Design for CNC Machining: What Good Looks Like Before the First Cut

Most of a part's cost is determined before programming begins. The number of setups, the required tooling, the inspection burden, the cycle time — all of it is baked into the design. A well-designed part is cheaper to quote, faster to run, easier to inspect, and more likely to come out right on the first run. A poorly designed one pays a tax on every operation.

This isn't a list of rules. It's a practical account of what separates parts that machine cleanly from parts that generate feedback, rework, and cost overruns.

Tolerances: The Single Biggest Cost Driver

Tight tolerances aren't free, and the cost isn't linear — it compounds. Tighter tolerances require slower feeds, additional finishing passes, sometimes specialized tooling, and more intensive inspection. When you call out ±0.001" on a feature that functions fine at ±0.005", you're not improving quality. You're paying for a constraint the part doesn't need.

As a general starting point:

• General machining: ±0.005" (0.13mm) — achievable on well-maintained equipment without special process control
• Precision features (bores, mating interfaces): ±0.002" (0.051mm) — slower operations, more care
• Tight-tolerance / reamed holes: ±0.0005" — requires dedicated tooling and deliberate setup strategy
• Plastics: ±0.010" as a baseline — material movement and thermal expansion work against you

The right approach is to reserve tight tolerances for features where they directly affect function: bearing bores, locating interfaces, sealing surfaces, and critical mating geometry. Everything else should carry block tolerances. Applying tight tolerances to non-functional geometry is one of the most consistent ways customers add cost without adding value.

One thing worth understanding: when tolerances tighten, inspection changes too. A part that can be gauged with calipers at ±0.005" may require a CMM run at ±0.001". That time costs money and usually gets passed through.

Wall Thickness: Thin Walls Cause Real Problems

Thin walls are a reliable source of scrap and rework. The cutting forces involved in milling and turning aren't trivial — thin sections vibrate, deflect, and sometimes break. Residual stress in the raw stock tends to release as material is removed, and thin walls amplify that movement.

As a practical starting point:

• Metals: Minimum recommended wall thickness is around 0.040" for most applications. Walls approaching 0.020" are at meaningful risk of chatter, deflection, or breakage.
• Plastics: More compliant, so the threshold is higher in practice — 0.060" is a reasonable working minimum. Thin plastic walls also tend to warp after machining as stress releases.

If thin walls are required functionally, design stiffening geometry (ribs, gussets) where the envelope allows. For aluminum parts with large pockets machined from plate, some bow or warpage should be expected — particularly in elongated parts. Stress-relieved stock helps; designing in machining stock that gets cleaned up in a final facing pass helps more.

Pockets: Depth, Width, and Corners

Every pocket design is a negotiation between feature geometry and available tooling. Standard end mills have a practical cutting length of roughly 3–4x their diameter before deflection and chatter become real problems. Deep, narrow pockets require long-reach tools, slower parameters, and more passes — all of which drive cost.

A few practical guidelines:

• Depth-to-width ratios beyond 4:1 start requiring non-standard tooling or multiple tool lengths. Beyond 6:1, you're in specialized territory.
• Internal corner radii must be greater than zero — a rotating cutter cannot produce a true inside square corner. As a general rule, r ≥ 1/3 of pocket depth keeps standard tooling accessible. Minimum practical radius is typically around 0.030".
• Adding a small amount of clearance beyond the exact tool radius (e.g., 0.140" radius for a 0.125" cutter) reduces cutting force and allows climb milling on the finish pass. It's a small detail that improves surface quality.

Undercuts — features below a ledge or on a surface the spindle can't reach axially — require T-slot cutters or lollipop-style tools and add setup complexity. When they appear in a quote, they're noted. Design them out if there's no functional reason they need to be undercuts.

Holes: Use Standard Sizes, Specify Depth Correctly

Non-standard hole diameters require interpolation with an end mill, which is slower and more expensive than drilling. Where function permits, design holes to standard fractional, letter, or numbered drill sizes — it reduces cycle time and is easier for the shop to tool without special orders.

Depth guidance:

• Recommended max drilled depth: 4x diameter with standard tooling
• Deeper holes are feasible with extended drills, typically up to 10x diameter, with increasing attention to chip evacuation
• Blind holes drilled with a twist drill have a conical floor at ~135°. If you need a flat bottom, that requires an end mill — design that in and call it out explicitly

For tight-tolerance holes, specify reaming. Reamed holes to ±0.0002–0.0005" are reliable and repeatable. Interpolated bores with an end mill can hold tight tolerances but require more setup care and process verification.

