CNC Machining Design Guide: Seven Decisions That Decide Cost and Tolerance Before the First Cut

A machined part is priced before the spindle starts. Corner radii, tolerance callouts, surface finish, setup count, thread depth, wall thickness and expected volume together determine how many operations the part needs — and how likely the first article is to pass inspection. This CNC machining design guide works through those seven decisions in the order they affect cost, so engineers can specify what the function needs and buyers can read a quotation for what it actually contains.

Where a Machined Part's Cost Actually Comes From

A machining quotation is not a single number derived from a single variable. It is the sum of five separate cost drivers, and each one is set by a different part of the drawing.

Two observations follow. First, most cost is decided by geometry, not by tolerance — a part with an unreachable corner is expensive at any tolerance. Second, the only cost driver that adds no functional value when over-specified is inspection, which is why tolerance discipline has a disproportionate effect on price.

Decision 1 — Geometry a Rotating Tool Can Reach

A milling cutter is round, and the shape it leaves in a corner is round. This is the most common source of avoidable cost in a machined part, and it is where design for manufacturability for CNC parts earns its name: the cheapest correction is the one made on the drawing rather than at the machine.

Internal corners. A square internal corner cannot be produced by a rotating tool without leaving a radius equal to the cutter radius, which means internal corner radius for CNC milling should be specified at or above the radius of the cutter that will reach that depth. Specifying a smaller radius forces either a smaller cutter (slower, more fragile, more passes) or a second process such as EDM. Design relief, a corner radius slightly larger than the tool, or a dog-bone relief at the corner avoids the problem entirely.

Pocket depth. As depth increases relative to cutter diameter, the tool becomes a long, thin cantilever. Deflection produces taper and chatter, and the shop compensates with lighter passes and more time. A pocket that is deep relative to its width is not impossible — it is simply priced as a different job.

Undercuts and side features. Any feature that cannot be approached from the tool axis requires a specialised cutter, an additional setup or a different process. If the feature is cosmetic, consider moving it to an accessible face; if it is functional, expect it to be quoted as its own operation.

The practical test is simple: point at every feature on the drawing and ask which direction the tool enters from. Features with no clean answer are where cost accumulates.

Decision 2 — Put Tolerance Where the Function Lives

Tolerance is a statement about function, not about seriousness. A drawing covered in tight tolerances does not communicate a demanding part; it communicates an unfinished analysis.

Start from the datum scheme. Every controlled feature should be measured from a datum that reflects how the part is located in assembly. When datums are chosen for drawing convenience rather than function, GD&T callouts for machined parts become internally inconsistent, and inspection turns into a negotiation.

Separate functional from non-functional dimensions. A bore that receives a bearing, a spigot that centres two components, and a face that seals all justify tight control. An outside profile that is never touched in assembly usually does not. Applying general tolerances to the second group and explicit control to the first is what CNC machining tolerance standards are designed for: ISO 2768 exists precisely so a drawing does not need a tolerance on every dimension.

Watch the stack, not just the feature. A single ± sign tells the shop about one dimension. What matters is whether the accumulated variation across three parts still closes the assembly. When a stack is tight, the cheaper fix is often to loosen one non-critical link rather than tighten two critical ones.

Remember that tolerance is inspection. Each controlled feature needs a method, a fixture and a record. Tightening a tolerance that no one measures adds cost without adding control; loosening a tolerance that the assembly depends on removes control without removing cost.

Decision 3 — Surface Finish Only Where It Does Work

Surface roughness is specified as Ra, an average deviation from the mean line. It is a functional requirement on some surfaces and a cosmetic one on others, and the two carry very different costs.

Where finish matters. Sealing faces, sliding surfaces, bearing journals, optical or contact surfaces and any face that receives an adhesive or a gasket. For these, surface roughness Ra for sealing surfaces is a genuine engineering decision — a finish that is too rough will not seal, and a finish that is unnecessarily fine requires slower feeds, additional passes and often a separate finishing operation.

Where finish does not matter. Non-contact external faces, clearance bores, internal cavities and any surface hidden in assembly. Specifying a fine finish here adds machining time while producing a surface no one will ever see or measure.

The useful discipline is to mark finish only on the faces that need it, using a general note for everything else. A drawing with three finish callouts on three functional faces is easier to quote, easier to inspect and cheaper to make than one with a blanket requirement across the whole part.

