below puts the crossover between a two-setup lathe-plus-mill route and a single-setup turn-mill route at 14 pieces. Substitute your supplier’s figures and the arithmetic still works.
- General tolerances default to ISO 2768-1:1989 unless the drawing says otherwise. MW+ works to ±0.01 mm as a general figure, ±0.005 mm as a precision band and ±0.001 mm on selected features, with Cpk ≥1.67 on controlled characteristics.
- The cheapest process is the wrong call when it costs you a datum. Splitting a part across two machines to save machine-hours can put your critical dimensions on two different setups.
- Why milling and turning carry different cost structures
- What actually drives the cost of a machined part?
- How does part type change the milling-versus-turning answer?
- A worked example: where the two-setup route stops paying
- Does a turn-mill centre change the economics?
- What tolerance and finish does each process hold?
- How does batch size move the decision?
- When the cheaper process is the wrong call
- What to send so the quote reflects the right process
- Frequently asked questions
Why milling and turning carry different cost structures
In turning the workpiece rotates and the tool is fed against it; in milling the tool rotates and traverses. That single difference propagates into everything that costs money.
A lathe cut is continuous; a milling cut is interrupted, each flute entering and leaving once per revolution. That cycling is why milling cutters generally wear faster than turning inserts in the same material, and why milling needs lighter finishing passes.
Workholding differs just as sharply. A round bar drops into a chuck or collet with no part-specific hardware; a prismatic plate or casting usually needs soft jaws or a dedicated fixture, which is engineering time spent before a chip is cut and charged against a single part on a one-off. That is the first reason two competent shops can quote the same geometry very differently.
Programming follows the same pattern. A turned profile is a two-dimensional contour; a milled pocket with compound draft is a toolpath simulated against holder, fixture and envelope, and that effort scales with complexity far more steeply. Our CNC milling services and CNC turning services pages list the envelopes that set those limits.
What actually drives the cost of a machined part?
Every quote decomposes into the same buckets; what changes between the processes is the weight each one carries. The table below carries no currency figures deliberately, because a rate quoted out of context tells you nothing useful.
| Cost bucket | Behaviour in turning | Behaviour in milling | Spread over |
|---|---|---|---|
| Programming and proveout | Short for plain contours; grows with live tooling | Grows steeply with surface complexity and axis count | The whole batch |
| Workholding | Standard chuck, collet or bar feed | Soft jaws or a dedicated fixture | The whole batch or programme |
| Setup and first-off | Fewer tools; datum is the rotational axis | More tools and offsets; datum set on the part | Each setup |
| Cycle time | Continuous cut, high removal on round stock | Interrupted cut, lighter finishing passes | Every part |
| Tooling consumption | Indexable inserts, predictable edge life | Faster wear on interrupted cuts | Every part |
| Number of operations | One for pure rotational geometry | One per accessible face without rotary axes | Every part, plus handling |
Notice how many of these are fixed per batch rather than per part. That is the whole game: at low volume you are buying setup, at high volume cycle time, and the winner flips somewhere in between.
How does part type change the milling-versus-turning answer?
It is more useful to go part type by part type than to compare processes in the abstract. If you are still deciding which process your geometry belongs on at all, our companion article on selecting the right process for your design answers that. This section asks what it costs.
Rotational parts
Shafts, pins, bushings, spacers, fittings, rings and adaptors are turning work. The real cost question is bar-fed versus chucked, and whether live tooling can absorb the secondary features in the same setup. Long slender small-diameter versions belong on Swiss-type equipment, where a guide bush supports the work close to the cut.
Prismatic parts
Brackets, housings, manifolds, plates, mould inserts and enclosures are milling work, and the cost driver is the number of faces you must reach. Each unreachable face is a new setup or a reason to move to multi-axis machining: a five-face bracket is five setups on a 3-axis machine and can be one on a 3+2 indexed machine.