Threads:

• Through-threaded holes are simpler and more reliable than blind tapped holes. Use them when geometry allows.
• Minimum thread engagement in aluminum: approximately 1x nominal diameter; in steel, 0.75x is often adequate, though application-specific
• Specify thread class (2B standard, 3B precision) on the drawing if fit tolerance matters. "Tapped hole" with no further callout leaves interpretation to the machinist.

Setup Count: Often More Important Than Complexity

Every time a part is re-fixtured, the machine stops, the program changes, and a new datum must be established. Each additional setup introduces a small but real risk of datum shift between operations — a hole pattern located on setup one may have a slightly different relationship to a bore machined on setup three than your drawing assumes. On simple parts this is usually negligible. On parts with tight positional relationships between features on different faces, it's worth designing carefully.

Parts that can be machined in two setups (typically op1/op2, top and bottom) are significantly cheaper and more consistently accurate than parts requiring four, five, or six setups. When designing:

• Consolidate features onto as few faces as possible
• Keep critical positional relationships between features that can be machined in a single setup
• For milled parts requiring features on multiple faces, design in clear primary locating geometry so the shop can re-datum consistently
• For turned parts, design features to be on-axis or accessible from one end where possible

When a part genuinely requires multi-face access, the right tool is 5-axis machining or a tombstone setup — not a series of manual re-fixtures. Plan for the process from the design stage.

Surface Finish: Functional Spec, Not Aesthetic Preference

Surface finish (Ra) is not a cosmetic preference — it's a functional specification that affects sealing, wear, fit, and sometimes fatigue life. Calling out a tighter finish than the surface requires adds machine time and sometimes post-processing cost.

Typical as-machined finishes:

• Flat and perpendicular milled surfaces: Ra 63 µin (1.6 µm) — standard, adequate for most structural and non-contact surfaces
• Curved and profiled milled surfaces: Ra 125 µin (3.2 µm) or better

When to specify tighter:

• O-ring seating faces, gasket surfaces: Ra 32–63 µin
• Bearing bores, shaft journals, sliding fits: Ra 32 µin or per bearing manufacturer specification
• Cosmetic visible surfaces: call it out as a cosmetic requirement; the machinist can optimize the finish pass

One detail often missed: for sealing surfaces, direction of lay matters. A finish with lay parallel to the sealing direction leaks more easily than one perpendicular to it. If it matters, specify it.

Engraved Text and Part Marking

Machined text is significantly more expensive per character than it looks. Each character is traced with a small ball or V-groove end mill, consuming machine time proportional to total path length. Across a production run, it adds up.

For production quantities, consider laser marking, electrochemical etching, or applied labels as alternatives. If machining is required:

• Sans-serif fonts machine cleanly and are easy to program
• Minimum practical text height: 0.100" for clean results
• Call out engraving depth explicitly — typically 0.010–0.020"

DFM Checklist

Before sending a file to any machine shop:

• [ ] Wall thickness ≥ 0.040" for metals, ≥ 0.060" for plastics
• [ ] Pocket depth-to-width ratio ≤ 4:1 for standard tooling
• [ ] Internal corner radii specified and non-zero (typically ≥ 0.030")
• [ ] Tight tolerances applied only to functional features; block tolerances cover the rest
• [ ] Hole diameters are standard sizes where function permits
• [ ] Blind hole floors specified (conical drill floor vs. flat end mill floor)
• [ ] Thread engagement, class, and fit specified
• [ ] Feature consolidation minimizes setup count
• [ ] Surface finish callouts on functional surfaces only, with lay direction noted where it matters
• [ ] 3D model and 2D drawing are synchronized

The Point

Good DFM doesn't mean simpler parts. It means parts whose geometry reflects how they'll actually be made. The goal is a design where the machinist reads the print, programs the part, and runs it — without writing back to ask questions or making judgment calls that should have been engineering decisions.

When we quote a well-DFM'd part, we can price it accurately and turn it around fast. When we quote something with avoidable issues, we have to account for the risk, and that goes into the number.

If you want DFM feedback before you finalize your drawing, submit the model early. We review geometry before quoting and flag what will drive cost — at no obligation.