Decision 4 — Count the Setups Before You Draw the Fixture

Every time a part is removed and re-fixtured, its position is re-established from a new reference. That is a source of variation that no tolerance callout can eliminate, and it is also a line item on the quotation. The relationship between setup count and machining cost is direct: each setup adds fixturing, handling, alignment and inspection, and none of that work changes the part.

Setups follow feature accessibility. If all critical features can be reached from one orientation, the part can often be produced in a single setup with a single datum. If critical features sit on four faces, the part needs four setups — or a multi-axis machine that reaches them without re-fixturing.

Multi-axis is not automatically better. Five-axis work reduces setup count and datum shift, which improves positional consistency across faces. It also concentrates capability into fewer, more expensive operations. The right answer depends on whether the part's function actually depends on relationships between faces. A part where one face carries all the critical features gains nothing from multi-axis work; a part with coaxial bores on opposite faces gains a great deal.

Design for the setup you want. Group critical features onto as few faces as possible. Keep non-critical features wherever they are convenient. This single change — grouping by function rather than by appearance — is often the largest cost reduction available without touching tolerance.

When the design genuinely requires many faces and relationships, that is the case where CNC turned and milled components in a single multi-axis operation make sense: turn the primary form, mill the features, and keep one datum throughout.

Decision 5 — Threads, Inserts and Mating Features Belong in the Same Drawing

A machined part rarely exists alone. It receives a screw, holds an insert or threads into another component, and the thread specification is part of the part, not an accessory to it.

Thread engagement. As a working rule of thumb, thread engagement of one to one and a half times the nominal diameter is used in steel, with more engagement in softer materials. Beyond that, additional depth adds little strength while adding tap wear, breakage risk and cycle time. Thread depth in machined holes should therefore be specified as a functional requirement, not as "as deep as possible".

Through versus blind. A through hole is cheaper to produce and easier to clean. A blind hole is necessary where the thread must not break through, but it requires depth control, chip clearance and a bottom clearance allowance. Specify the thread depth and the drill depth separately; conflating them is a frequent source of a tapped hole that is shallower than intended.

Thread callouts. State the standard explicitly — ISO metric coarse or fine, UNC, UNF — along with tolerance class and whether the thread is cut or rolled. Mixed callouts on one drawing invite a guess, and a guess about thread class is a costly one.

Inserts in machined and moulded parts. Where a heat insert nut is planned, the boss that receives it should be designed with the insert, not after it. Boss wall thickness, boss height, hole diameter and the allowance for melt flow all come from the insert specification. A boss designed first and fitted with an insert later usually ends up either too thin to hold torque or too large to fit the envelope.

Mating hardware. If the part mates with a specific screw, the drawing should name that screw and its standard. Two suppliers reading "M4 screw" without a standard can produce two different interfaces, and the machined part is the one that has to be reworked.

Decision 6 — Wall Thickness, Rigidity and Chatter

Thin walls are where machining stops being a cutting problem and becomes a stiffness problem.

Deflection, not just removal. A thin wall moves away from the cutter. The tool pushes, the wall yields, and the result is a wall that is thicker at the base, thinner at the top and out of tolerance in the middle. The remedy is support — leaving stock, machining in stages, adding temporary ribs — all of which is legitimate work and all of which is quoted.

Chatter. As wall stiffness falls, vibration rises. Chatter shows up as a poor finish and as dimensional variation, and the usual response is reduced depth of cut and more passes. Thin wall machining distortion is therefore a cost driver, not merely a quality risk.

Residual stress. Removing a large proportion of a billet releases internal stress, and the part can move after the final pass. Symmetrical removal, stress-relief between roughing and finishing, and choosing a stock form closer to the finished envelope all reduce the effect.

Practical design guidance. Keep walls as uniform as the function allows, avoid isolated thin sections that serve no purpose, and add ribs where stiffness is needed instead of thickening an entire face. Where a thin wall is functionally necessary, say so on the drawing — the shop will plan around it, and planning is cheaper than reworking.

Decision 7 — Know When CNC Stops Being the Right Process

CNC machining is the correct answer for prototypes, low and moderate volumes, complex geometry, tight tolerance and parts where the design is still moving. It stops being the correct answer when the geometry is stable and the volume is high enough that tooling cost amortises.

