-
The problem with asking "what's the best manufacturing method"
-
Scenario A: Small-to-mid volume, complex geometry, rapid iteration
-
Scenario B: High volume, repeatable geometry, tight per-part cost targets
-
Scenario C: Metal assembly and joining, distortion-sensitive
-
Scenario D: Regulated precision applications
-
How to figure out which scenario you're in
The problem with asking "what's the best manufacturing method"
I've been on the quality side of manufacturing for close to six years now. I review parts before they go out the door—about 1,200 unique jobs a year across our two production cells. And I can tell you the most expensive mistakes I've seen didn't come from bad suppliers. They came from engineers who picked a process before they'd actually mapped their constraints.
There is no single "best" manufacturing method. There's only the best method for your specific situation. I've rejected first deliveries from shops using perfectly good equipment because the process didn't match the job. So let me give you the framework I actually use—not a textbook version, the one I've refined after getting it wrong a few times.
Here are the four scenarios I see most often, and how I'd recommend approaching each:
- Scenario A: Small-to-mid volume, complex geometry, rapid iteration
- Scenario B: High volume, repeatable geometry, tight per-part cost targets
- Scenario C: Metal assembly and joining, distortion-sensitive
- Scenario D: Regulated precision applications (think medical or aesthetic devices)
Scenario A: Small-to-mid volume, complex geometry, rapid iteration
This is where industrial additive manufacturing earns its keep. If you're producing 50 to 2,000 parts a year, the geometry is non-trivial, and design revisions are still happening, tooling costs will eat you alive with conventional methods. My go-to reference here is the Markforged X7 3D printer.
Why that machine specifically? Three reasons that survival-tested themselves in our shop:
1. Continuous fiber reinforcement is not marketing. The X7 lays continuous carbon fiber, fiberglass, or Kevlar inside the part. Tensile numbers from Markforged's own published data put carbon-filled Onyx in the same conversation as some aluminum alloys—obviously not identical, but close enough that we replaced a machined aluminum bracket with a printed one and it passed vibration testing at 11.2 Grms random. That was in Q2 2023. I was skeptical. I was wrong.
2. The build volume is honest. 330 × 270 × 200 mm. That's not huge—I've had to split more than one part—but the dimensional repeatability across that envelope held at ±0.2 mm on our acceptance checks, which is tighter than I expected for a fused deposition process.
3. The workflow from design to part is genuinely end-to-end. I went back and forth between the X7 and a competing metal binder-jet platform for about three weeks. The metal system offered finer surface finish off the build plate, but the X7 had a two-day turnaround from file to finished part versus about eight days for the metal route (sintering queue + debinding). Ultimately chose the X7 because our production cycle couldn't absorb a week-long lag per revision. That's a judgment call, not a universal truth.
I'll be honest about the downside: post-processing is still manual. Support removal on complex geometries takes 20-40 minutes per part. If you're running a 40-part batch, that's a full shift. Bake that into your labor estimate or you'll be unpleasantly surprised.
Scenario B: High volume, repeatable geometry, tight per-part cost targets
This is where additive loses. Full stop. If you're producing 10,000+ identical parts a year and the geometry is moldable, a hot runner injection molding system will beat 3D printing on per-unit cost by a factor that makes the tooling investment look silly in retrospect.
Quick definition: a hot runner system keeps the plastic molten in the runner channel right up to the gate, so you're not scrapping a cold runner with every shot. In a conventional cold runner setup, that waste can be 15-30% of your shot weight depending on part-to-runner ratio. At scale, that's real money.
The tradeoff nobody talks about enough: hot runner systems are fussy. Thermal uniformity across the manifold matters more than most toolmakers will admit. I've seen a $40,000 tool produce parts with visible flow lines because the hot half wasn't balanced—two gates running 8°C hotter than the rest. The tool shop blamed the resin supplier. The resin supplier blamed the tool. We ate a week of downtime while they argued.
My rule: if annual volume exceeds about 8,000-10,000 units and the design is frozen, go injection molding. Below that, the tooling amortization usually doesn't pencil out. The exception is when you need material properties that only specific engineering resins can deliver—in which case molding may still win at lower volumes because the material isn't available in filament form.