Request a DFM review from Axion Labs →

CNC Machining Material Selection: A Decision Guide, Not a Catalog

The most common material mistake we see isn't choosing something too weak — it's choosing something harder, heavier, or more expensive than the application actually justifies. A bracket that works fine in 6061 gets specced in 316 stainless because the engineer defaulted to "corrosion resistant." A housing gets quoted in PEEK because someone in the meeting mentioned "high performance." The result is a part that costs three times more, takes longer to machine, and performs identically in service.

Material selection should be driven by function, environment, and manufacturing cost — in that order. This guide walks through the materials we machine most often and the decisions that matter.

The Properties That Actually Drive Selection

Tensile strength is a starting point, not a conclusion. In practice, the questions that drive material choice are:

• What load case does this part see? Static, cyclic fatigue, and impact all point to different materials and required properties.
• What environment does it live in? Temperature, chemical exposure, UV, and humidity can eliminate entire material families.
• How tight are the tolerances? Some materials hold dimensions well; others move after machining. Nylon in a high-humidity environment is a different part than Nylon in a dry lab.
• What does it need to look like? Anodize requires aluminum. If aesthetics matter, that constrains the list.
• What's the volume and lead time? Common alloys (6061, 1018, 304 SS) ship from every distributor. Exotic alloys in non-standard stock sizes add days or weeks.

Aluminum: The Right Default — With Caveats

Aluminum is the most commonly machined metal for good reason. It machines fast, tolerates high spindle speeds, finishes cleanly, and works well with anodizing or chromate. When in doubt, start here and justify switching away.

6061-T6 is the true all-purpose choice. Good corrosion resistance, excellent weldability, readily available in plate, bar, and extrusion. Tensile strength around 45 ksi. For structural brackets, enclosures, housings, and general-purpose machined components, 6061 is usually correct.

7075-T6 offers roughly double the strength of 6061 (~83 ksi) with a meaningful trade: lower corrosion resistance, poor weldability, and inconsistent hard anodize results. It's justified in high-stress aerospace or performance applications. We see it over-specified frequently — if your load analysis doesn't require it, 6061 costs less and machines more predictably.

2024-T3 has high copper content (3.3–4.6%) and excellent machinability, particularly for threading and high-speed work. Common in aircraft structures and fastener-type components. Low corrosion resistance — not appropriate for outdoor or marine use without protective coating.

As a general rule, aluminum holds ±0.005" without difficulty on well-maintained equipment. ±0.001" is achievable with proper fixturing and tooling, but it isn't free — factor that into your tolerance callouts.

Steel: When You Actually Need It

Steel's advantages over aluminum are strength, hardness, and wear resistance. Its disadvantages are weight and machinability — it cuts slower, wears tooling faster, and adds cycle time. Specify it when the part genuinely needs what steel provides.

1018 cold-rolled is the standard low-carbon choice. Excellent machinability, easy to weld, moderate strength (~64 ksi tensile), not hardenable. Right for shafts, spacers, brackets, and structural weldments where hardness isn't required.

4140 alloy steel is the workhorse for tooling, fixtures, gears, and high-load shafts. Tensile strength ranges from roughly 95 to 148 ksi depending on heat treat condition. Machines well in the annealed state; pre-hardened 4140 cuts noticeably harder. Specify the condition clearly on your drawing.

Tool steels (D2, A2, H13) are for applications requiring hardness — punch and die components, mold inserts, wear surfaces. Machine in the annealed state; finish dimensions after heat treat typically require grinding. Don't expect a general-purpose machine shop to take these to final hardened dimensions without grinding capability.

Stainless Steel: Frequently Over-Specified

Stainless is the material customers reach for whenever "corrosion resistance" appears in the requirements, without always checking whether it's actually needed. It costs more to machine than carbon steel, requires slower speeds, more coolant, and more tool wear. Specify it when the environment genuinely demands it.

303 is the free-machining grade — added sulfur improves chip breaking and makes it significantly easier to machine than 304 or 316. Use it for turned parts, fasteners, and fittings where corrosion resistance matters but chloride exposure is limited. Its sulfur content reduces weldability and slightly reduces corrosion resistance compared to 304.

316/316L is appropriate for marine environments, food processing, medical, and applications with chloride exposure. The molybdenum addition (2–3%) is what separates it from 304 in harsh environments. Harder to machine than 303 — budget for longer cycle times.

17-4 PH (H900 condition) reaches tensile strengths around 190 ksi while retaining corrosion resistance. Used in aerospace and medical applications requiring both. Specify the H condition on the drawing; it changes the material properties substantially.