Knowing when to switch from CNC to casting or forging is a geometry question before it is a volume question, and getting the order right prevents tooling being bought for a part that was never suitable for it.

The decision factors are geometry, material and volume — in that order. A part with draft, uniform section and no undercuts can be cast or forged and then machined only on its functional surfaces. A part with deep pockets and thin ribs may never be economical to mould. A simple rotational part at volume is usually better cold-formed or headed than turned.

Prototype to production CNC machining is a transition, not a switch. The intermediate step is often to keep the CNC process and change the blank: machine from a near-net forging or casting instead of a solid billet, which cuts stock removal while preserving the geometry that is already validated. This keeps the drawing stable and defers tooling investment until the design has stopped changing.

Where the crossover sits depends on the part. It is a function of removal ratio, tolerance, surface requirement, tooling cost and expected life of the design. Any supplier quoting a fixed volume threshold without seeing the part is guessing.

What a Useful DFM Review Returns

A DFM review before quotation is not a courtesy. It is the step where the cost of a part is either accepted or designed out, and it works best when the input is complete.

Send the geometry, not a description. A 3D model establishes shape; a 2D drawing establishes tolerance, datum and finish. Sending only one of the two guarantees a round of questions.

State the function. Which faces seal, which bores locate, which surfaces are seen, and what the part assembles into. Function is what allows a supplier to distinguish a tolerance that must be held from one that was copied from a template.

State the volume and the stage. A first prototype and a production release are different exercises, and the recommended process route differs accordingly.

Expect specific feedback, not a price alone. A useful review flags unreachable corners, over-tight tolerances, unnecessary finish callouts, features that force extra setups and thread specifications that will not produce the intended engagement. Those comments are worth more than a small price difference, because they recur on every order.

Send the Drawing and Let the Process Be Decided

Design decisions are easier to make with the manufacturing route visible. GANGGU has produced hardware fasteners and precision CNC machined parts since 2011, which means the same review can cover a machined component and the screw or insert that fits into it.

Send a 3D model and a 2D drawing to ggscrew.com/contact, with the application, expected volume and any surface that must seal or locate. If the geometry suggests a different process route, that will be said before a quotation rather than after a failed first article.

To see the range of machined work, browse custom CNC machined parts or CNC parts by industry.

Frequently Asked Questions

What internal corner radius should I specify?

At least the radius of the cutter that can reach that depth, and preferably larger. A corner radius smaller than the available tool forces a smaller cutter, additional passes or a secondary process such as EDM, all of which increase cost without improving function.

How tight should tolerances be on a machined part?

Only as tight as the function requires. Control the features that locate, seal, bear load or mate, and apply general tolerances elsewhere. Every additional controlled feature adds inspection work, and tolerance that is never measured adds cost without adding control.

Does a surface finish callout change the price?

Yes, when it is applied broadly. A fine Ra requirement on a functional face is a legitimate cost. The same requirement applied to every surface adds finishing passes and inspection on faces that never need them.

How deep should a threaded hole be?

Engagement of roughly one to one and a half times the nominal diameter is a common working rule in steel, with more in softer materials. Specify thread depth and drill depth separately, and state the thread standard and tolerance class so the hole is produced as intended.

Why does an extra setup increase cost?

Each setup re-establishes the part's position from a new reference, which adds fixturing time, inspection and positional variation between faces. Grouping critical features onto fewer faces usually reduces cost more than loosening any tolerance.

When should a part move from CNC to casting or forging?

When the geometry is stable, the volume justifies tooling and the material suits the process. A common intermediate step is to keep the CNC route but machine from a near-net blank, which reduces stock removal without changing a validated drawing.

Can thin walls be machined reliably?

Yes, with planning. Thin sections deflect under cutting force and vibrate, so the process needs support, staged machining and sometimes stress relief. Uniform wall thickness and ribs where stiffness is needed make the part both cheaper and more consistent.

What should I send for a DFM review?

A 3D model for geometry, a 2D drawing for tolerance, datums and finish, plus the application, expected volume, critical surfaces and the mating hardware. Complete input usually produces a single review with actionable comments instead of several rounds of clarification.

About GANGGU — Dongguan Ganggu Hardware Product Co., Ltd. has developed and manufactured hardware fasteners and precision CNC machining services since 2011, with production bases in China and Vietnam under an ISO9001:2015 quality management system. Contact: angel@ganggus.com ｜ WhatsApp +86 18602085825