Hybrid parts
Flanged shafts, eccentric bushings, valve bodies, rotors with keyways or cross-drilled ports carry both rotational and prismatic features, so the decision is genuinely open. They are where a wrong process choice costs the most.
| Part type | Dominant process | What makes it cheap | What makes it expensive |
|---|---|---|---|
| Plain shaft, pin, spacer | Turning, bar-fed where diameter allows | No part-specific fixture; short cycle | Any feature needing a second machine |
| Flat plate with holes | Milling, single face | One orientation, standard vice | Flatness as stock relieves; may need a re-cut |
| Multi-face bracket or housing | Milling, 3+2 indexed if available | Setup elimination; datums established once | Setup count on a 3-axis machine; fixture design |
| Mould insert, freeform cavity | Milling, simultaneous where surfaces blend | Nothing; inherently programming-heavy | Finish specification and polish allowance |
| Flanged shaft, valve body | Open: two-setup route or turn-mill | Single-setup turn-mill above the crossover quantity | Two setups below it, plus inter-operation handling |
| Small slender rotational part | Swiss-type turning | Guide bush allows high length-to-diameter ratios | Features outside the guide bush zone |
A worked example: where the two-setup route stops paying
Take a flanged shaft: a Ø50 mm body 120 mm long, with a turned flange, two spanner flats and two cross-drilled ports. It can be made two ways.
- Route A — turn the body and flange on a lathe, then move to a milling centre for the flats and cross-holes: two setups, two machines, one transfer.
- Route B — one setup on a turn-mill centre with live tooling; everything cut before the part leaves the chuck.
The inputs below are illustrative process-sheet figures of the kind your supplier can give you for your own part, not published benchmarks. Substitute the numbers your shop quotes and the method is unchanged.
| Input | Route A (lathe + mill) | Route B (turn-mill) |
|---|---|---|
| Setup and proveout | 35 min lathe + 60 min mill | 120 min, combined |
| Total fixed minutes per batch | 95 min | 120 min |
| Cycle | 2.5 min turning + 3.5 min milling | 5.2 min, combined |
| Inter-operation handling | 1.0 min | None |
| Total variable minutes per part | 7.0 min | 5.2 min |
Machine minutes per part for a batch of Q pieces is the fixed time spread over the batch plus the variable time every part incurs:
M(Q) = Fixed ÷ Q + Variable
Route A gives M = 95 ÷ Q + 7.0. Route B gives M = 120 ÷ Q + 5.2. Set them equal to find the crossover:
- 95 ÷ Q + 7.0 = 120 ÷ Q + 5.2
- 7.0 − 5.2 = (120 − 95) ÷ Q
- 1.8 = 25 ÷ Q
- Q = 25 ÷ 1.8 = 13.9, so 14 pieces
Below 14 pieces the two-setup route uses less machine time, because the turn-mill setup has not been paid off; above it the single setup wins and keeps winning. At Q = 10 it is 16.5 min per part against 17.2; at Q = 50, 8.9 against 7.6; at Q = 500, 7.19 against 5.44 — about a quarter of the machine time saved.
Two things follow. The crossover is low, often inside prototype quantities, which is why single-setup routes get proposed more often than buyers expect. And machine minutes are not money: a turn-mill hour is dearer than a plain lathe hour, which pushes the crossover up. Ask your supplier to rerun the arithmetic with their own rates.
To sanity-check the cycle figure itself, spindle speed n in rev/min is n = (1000 × Vc) ÷ (π × D), with cutting speed Vc in m/min and diameter D in mm. For the Ø50 body at Vc = 200 m/min, n = 200 000 ÷ 157.08 = 1273 rev/min; at 0.15 mm/rev the tool advances 191 mm/min, so a 120 mm pass takes 0.63 min, about 38 seconds. Multiply by the pass count for a floor on the turning cycle. Anything well above it is tool changes, air moves and in-process gauging.
Does a turn-mill centre change the economics?
In one specific way: it converts a setup into a tool change. Flats, cross-holes, keyways and light contouring all happen while the part is still gripped on its original datum.