A useful check: if the part will be painted, powder coated, or plated, you may not need stainless at all. Many stainless specs are driven by surface appearance, not corrosion performance — a coated carbon steel part often serves the same function at lower cost.

Titanium (Ti-6Al-4V): High Performance, High Cost

Grade 5 titanium offers roughly the strength of steel at about 45% of the weight, with excellent corrosion resistance and biocompatibility. Tensile strength around 130 ksi. The case for titanium is clear in aerospace structures, medical implants, and high-end performance hardware where strength-to-weight is a hard requirement.

The case against is equally clear everywhere else. Titanium machines slowly. Its low thermal conductivity traps heat at the cutting edge, requiring conservative speeds, appropriate tooling, and flood coolant management. Cycle times are typically several times longer than equivalent aluminum parts. Material cost is high, and the combination means titanium parts carry a significant price premium.

If your application can be served by 7075 aluminum or 4140 steel, use one of those. Titanium earns its cost when nothing else meets the requirement.

Brass and Copper: Niche, But Worth Knowing

C360 free-machining brass is one of the easiest metals to machine — fast cycle times, excellent surface finish, clean chips. It's the right choice for fittings, valves, electrical contacts, and decorative components. Lead content makes it a good cutting brass; check regulatory requirements for food or medical contact.

Copper (C110 and similar) is soft, gummy, and prone to built-up edge on tooling. Requires sharp tools and careful chip management. Justified when electrical or thermal conductivity is the primary requirement — bus bars, heat sinks, RF shielding.

Engineering Plastics: Underused and Misunderstood

Plastics often outperform metals on weight, chemical resistance, or electrical isolation — and they machine faster and more cheaply. They're also less forgiving on tolerances because thermal expansion, moisture absorption, and material variability all work against you.

Delrin (POM/acetal) is the most commonly machined plastic for good reason. Dimensionally stable, low friction, self-lubricating, and easy to machine cleanly. Good for gears, bushings, sliding components, and food-contact parts. Works well to approximately ±0.003–0.005" under controlled conditions; tighter than that starts requiring process care.

PEEK is a high-performance thermoplastic with excellent mechanical properties, chemical resistance, and biocompatibility. Retains meaningful strength at elevated temperatures — continuous service in the 450–480°F range for unfilled grades. It's the right choice for demanding medical, semiconductor, and aerospace applications. It's expensive — both the material and the machining — so don't specify it unless the application actually requires it. We see PEEK specified on parts that could have been Delrin or even nylon.

Nylon (PA6/PA66) is tough, wear-resistant, and cost-effective. The meaningful limitation is moisture absorption — nylon can absorb several percent of its weight in water, which affects dimensional stability. For loose-tolerance applications in dry environments it's excellent; for tight-tolerance parts in humid or immersed environments, it requires careful design and often conditioning before final machining.

PTFE (Teflon) has outstanding chemical resistance and the lowest friction coefficient of any common solid. It's soft, which creates problems holding tight tolerances — material creep and springback during machining are real. Use it for seals, gaskets, and chemical-resistant liners where loose tolerances are acceptable.

Polycarbonate offers good impact resistance and optical clarity. Machines well, but is susceptible to chemical attack from solvents, some lubricants, and certain cutting fluids. Right for machine guards, optical components, and electrical enclosures.

Quick-Reference Comparison

Tensile strength values are representative; actual values vary with temper, heat treat condition, and test standard.

How We Approach Material Questions at Axion

When a job comes in without a material specified — or with one that looks questionable for the application — we ask before quoting. Material affects cycle time, tooling, inspection approach, and cost in ways that aren't always obvious from the CAD model.

The most common conversation is about stainless and titanium: customers specifying them for properties the application doesn't actually require. The second most common is plastics: customers defaulting to aluminum when a well-chosen engineering plastic would serve the same function at lower weight, lower cost, and better chemical resistance.

If you're not sure what material is right for your part, describe the application when you submit your RFQ. We'll tell you what we'd use and why.

Request a quote from Axion Labs →

GD&T for CNC Machined Parts: A Practical Guide From the Shop Floor

GD&T is widely misunderstood, inconsistently applied, and regularly blamed for manufacturing problems it didn't cause. The actual problems are usually simpler: drawings that reference datums never defined, tolerances tighter than any machine can verify, and callouts applied to every feature on a print regardless of whether they affect function.

