Why CNC cost is mostly a design problem.
Engineers often assume CNC cost is set by the shop — labor rate, overhead, how aggressive the quote model is. In our experience quoting 14,000+ parts a year, the shop accounts for maybe 30% of the variance between a cheap part and an expensive one. The other 70% is baked into the CAD before anyone quotes it.
Below are the ten design choices we see most often that inflate cost without adding functional value. Each includes the approximate cost multiplier based on comparable jobs we quoted in 2025, and a suggested fix.
§01 — Sharp internal corners.
Internal corners on milled parts must be cut with a round tool. A corner with R0.5 requires a Ø1 mm end mill — which has low rigidity, slow feed, and short tool life. Zero-radius corners can't be milled at all; they require EDM or sinker broaching.
Cost delta: An aluminum bracket with R0.5 internal corners throughout quoted at $42/part (qty 50). The same part with R3 corners quoted at $18. A 2.3× increase for corners that had no functional purpose.
§02 — Deep pockets and thin walls.
Deep pockets (depth > 3× tool diameter) require either long reach tooling (fragile, slow) or plunge machining (very slow). Thin walls below 1 mm chatter, work-harden, and often require spark-out passes.
Cost delta: A 6061 housing with 0.8 mm walls and 40 mm deep pockets quoted at $180/part. Same housing with 1.5 mm walls and 30 mm pockets: $72. That's a 2.5× premium for thinner walls that weren't load-critical.
§03 — Tight tolerances on every dimension.
This is the single biggest cost inflator we see. Drawings with ±0.02 mm on every dimension, when only two or three dimensions actually matter for fit. The shop has to inspect and prove every feature — CMM time, scrap rate, and setup rigidity all go up.
Cost delta: A titanium medical bracket had 14 dimensions at ±0.01 mm. We requoted with only 3 critical fits at ±0.01 mm and the rest at ISO 2768-f. Price dropped from $412 to $147 per part (qty 25). Same finished geometry, same function.
ISO 2768-mK in the title block. Add tight tolerances only to features that must mate with specific parts or must meet specific assembly loads. Read our full tolerance chart for guidance on which tier to pick.
§04 — Exotic threads, non-standard hole sizes.
Custom-pitch threads, imperial threads on metric parts, odd hole sizes (e.g. Ø9.37 mm when Ø10 would work) all require special tooling or boring. Non-standard threads especially kill cost — every custom tap is a tool order and a setup.
Cost delta: A stainless manifold specified M8×1.0 fine-pitch in a place where M8×1.25 standard would have worked. Setup and tooling added $14/part; part quantity was 200. Total avoidable cost: $2,800.
§05 — Ra 0.4 surface finish on non-functional faces.
A blanket "Ra 0.4 all over" note on the drawing forces finish passes on every face — even the ones nobody ever sees. Each finish pass adds cycle time; a face-milled aluminum part at Ra 0.4 cuts 3–4× slower than the same face at Ra 1.6.
Cost delta: Same 6061 top plate, two versions: Ra 0.4 all over = $58/part. Ra 0.4 only on sealing face, Ra 1.6 elsewhere = $31/part. 1.9× premium for a finish nobody needed.
§06 — Specifying 7075 instead of 6061.
7075 is great if you need the strength. But many designs specify 7075 "because stronger is better" when 6061 would have worked fine. 7075 costs 2.5–3× more in raw material, and isn't as easy to anodize to a uniform color.
Cost delta: Drone gimbal frame in 7075: $86/part. Same frame in 6061-T6: $39/part. Strength margin against worst-case load was 8× in both — i.e. neither material was stress-limited. Pure material overspec.
§07 — Designs that need 4+ setups.
Every setup on a 3-axis machine adds 10–20 minutes of re-fixturing, re-zeroing, and re-probing. A part that needs 5 different orientations may spend more time being repositioned than being cut. 5-axis avoids this — but not every shop has the capability, and 5-axis time is billed 50–80% higher per hour.
Cost delta: Stainless valve body designed with features on 5 faces: $94/part, 3-axis, 5 setups. Redesigned with 4 faces (combined two features that could be on the same plane): $54/part, 3-axis, 3 setups. 1.7× premium for the extra setup.
§08 — Arbitrary chamfer callouts.
"0.5 × 45° chamfer all edges" looks innocent. But on a part with 30 edges, that's 30 individual tool paths — often a separate toolpath for each edge because they're on different faces. And chamfers break into corners in ways that sometimes require a different tool to finish cleanly.
Cost delta: Simple aluminum housing, 24 edges chamfered 0.5×45°: 18 minutes added per part. At qty 100, that's 30 hours of CNC time.
§09 — Tight hole position tolerance from a rough face.