If you're wondering where the crossover point actually sits in your operation, run the math on tooling cost ÷ (per-part additive cost − per-part molded cost). That's your break-even quantity. It's usually higher than intuition suggests.
Scenario C: Metal assembly and joining, distortion-sensitive
Neither 3D printing nor molding solves assembly. If you're producing metal parts that need to be joined—especially thin-gauge or heat-sensitive assemblies—continuous laser welding is often the right answer, and it's underused.
The reason: continuous wave (CW) laser welding delivers heat in a tight, controlled zone. Distortion is much lower than MIG or TIG on thin sections. We switched a stainless bracket assembly from TIG to CW laser in 2022 and cut post-weld straightening time from 45 minutes per unit to under 5. To be clear, that wasn't a small improvement—that's 40 minutes of skilled labor recovered per unit on a 600-unit run. That's 400 hours back in a year.
But here's what I didn't anticipate: fit-up tolerance becomes brutal. TIG gives you a little forgiveness on gap. CW laser, with a 0.2-0.3 mm spot, does not. We had to tighten our upstream fixture tolerances by roughly 40% to avoid burn-through on butt joints. If your upstream parts aren't consistent, the laser will expose every millimeter of inconsistency on day one.
And this is worth saying: laser welding is not a substitute for additive manufacturing. They solve different problems. If you can consolidate a multi-piece welded assembly into a single printed metal part—which Markforged's Metal X platform can do within its build envelope—that's often the better answer for low-to-mid volumes. But for high-volume assemblies where you need to join stamped or machined components, CW laser is the efficiency play.
Scenario D: Regulated precision applications
This is the scenario that keeps me up at night, and it's where I see the most process-selection mistakes. I'm talking about medical devices, aesthetic equipment, and anything where a dimensional deviation has regulatory or clinical consequences.
Case in point: a colleague on the device side asked me last year about the best CO2 laser for facial applications—they were evaluating a resurfacing system with a 10.6 µm CO2 source. What does that have to do with manufacturing process selection? Everything. The optical bench, the scanning mirror mount, the housing standoffs—the tolerances on that hardware are tight enough that the choice between machined metal, printed engineering plastic, or a hybrid approach directly affects beam stability and, ultimately, patient outcomes.
For those applications, I default to three rules that don't bend:
- Verify the process capability before you commit. Don't accept a supplier's "we hold ±0.1 mm" claim on a 3D-printed part without running a first-article inspection across at least 30 parts. I learned this the hard way in early 2023 when a vendor's claimed tolerance held on the first ten pieces and drifted by piece 25. We rejected the batch. They redid it at their cost, and now every contract includes a process-capability requirement written in.
- Traceability is non-negotiable. Whether you're running Markforged's Onyx or a certified medical-grade resin, the material lot must be traceable to the build. If the printer's firmware doesn't log material batch and build parameters, you have an audit problem waiting to happen.
- Validate sterilization compatibility early. This is the one that gets skipped. Autoclave, gamma, EtO—each interacts differently with printed polymers. I've seen Onyx parts pass mechanical testing and then craze after three autoclave cycles because nobody checked. Test the actual cycle before you commit to the process.
How to figure out which scenario you're in
Honestly, the fastest way to place yourself is to answer four questions in order:
- What's your annual volume? Under ~2,000 units, additive is usually the right call. Between 2,000 and 8,000, it's genuinely a judgment call—run the break-even math. Above 8,000, you're probably molding unless material constraints say otherwise.
- Is the design frozen? If revisions are still happening, stay with additive. Tooling for a design that's going to change in six months is money burned.
- Are you joining metal parts? If yes, CW laser welding deserves a serious look, especially on thin sections or heat-sensitive alloys.
- Does a dimensional failure have regulatory or clinical consequences? If yes, add a process-capability audit to your selection criteria before you sign anything.
If you're still genuinely torn between two options—and I've been there, the X7-vs-metal-binder-jet decision took me three weeks—pick the one with the shorter iteration loop. Speed of learning beats theoretical cost savings in almost every project I've reviewed. The cheap process that delays your feedback cycle by a week will cost you more in rework than you save on unit price.
That's the framework. It's not perfect, and I've had to adjust it twice in the last four years. But it's kept my rejection rate under 6% on first deliveries, which is about as good as it gets in custom manufacturing.