Used correctly, GD&T makes a drawing cleaner and more functional — it communicates exactly what matters, and nothing more. This guide covers how GD&T applies to CNC machined parts specifically: which callouts we see on most jobs, what they actually control, what inspection they require, and the mistakes that drive cost without improving outcomes.

What GD&T Controls That Linear Tolerances Don't

Standard dimensional tolerances control size. A hole dimensioned at 1.000" ±0.005" in X and ±0.005" in Y defines a square tolerance zone — 0.010" × 0.010". The actual hole can be anywhere within that square, including in the corner, and still be "in tolerance." But a circular pin being located to mate with a circular clearance hole doesn't care about corners. The functional tolerance zone is circular.

True position replaces those two linear tolerances with a circular zone of diameter φ0.014" — same worst-case footprint, but now the zone reflects how the feature actually functions. On a bolt pattern, this matters: parts that fail coordinate tolerances sometimes pass true position and assemble fine, and vice versa.

This is the central purpose of GD&T: making the drawing reflect the geometry of the actual functional requirement, not just the convenience of a coordinate dimension.

It's governed by two major standards:

• ASME Y14.5 (current edition: 2018) — the standard for most North American work
• ISO 1101 / GPS — common in European and international contexts

The two are similar in intent and share most symbology, but differ in some interpretation details — particularly around material condition modifiers and datum precedence. Specify which standard applies in your title block.

The GD&T Controls You'll Actually Use

GD&T has five formal categories and fourteen-plus symbols. On most CNC machined parts, the following cover the overwhelming majority of practical requirements:

True Position

The most commonly applied GD&T control in machined parts. Controls where a feature — typically a hole or boss — is located relative to a datum reference frame.

True position gives you a cylindrical tolerance zone whose axis must contain the feature axis. Combined with an MMC (Maximum Material Condition) modifier on bolt-pattern holes, it provides bonus tolerance as the hole size departs from its minimum — a real advantage in clearance-fit applications.

Use it for: bolt patterns, locating pins, holes that must align with mating components. The right tolerance value always comes from a stack-up calculation, not convention.

Flatness

Controls the form of a single surface — how flat it is, independent of any datum or other feature. Two parallel planes define the zone; the surface must lie entirely within them.

Commonly confused with surface finish (Ra). They're not the same. Flatness is macrogeometry — the overall shape of the surface. Surface roughness is microgeometry — the texture. A surface can be flat but rough, or smooth but warped. Both may need to be controlled, but they're specified separately.

Use it for: sealing faces, mating surfaces, precision bases. A flatness callout of 0.001" on a general mounting plate is a waste of money; the same callout on a valve body sealing face may be exactly right.

Perpendicularity

Controls the 90° angular relationship between a feature and a datum. Applied to a surface, it controls how square the wall is. Applied to a hole axis, it controls whether the bore runs truly perpendicular to the datum face — important for press fits, clearance alignment, and thread engagement across interfaces.

This callout comes up frequently on machined parts with bores that must align with mating bores on an assembly. The perpendicularity tolerance typically needs to be tighter than the general angular block tolerance, and it often matters when multiple setups are involved.

Runout (Circular and Total)

Runout controls how a surface varies as the part rotates about a datum axis. Circular runout measures each individual cross-section; total runout controls the full surface simultaneously.

For rotating components — shafts, spindle housings, anything on an axis — runout is more practical than concentricity. It's measurable with a dial indicator against the datum, which most inspection setups can accommodate without a CMM. Total runout is called out most frequently on turned components where the relationship between an OD, shoulder face, and bore matters for assembly or rotational balance.

The Others

Cylindricity, profile, angularity, parallelism, concentricity, and symmetry all have legitimate applications. Cylindricity is useful for precision bores where form across the full length matters beyond what a diameter tolerance controls. Profile of a surface works well for complex curves and freeform geometry. Concentricity is technically correct but rarely the right choice over runout — it's more difficult to measure and more rarely necessary.

Datums: Where Most Drawing Errors Start

Every GD&T callout references a datum or datum reference frame. Getting this wrong is the most common fundamental error we see on customer drawings — and it usually shows up as either datums referenced in feature control frames that aren't labeled on the drawing, or datum surfaces that look fine on paper but are difficult to contact repeatably in the shop.

The principle that matters most for machining: your datum surfaces should mirror your fixturing and machining surfaces. If datum A is the bottom face, and the part sits on its bottom face in the vise, the machinist and inspector are working from the same reference. If datum A is a small, non-planar edge that can't be fixtured against, you've created a gap between what the drawing specifies and what can be practically verified.