Datum selection matters. If you specify Ø0.05 true position on a hole, but the datum is a rough bandsaw-cut edge, the shop has to first mill that edge true — adding setup time — or measure from an imaginary feature that no tool can find.
Cost delta: Steel linkage plate with ±0.02 mm hole location datum'd to a rough sheared edge: $48/part. Same plate re-datum'd to a machined pad: $22/part.
§10 — "No burrs allowed" without defining "allowed."
Nothing sinks a part to 100% inspection faster than a subjective quality note. "No burrs" gets interpreted conservatively: the inspector will reject anything the fingernail catches. This means hand-deburring every edge, every part, which is expensive and still doesn't guarantee acceptance.
Cost delta: Brass fittings with "no burrs allowed, all edges smooth to touch": $19/part after 12% scrap. Same fittings with "burrs < 0.1 mm per ISO 13715": $11/part, 2% scrap.
The one-page checklist.
Before you release a CNC drawing, walk through these ten questions:
| Check | If yes, fix |
|---|---|
| 1. Internal corners < R1.5 mm? | Open up to ≥ R1.5 unless mating requires sharp |
| 2. Walls < 1.5 mm or pockets > 3× tool dia deep? | Thicken walls, reduce depth, or flag as thin-wall |
| 3. Tolerances tighter than ±0.05 on non-fit dimensions? | Default title block to ISO 2768-mK |
| 4. Non-standard threads or hole sizes? | Use M3–M12 coarse or UNC/UNF; H7 reamed holes |
| 5. Ra < 1.6 specified "all over"? | Specify per face; default Ra 1.6 or 3.2 |
| 6. 7075, titanium, or Inconel called out? | Verify strength margin; 6061/304 may work |
| 7. Part requires > 3 setups? | Consolidate features onto fewer faces |
| 8. Explicit chamfers on every edge? | Replace with "break sharp edges 0.2–0.5" note |
| 9. Datums on raw / unmachined surfaces? | Datum from machined faces only |
| 10. Subjective quality notes ("smooth," "no burrs")? | Reference ISO 13715 with numeric limits |
Work through this list before you send the RFQ. If you're not sure on any of them, send us the drawing — we'll DFM-review it and come back with markups. No obligation, no fee.
Real cost-delta data from recent quotes
Theoretical rules are less persuasive than numbers from actual parts. Below are five anonymized cost comparisons from recent RFQs — same part geometry, same material, same supplier (us), quoted twice: once as designed, once as redesigned per DFM feedback. These are not cherry-picked; they're representative of what typical DFM review yields.
| Part type | Original cost | After DFM fix | Delta | Fix applied |
|---|---|---|---|---|
| Aluminum sensor housing, qty 50 | $62.40 ea | $38.10 ea | -39% | Replaced 3 blind-pocket corners from R0.5 to R2.5 |
| Stainless 316 bracket, qty 200 | $41.20 ea | $27.90 ea | -32% | Dropped ±0.02mm to ±0.1mm on non-mating dimensions |
| Titanium Gr5 part, qty 10 | $485 ea | $310 ea | -36% | Removed surface callout "mirror polish all faces" — 2 faces only |
| Aluminum 7075 bracket, qty 100 | $88 ea | $44 ea | -50% | Switched to 6061-T6 — application didn't need 7075 strength |
| Complex 5-axis part, qty 20 | $275 ea | $165 ea | -40% | Split into 2 pieces machinable on 3-axis + dowel-pin assembly |
Average savings across these five: 39.4%. That's not an outlier — it's what typical DFM review uncovers when someone actually takes a critical look at the drawing. The savings almost always exist because the drafter was optimizing for something other than cost (habit, perceived safety margin, copied from a past part) and didn't stop to ask "what does this actually drive in manufacturing?"
The tolerance-cost relationship is not linear
Most designers know that tighter tolerance costs more. But the cost curve is dramatically steeper than most estimates suggest. Concrete data from our shop on a representative aluminum feature (Ø20mm hole, 40mm depth):
| Tolerance | Process | Relative cost |
|---|---|---|
| ±0.2 mm (IT13) | Drilled only | 1.0× (baseline) |
| ±0.1 mm (IT11) | Drilled + reamed | 1.4× |
| ±0.05 mm (IT9) | Drilled + reamed + measured | 1.8× |
| ±0.025 mm (IT8) | Boring + precision reaming | 3.2× |
| ±0.013 mm (IT7) | Single-point bored + ground | 5.8× |
| ±0.008 mm (IT6) | Jig-bored + ground + CMM inspected | 9.5× |
| ±0.004 mm (IT5) | Precision ground + gauged 100% | 18–25× |
The jump from IT11 to IT7 is a 4× cost multiplier on that single feature. Yet on many drawings, we see ±0.025mm called out on features that functionally need ±0.1mm — because the designer applied a standard block tolerance to every dimension. This is the single most common source of cost inflation in CNC quoting.