That matters more for tolerance than for cost. Cut in one setup, the relationship between flats and diameter is a machine-accuracy question, characterised by positioning repeatability under ISO 230-2:2014. Split across two machines it becomes a re-datuming question, and the error budget must carry the fixture and the operator too.
| Consideration | Two-setup route | Single-setup turn-mill |
|---|---|---|
| Fixed cost per batch | Lower — two simple setups | Higher — one complex setup and longer proveout |
| Variable cost per part | Higher — two cycles plus handling | Lower — one cycle |
| Datum integrity | Features on different setups reference different datums | All features share one datum |
| Handling damage risk | Present at every transfer | Eliminated |
| Machine hour cost | Lower per machine | Higher |
| Best suited to | Very small batches, simple secondary features | Repeat work, tight feature-to-feature relationships |
What tolerance and finish does each process hold?
Neither process has an inherent accuracy advantage across the board. What differs is which characteristics come easily.
Turning produces diameters and lengths on one rotational datum, so concentricity between turned features is close to free. Milling produces position and profile characteristics, defined in ISO 1101:2017 and in the North American convention ASME Y14.5-2018, whose accuracy depends on how well the drawing’s datum scheme survives the setup plan. Naming a tolerance without naming its datums is the commonest reason a drawing cannot be quoted as drawn.
Where no individual tolerance is indicated, ISO 2768-1:1989 supplies default linear and angular limits. Its companion for geometrical defaults, ISO 2768-2:1989, is withdrawn and replaced by ISO 22081:2021; drawings in circulation still call out the old part, so state which convention governs. For fits between mating turned and bored features, use the ISO code system in ISO 286-1:2010 rather than inventing a bilateral band.
| Characteristic | Usually easier in turning | Usually easier in milling | Governing reference |
|---|---|---|---|
| Diameter and length | Yes, single datum and continuous cut | Achievable, but by interpolation | ISO 286-1:2010 |
| Concentricity and run-out | Yes, inherent to the process | Depends on the setup plan | ISO 1101:2017 |
| True position of a hole pattern | Only radial patterns, with live tooling | Yes, this is what milling is for | ISO 1101:2017 / ASME Y14.5-2018 |
| Profile of a freeform surface | No | Yes, the reason simultaneous multi-axis exists | ISO 1101:2017 |
| As-machined surface texture | Continuous cut, consistent lay | Finish depends on step-over | ISO 21920-1:2021 / ISO 21920-2:2021 |
Surface texture parameters are now defined in ISO 21920-2:2021, superseding ISO 4287:1997, and the indication rules in ISO 21920-1:2021, superseding ISO 1302. Both conventions are still in circulation, so the drawing must say which governs. MW+ works to ±0.01 mm generally to ISO 2768-m, ±0.005 mm as a precision band and ±0.001 mm on selected features, with Ra 3.2 µm as-machined, Ra 0.4 µm fine-machined and Ra 0.1 µm polished — the same figures quoted for CNC precision parts.
How does batch size move the decision?
The crossover calculation generalises: of any two routes, the lower fixed cost wins at low quantity and the lower variable cost at high quantity, and there is always a quantity at which they swap.
| Batch band | What dominates the unit cost | Which route usually wins | What to optimise |
|---|---|---|---|
| 1–10 pieces | Setup, programming, fixture | Fewest fixtures, even if the cycle is slow | Standard workholding; relax non-critical tolerances |
| 10–100 pieces | Setup still significant; cycle starting to matter | Usually the consolidated route | Settle the datum scheme |
| 100–1,000 pieces | Cycle time and tool life | The route with the shortest proven cycle | Cutting data, tool grade, in-process gauging |
| 1,000+ pieces | Cycle time, tool consumption, automation | Bar-fed turning; pallet-loaded milling | Unattended running, tool life, sampling plan |
At the top of that range the sampling plan becomes part of the cost, as our piece on scaling CNC production from one part to a hundred thousand sets out.
When the cheaper process is the wrong call
Everything above points towards consolidating operations and putting geometry on its natural machine. There are real cases where that advice is wrong, and a supplier who never says so is not being straight with you.
When the design is not frozen
A consolidated single-setup programme is expensive to change. On the third iteration of a housing with a fourth expected, the flexible two-setup route is cheaper in total even though it loses on paper above the crossover. Buy consolidation once the geometry stops moving. The cheapest route is also wrong when it runs through the machine class that is fully booked, so ask about queue position as well as price.