Datum hierarchy: Primary datum (A) constrains the most degrees of freedom — minimum three contact points for a planar surface. Secondary (B) adds orientation — two points. Tertiary (C) locks the final constraint — one point. The order matters: reordering A and B changes how the measurement is set up and can change whether a part passes or fails inspection.

Practical rules:

• Datum off large, accessible, machined surfaces — not small edges, thin walls, or surfaces that warp
• When parts require multiple setups, break tolerance zones by setup to avoid accumulating datum shift between operations
• Confirm with your shop which face represents datum A before cutting begins — don't leave it as an assumption

CMM Inspection and What It Actually Costs

The inspection method required is a direct function of which GD&T callouts appear on your drawing.

Flatness on a simple surface can be checked with a surface plate and a dial indicator. Perpendicularity of a wall to a face can often be verified with a precision square. Runout is straightforward with an indicator on a known axis.

True position, cylindricity, and profile of a surface typically require a CMM — a multi-point 3D measurement against a full datum reference frame. CMM time is real time, usually quoted per feature or per setup. When you add a true position callout to a 24-hole bolt pattern, you're not just adding a symbol to the print — you're potentially adding 24 measurement points to an inspection report.

We don't tell customers to avoid GD&T. We tell them to apply it where it's justified. Every callout that requires CMM verification should be there because the feature matters functionally — not because it seemed thorough.

Over-Tolerancing: The Most Expensive Drawing Mistake

Over-tolerancing is applying tighter geometric controls than the part actually requires. It's common, and it's consistently the source of parts that are expensive to produce and no better in service.

A few examples we see regularly:

• Flatness of 0.001" on a non-sealing mounting bracket — 0.010" would be indistinguishable in function
• True position of φ0.002" on clearance holes with 0.015–0.020" of pin-to-hole clearance — the extra precision disappears immediately in assembly
• Cylindricity on a bore that seats a commercial bearing — the bearing spec is looser than the callout

The machinist response to over-tolerancing is rational: slower speeds, more finishing passes, 100% inspection of that feature. All of that time costs money. The quality outcome doesn't change because the part doesn't need the precision.

The right approach: reserve geometric tolerances for features that are critical to function. If you can't trace a path from that tolerance to a specific assembly requirement, consider whether it needs to be there.

Common Drawing Errors That Cause Problems in the Shop

These aren't hypothetical — they appear in submitted drawings regularly:

Unreferenced datums. Feature control frames call out "A" or "B" but no surface on the drawing is labeled as datum A or B. The machinist has to guess, or ask.

CAD/drawing discrepancies. The 3D model and the 2D drawing don't agree. A hole that was moved in CAD wasn't updated on the print. The shop uses the print, makes a part to the wrong dimension. Always reconcile the two before submitting.

Over-constrained drawings. More geometric requirements than the part has degrees of freedom to satisfy. This creates specifications that are mathematically impossible to meet simultaneously.

Missing material condition modifiers. On bolt-pattern true position callouts, the lack of an MMC modifier means you're not getting bonus tolerance. For clearance-fit fastener patterns, specifying MMC is almost always the right call.

Profile without datums. Profile of a surface without datum references is an incomplete control — it specifies what the shape should look like but not where it should be located.

Before Adding Any GD&T Callout, Answer These Four Questions

1. What specifically fails if this isn't controlled? If the answer is vague, the callout is probably unnecessary.

2. What's the required value? Back-calculate from your assembly stack-up or mating part clearance. Don't pick a number that "seems tight enough."

3. Can it be verified? If it requires a CMM program to confirm, that's a real cost — make sure it's worth it.

4. Is the general block tolerance adequate? ISO 2768-m or your title block standard covers most non-critical geometry. Default to that unless you have a specific reason not to.

A Note on Drawing Quality

A well-controlled GD&T drawing builds trust. When we receive a drawing with clean datum structure, appropriate callout density, and tolerances that clearly reflect functional intent, we can quote it accurately and run it confidently. When we receive a drawing with every surface controlled to tight GD&T, datums that don't match the fixturing geometry, and callouts that seem generated rather than designed — we slow down and ask questions. That friction has a cost.

Good drawings reduce ambiguity between what the designer intended and what the machinist produces. That's the goal. GD&T is a tool to get there — not a measure of how thoroughly a drawing was reviewed.

