3D Printing, CNC Machining, or Injection Molding? Oshcut's Quality Inspector Explains How to Decide

There's no universal answer to "what's the best manufacturing process?" The right choice depends on what you need the part to do, how many you need, and what you're willing to tolerate on specs. It's a decision I watch people get wrong on a regular basis—and I say that as someone who has rejected more than my share of first deliveries.

I'm the quality manager at Oshcut, a custom manufacturing company. Roughly 200+ unique items cross my desk every year, and every one of them gets checked against specifications before it ships. I've sent back more than a few with "requires rework" scribbled on the paperwork. Here's how I'd guide you through process selection, based on that experience.

Classify your situation before you call anyone

Whatever you're trying to make, you're in one of three camps. Figuring out which one takes about five minutes.

• Camp 1 – Prototype. You need a handful of parts to test form, fit, or just to see the thing in real life.
• Camp 2 – Low-volume production. You need 20, 100, or maybe 500 parts that go into something you're selling.
• Camp 3 – High-volume production. You're thinking in thousands, the part is plastic, and the cost per unit is the main driver.

Most buyers start by asking for a price per part. The better question is total landed cost—which includes setup fees, finishing options, and shipping. I've seen quotes where the per-part price looked fantastic and the setup fee plus shipping turned the total into something hideous. Do the math before you fall in love with a number.

Camp 1: Prototypes — go 3D printing, and ask about cold pulls

For a visual model or a form-fit check, 3D printing is usually the quickest route. FDM is fine for rough geometry; SLA gives you smooth surfaces and finer details. But not every printed part is the same quality, and the difference often comes down to something buyers don't think about: nozzle maintenance.

What is cold pull in 3D printing?

A cold pull is a cleaning procedure for a 3D printer nozzle. You heat the nozzle, insert a piece of cleaning filament, let the heat soften the deposits, then drop the temperature slightly and pull the filament out. The pull action drags old filament residue and charred debris out of the nozzle with it. It's simple, it's fast, and it prevents the artifacts I see on low-quality prints: inconsistent layers, stringing, random under-extrusion.

I rejected a batch in 2024 because the parts looked fine on the first layer but every surface above it had streaks. The shop had switched from black to white filament without a cold pull between colors. The contamination was obvious under a magnifier. (Ever wonder why "white" prints come out grayish? Now you know.)

So when you order printed parts, ask the shop how they manage nozzle hygiene. A fair question—and the answer separates operators who treat 3D printing as a discipline from those who treat it as a toy.

When a 3D print is the wrong tool

If your prototype is round, threaded, or has to survive actual physical handling, a printed part may not tell you anything useful. Go straight to machining instead. An industrial CNC lathe machine will turn out a metal part with real strength and realistic tolerances—which matters for bushings, adapters, or any cylindrical component where layer lines don't cut it.

For prototype quantities, a CNC lathe might cost a little more than a print, but the information you get from a machined part is different in kind: the feel, the weight, the way threads engage. All of it matters before you commit to production.

Camp 2: Low-volume production — CNC machining is the sweet spot

Need functional metal parts in quantities of 10 to 1,000? CNC machining is where I'd steer you. Materials are real: aluminum, stainless, steel, brass. Tolerances are dependable. Surface finish is predictable. And unlike 3D printing, the part behaves like the material it's made of, not like a plastic approximation of it.

The one piece of advice I'd hammer home: your tolerance spec is probably too tight. Most engineers add ±0.001" to every feature out of habit, without asking what actually matters. A lot of features can operate at ±0.005" or ±0.010" without any functional problem at all.

Here's a concrete example. In Q1 2024, we ran a small production batch of mounting brackets where a critical bore came out dead on nominal—±0.0005", beautiful work. When I sat down with the customer afterward, we agreed that loosening that bore to ±0.005" would have saved about 15% on unit cost. Nothing in the application needed the tighter spec. We revised the drawing, and the next order cost meaningfully less.

Before you place a machining order, look at your drawing and ask: if this dimension drifts by a few thousandths, what actually happens? If the answer is nothing, loosen it. Save your budget for the dimensions that genuinely matter. (Note to self: I really should put this prompt on the quote form.)

Camp 3: High-volume plastic parts — injection molding, with eyes open

Production runs in the thousands? Injection molding wins on cost per part once the tooling is paid. But here's the frustration I see constantly: many molders impose minimum order quantities that are wildly above what a small business actually needs. Five to ten thousand parts when you want 500? That's not a challenge; that's a "go away" signal.

To be fair, minimum order quantities aren't always about gatekeeping—tooling amortization is real math. But that's an explanation, not an excuse. Small doesn't mean unimportant; it means potential. When I was starting out, the suppliers who took my $200 orders seriously are the ones I still send $20,000 orders to today. If a molder is too big for your order, let them be too big. Move on to someone whose model doesn't crush you with minimums.

ERP for plastic injection molding: a checklist item

If you're scoping a molding project, ask whether the shop runs ERP. An ERP for plastic injection molding tracks raw material lots, machine settings, and inspection results through every run. When something fails in the field, a shop with solid ERP identifies the exact suspect parts. A shop without it shrugs.

Cheaper shops often skip ERP until they get burned. I've audited both types, and the ERP users aren't usually the cheapest option—which, honestly, is fine when you're making hundreds or thousands of parts that your customers will rely on.

And one more thing if you'll be marketing these parts: watch your environmental claims. Per the FTC Green Guides (16 CFR Part 260), calling a product "recyclable" requires that a substantial majority of consumers have access to facilities that actually recycle it. It's easy to assume materials are recyclable in practice. Check before you print "recyclable" on a label.

So which camp are you in?

Here's the quick test I give clients when they're torn:

1. Need 1–10 parts for concept validation, fit, or a presentation → 3D printing.
2. Need 10–1,000 functional parts in metal → CNC machining.
3. Need 1,000+ plastic parts with low per-unit cost → injection molding.

If you're genuinely between categories, talk to someone who offers more than one of these processes. That's actually an advantage at Oshcut: we're not pushing a single technology. We'll tell you when a part doesn't need five-axis machining and can be 3D printed without losing sleep.

Contacting Oshcut—and what to send with your RFQ

If you're ready for a quote, the Oshcut contact route is simple: use the RFQ form on the website or email us directly. What you should have ready:

• A CAD file or drawing with dimensions and tolerances marked
• Material requirements—or a note that says "you tell me what's best"
• Quantity and a realistic target timeline
• Context: what the part does, and how it's used

That last point matters more than people think. In November 2024, a customer sent a drawing for a bracket that was, by every appearance, meant to hold electronics. Because the drawing said nothing about vibration, we suggested a material change costing thirty cents more per part. They came back later to say their first prototype had cracked in service. The fix cost less than a USPS stamp—which, by the way, is $0.73 for a First-Class letter as of January 2025 (usps.com/stamps), in case you're mailing a sample to a colleague.

The last variable is trust

I've been doing this for four years, and I'll be honest: I've made the wrong call under time pressure. In one instance, I had two hours to decide whether to approve a final production run. The upside of saying yes was hitting the launch date. The downside was a $3,500 redo if the parts failed. Normally I'd run a full verification protocol—I implemented one back in 2022—but there was no time. I approved on the strength of a quick visual check and a good relationship with the vendor. It worked out, but only because their quality system was solid.

The lesson hasn't changed: find a partner whose standards hold your parts—and your project—to a level you can trust. Whether you order one prototype or a thousand parts, the quality process matters equally. And a supplier that respects small orders will almost always respect big ones. Those are the shops worth building with.