When splitting the part splits the datum
Moving a secondary operation to a cheaper machine saves machine-hours and can still be wrong, because it puts two related characteristics on two different setups. If the drawing calls a tight positional relationship between a bore and a face and those are cut separately, you have bought a tolerance stack you did not budget for. Work out where your critical relationships sit before accepting a split route.
When the part is small enough that neither answer is the right one
Very small features, thin webs and micro-holes are not a milling-versus-turning question at all, but a question of whether the geometry belongs on micro-machining equipment, a Swiss-type lathe, or wire EDM — the last worth considering wherever a rotating tool cannot reach the feature at all.
What to send so the quote reflects the right process
Most process mis-selection traces back to an incomplete enquiry. A supplier who cannot see your quantity, your critical dimensions or your stock form quotes the safe route, which is rarely the cheapest.
| What to send | Why it changes the process choice | Cost of omitting it |
|---|---|---|
| 3D model in STEP or IGES | Shows reachability and sets the axis count | Quote assumes the most conservative setup plan |
| 2D drawing with datums and GD&T | Determines which features must share a setup | Route split across machines, tolerance stack appears late |
| Quantity now and expected annual volume | Sets which side of the crossover you are on | Priced as a one-off; no case made for consolidation |
| Material, temper and acceptable substitutes | Drives stock form, tool grade and cycle | Lead time lost sourcing a grade that was never critical |
| Which dimensions are genuinely critical | Lets the rest run to a general tolerance | Every dimension inspected as if critical |
| Surface texture and where it applies | Decides finishing passes and post-processing | Blanket Ra, paid for on faces nobody sees |
Every MW+ order ships with a certificate of conformance, a CMM inspection report and material certificates; AS9102 first article inspection and PPAP Level 3 are quoted per programme on request. The methods behind those reports, including CMM acceptance and reverification under ISO 10360-2:2009, sit on our CNC machining quality control page. Quotes come back within 24 hours; to talk a route through first, request a CNC machining quote.
Frequently asked questions
Is turning always cheaper than milling?
No. Turning is usually cheaper for rotational geometry because workholding is standard and the cut is continuous, but that advantage vanishes the moment the part needs features a lathe cannot reach. A prismatic housing turned from bar would waste most of the material and still need milling. Compare routes for your geometry and quantity, not processes in the abstract.
Why did two shops quote my part at completely different prices?
Almost always because they assumed different routes: one setup on a turn-mill centre against two across a lathe and a mill, or soft jaws against a dedicated fixture. Ask each supplier how many setups they assumed and what workholding. That single question explains most of the spread and tells you which quotes are actually comparable.
Should I specify the process on my drawing?
Generally no. Specify the result — dimensions, datums, tolerances, surface texture, material — and let the supplier choose the route. Dictating the process removes their ability to propose a cheaper one and gives you no protection the tolerances did not already give. The exception is where a customer specification or an approval genuinely requires a named process.
Does a tighter tolerance cost more on a lathe than on a mill?
It depends which characteristic you tighten. Tightening a diameter on a lathe is comparatively cheap because the process already works on the rotational datum. Tightening a positional relationship between features on different faces is expensive on either machine, because the cost sits in the setup plan and the inspection. Our comparison of ±0.01 mm against ±0.005 mm works through what each band changes.
Can one supplier do both, or should I split the work?
A supplier running both under one roof can propose the consolidated route and carry the tolerance stack themselves. Split across two, one owns the datum and the other inherits it, and any argument about a positional characteristic becomes a dispute rather than a correction. MW+ runs 60+ machining centres across milling, turning and multi-axis work on one 15,000 m² site.
Does the cheapest machine hour give the cheapest part?
Not reliably. A dearer machine that finishes the part in one setup can beat a cheaper one needing three, and it removes the handling and re-datuming risk. Compare finished-part cost including inspection and rework allowance, not the hourly figure — our article on what actually drives machine shop rates unpacks the structure.