Send us a drawing for review →

How to Source CNC Machined Parts: What Matters When Evaluating Shops and Structuring Orders

Getting good machined parts isn't just a function of finding a shop that can run the job. It's a function of communicating clearly, understanding what drives cost, and choosing suppliers whose capabilities match your actual requirements. Most sourcing problems we hear about come back to one of three things: an incomplete RFQ, a tolerance spec the supplier couldn't hold and didn't say so upfront, or a prototype shop that couldn't scale when the design got locked.

This guide covers the mechanics of sourcing CNC machined parts — what information needs to be in your RFQ, what actually drives the quote number, and how to evaluate a shop before committing production work.

Two Sourcing Models — and When Each One Makes Sense

The CNC machining supply market broadly divides into two approaches:

On-demand quoting platforms let you upload a 3D model, select material and finish, and receive a price without talking to anyone. Turnaround on standard geometry and materials can be very fast. The tradeoff is that the feedback loop is automated — DFM review is algorithmic, and complex or unusual work often doesn't quote as expected. These platforms generally serve straightforward work well. When parts involve tight tolerances, unusual geometry, custom fixturing, or ongoing production relationships, the automated model starts to show its limits.

Direct machine shop relationships offer more communication bandwidth: the ability to discuss the part before quoting, get real DFM feedback, negotiate on lead time and quantity, and build familiarity with your parts over time. This matters more as complexity and volume increase. A shop that has run your parts twenty times understands your tolerances, your quality expectations, and where you're flexible — that institutional knowledge reduces risk on every subsequent order.

Neither model is appropriate for all work. Many engineering teams use both: on-demand platforms for quick prototype iterations, direct shop relationships for production and complex geometry. The mistake is treating them as interchangeable.

What Drives CNC Machining Cost

Understanding the cost structure helps you write better designs, ask better questions, and evaluate quotes with more confidence.

Material typically represents somewhere between 20–40% of part cost on metal components, though this varies significantly with part geometry and material choice. Aluminum is the lowest-cost structural metal to machine. Titanium and nickel alloys carry both high material cost and high cycle-time cost — a combination that multiplies quickly.

Setup and fixturing are often the largest single cost element on low-volume work. Every setup requires machine downtime, program loading, fixture positioning, and first-article inspection before cutting begins. A part requiring five setups can cost considerably more than a geometrically simpler part requiring two — even if the final machined geometry looks similar. This is why design for minimal setups is a real cost lever, not just an abstract principle.

Cycle time is a function of material removal rate, toolpath complexity, and the constraints imposed by tolerances and surface finish requirements. Tight tolerances don't just affect individual operations — they affect the pace of the entire job and the inspection burden at the end.

Tolerances and inspection are frequently underestimated cost drivers. Standard tolerances can be verified efficiently with conventional gauging. Tight-tolerance features often require CMM time, dedicated gauges, or 100% inspection rather than sampling. That inspection time is real, and it typically shows up in the quote.

Quantity is the most obvious cost driver but also the most misunderstood. The per-piece cost reduction going from one to ten parts is steep. From ten to fifty, meaningful. From fifty to two hundred, smaller. Most of the unit-cost curve is driven by setup amortization — the fixed cost of programming and fixturing spread across more parts. Don't ask for a production price on a prototype, and don't expect a prototype price to hold at production volume.

Post-processing — anodizing, passivation, black oxide, powder coat, plating — all add lead time and cost. These services are typically subcontracted unless the shop has them in-house, which adds a handoff. Plan for it in your lead time estimate and request vendor certifications when they're relevant.

What Makes an RFQ Work

A good RFQ is complete. The more the shop has to guess, the less reliable the quote — and gaps in an RFQ almost always resolve in the shop's favor when the job is running, either by adding cost or pushing back delivery.

Minimum contents of a useful RFQ:

• STEP file (or other solid model format) — communicates geometry
• 2D drawing with material, tolerances, surface finish callouts, and GD&T where applicable. A 3D model without a drawing communicates geometry only; it does not communicate how the part should be made or inspected.
• Quantity tiers — if volume matters, request pricing at multiple quantities (1, 10, 50, etc.)
• Required certifications — material certs, certificates of conformance, first article, ITAR, AS9100. Don't mention these for the first time after the part is shipped.
• Delivery requirement — actual date or lead time expectation
• Post-processing — specify finish, color if relevant, plating spec, or "as-machined" explicitly

The most common RFQ failures we see:

• Sending only a CAD model with no drawing, then specifying tolerances after the quote comes back
• Writing "per drawing" when no drawing exists or the drawing has no tolerance callouts
• Specifying material as "aluminum" or "steel" without alloy and temper — 6061-T6 and 6061-O are not the same material; 4140 annealed and 4140 pre-hardened are not interchangeable
• Not specifying thread callouts to a standard (UNC, UNF, metric) — "tapped hole" is not a specification

A complete, unambiguous RFQ typically gets a more accurate quote, faster. It also tells the shop something about the customer — shops prioritize work from customers who communicate clearly.