The "scope creep" pattern on revisions
Another pattern we see repeatedly: a drawing goes through 4-5 revisions during development, each adding features. The final drawing has:
- Every internal corner with its own radius callout (originally they were all implied general fillets)
- Every edge marked "chamfer 0.5×45°" explicitly (originally "break sharp edges")
- Every hole with individual diameter tolerance (originally a single block tolerance)
- Every datum reference explicitly geometrically defined (originally just implied by function)
Each addition was well-intentioned — usually added during design review or in response to a past failure. But the cumulative effect is a drawing that specifies 40-60 explicit features where 10-15 matter. Every explicit feature on the drawing becomes a thing the manufacturer must measure and certify. On a $50 part, this can easily add $15-25 in inspection time alone.
Before releasing a drawing for production, review every callout and ask: "If I removed this, would the part still function?" Anything that doesn't affect form, fit, or function is pure cost without benefit.
What we won't fix for you (and why)
A DFM review isn't about making every part cheap. Some features are expensive because they matter — and we won't second-guess those:
- Critical fit tolerances on mating bores — if your bearing seat needs ±0.01mm, that's what it needs. We'll confirm the callout is on the right surfaces, but we won't suggest relaxing it.
- Certified material for regulated applications — AMS spec aluminum, aerospace-grade titanium, USP Class VI plastics. The premium is non-negotiable when certification is the whole point.
- Specified surface finishes on sealing surfaces — O-ring grooves, hydraulic cylinder bores, optical mounts. Ra 0.4 μm isn't overkill when a seal leaks below it.
- Complex geometry driving 5-axis requirements — impellers, turbine blades, medical implants. The cost reflects reality; no amount of redesign eliminates the complexity.
What DFM review does eliminate: tolerance, material, and process specs that were added out of habit rather than function. Those are the easy 30-50% savings.
Volume effects — where cost comes from at different quantities
Cost drivers change dramatically with quantity. A design decision that barely affects cost at qty 10 can be catastrophic at qty 10,000, and vice versa:
| Quantity | Dominant cost driver | What to optimize |
|---|---|---|
| 1–10 (prototype) | Setup + programming | Simplify geometry, single-setup when possible |
| 10–100 (pilot) | Cycle time per part | Remove unnecessary features, widen tolerances |
| 100–1,000 (small production) | Cycle time + material | Optimize stock size, reduce rough machining |
| 1,000–10,000 (production) | Material + cycle time | Fixture design, material grade review |
| 10,000+ (high volume) | Raw material cost | Material grade, consider casting/forging alternatives |
At prototype quantities, removing one chamfer saves nothing (setup dominates). At 10,000 units, removing one chamfer saves 2-3 minutes per part times 10,000 parts = 333-500 hours of machining time = $5,000-7,500 in machine cost. The same design change has 1000× different value depending on quantity.
Tell your supplier the target production volume when quoting. It changes which features are worth removing, which materials are worth switching, and which tolerances are worth revisiting. A good supplier will suggest different optimizations at different volumes.
Supplier-side levers you can ask about
Beyond design changes, there are procurement-side conversations that can yield 10-20% savings:
- Annual blanket orders: commit to buying 1,000 units over a year, but take delivery in 100-unit batches. The supplier commits capacity; you get better pricing. Cash-flow friendly because you pay as you receive.
- Stocked inventory programs: for repeating orders, some suppliers will hold raw material (certified to your spec) and finished goods. You pay a small carrying cost in exchange for faster lead time and locked-in pricing.
- Material pooling: if you have several parts using the same material, consolidate quotes so the supplier can buy larger stock lots. Typical savings 5-10% on material.
- Inspection plan negotiation: moving from 100% inspection to AQL-based sampling on mature parts can cut per-part inspection cost 30-50%. Requires production history first.
- Multi-part quotes: quoting 5 parts together instead of 5 separate RFQs often yields 5-8% total discount because setup efficiency compounds.
The takeaway
The opportunities above compound. A part that goes through a thorough DFM review — consolidating features, right-sizing tolerances, reconsidering materials, choosing the right process, optimizing for quantity — typically yields 30-50% cost savings on production. On a $50,000 annual order, that's $15,000-25,000 in savings per year.
None of this comes from negotiating harder with the supplier. It all comes from the engineering stage. The supplier you chose is quoting what you drew. If you drew something expensive, they quote expensive. If you drew something cost-optimized, they quote cost-optimized. The single most leveraged place to reduce manufacturing cost is the drawing, not the negotiation.
This guide reflects how we think about DFM when reviewing customer RFQs. We make suggestions not because we're trying to reduce our own revenue — we want long-term customers who understand what they're paying for, not one-time buyers who feel they overpaid.