Evaluating a Machine Shop

Before placing production work with any shop, consider these questions:

Capability

• What equipment do they run? 3-axis, 4-axis, 5-axis, turning, Swiss, EDM? The machine mix determines what geometries are accessible.
• What tolerances can they reliably hold, and can they demonstrate it? Inspection reports from previous jobs are more credible than claims on a website.
• Do they have in-house inspection — CMM, surface plate, optical comparator, dedicated gauging? Or is inspection subcontracted?

Quality System

• ISO 9001 certification is a baseline indicator of a documented quality management system. It's not a guarantee of good parts, but its absence on production work should raise questions.
• AS9100D for aerospace; ISO 13485 for medical devices. If your application requires these, verify registration — don't accept self-reported compliance.
• Material traceability: can the shop provide material certs and trace a lot number back to the raw stock used? For regulated industries, this is non-negotiable.

Engineering Engagement

This is undervalued and hard to assess from a website. The question is: when you send a drawing, do you get feedback on the design, or just a price?

A shop that reviews your drawing, identifies manufacturability issues, and tells you about them before cutting has started is worth something concrete. It reduces scrap, rework, and the back-and-forth that happens when a part comes back wrong. The alternative is a shop that executes exactly what's on the print — including the mistake you didn't catch — and invoices accordingly.

Ask directly: do they review drawings for DFM before quoting? Who does the review — a salesperson, or an engineer or machinist? What does their feedback look like?

Track Record

Ask for inspection reports from recent production runs on comparable parts. Request references from customers in similar industries. Start with a prototype or a small batch before committing to a production program — the fastest way to assess a shop's actual capability is to run a part through it.

Prototype to Production: Managing the Transition

Prototype and production work have different requirements, and treating them the same is a reliable way to create problems.

At the prototype stage, speed and design flexibility are the priorities. You expect to revise. Per-part cost is high and acceptable — the value is the learning. Tolerances may be loosened to reduce cycle time; finish may be as-machined rather than anodized. That's appropriate.

At the production stage, repeatability and per-piece cost matter. The design should be frozen. Process controls, inspection plans, and quality documentation need to be in place. Surprises — the kind that are acceptable during prototyping — are not acceptable at volume.

The transition typically needs to address:

1. Design lock. The design should be formally released before first article inspection. Changes after first article are expensive — they may invalidate the inspection, require refixturing, and restart the qualification process.

2. Fixturing ownership. Custom fixtures add cost to the first run. Clarify before that run who owns the fixture if you change shops — this becomes a real issue more often than buyers expect.

3. Inspection plan. Which dimensions are controlled? What's the inspection frequency — 100%, first article only, periodic sampling? What AQL level applies? Document this agreement in writing.

4. Process documentation. For regulated industries (medical, aerospace, defense), a control plan and FMEA may be required. Even for commercial work, understanding how the shop will detect and flag non-conforming parts before shipping is basic due diligence.

5. Cost negotiation. Production pricing should reflect amortized setup, tooling, and programming. If the prototype quote included NRE, that should not be charged again on subsequent runs using the same fixtures and programs.

Why the Shop Relationship Matters

The cleanest way to put it: a shop that knows your parts doesn't have to figure them out again on every run. The first time a job runs, the machinist is solving problems — finding the right workholding, optimizing the toolpath, verifying that the tolerances are achievable. By the fifth run, that work is done. The fixturing is proven, the program is dialed, the inspection is efficient. Per-part quality goes up and per-part cost goes down.

That's the value of a production relationship versus transactional sourcing. It shows up in consistent part quality, tighter lead times, and fewer conversations that start with "we have a quality issue."

The question when evaluating a shop isn't just "can they make this part?" — it's "are they the kind of shop we want knowing our parts?"

At Axion Labs, we review every drawing before quoting and provide DFM feedback as part of the process — not as an add-on. If you're sourcing a part for the first time or looking for a production partner for ongoing work, send us your drawing and let's talk about what you're building.