Why I stopped asking "which process is best"
I'm a procurement engineer handling prototype and production-part sourcing for a mold-and-die shop in the Midwest. Nine years. I've personally made — and documented — 23 significant sourcing mistakes, and they've cost us somewhere around $11,000. Maybe $9,500, I'd have to pull the spreadsheet. The reason I keep the list is that roughly half of those mistakes were the same mistake wearing a different hat: I picked a manufacturing process before I understood what the part actually had to do.
People ask me "should I use CNC, resin printing, or laser cutting" as if there's a correct answer sitting somewhere. There isn't. It depends on three things, and the annoying part is that the answer flips when any one of them changes. What I can do is tell you how I sort parts into buckets now, and what happened when I sorted them badly.
The three questions that actually classify your part
Before I open a single quote request, I ask:
- Is the part mostly flat? If it's a sheet, plate, or bracket that's essentially 2D with thickness, laser cutting is in play. If it has undercuts, internal channels, or compound curves, it isn't.
- Does it need to be metal, or does it need to behave like metal? Not the same question. Confusing them once cost me about $2,400.
- How many, and how tight? One piece at ±0.1 mm is a different problem than 500 pieces at ±0.05 mm.
That's it. Everything else is detail. Here's what each bucket looks like in practice.
Scenario A: metal parts that have to hold tolerance and repeat
If the part is metal, round or prismatic, and you need the 200th piece to match the first, you're in CNC territory — turning, milling, or both. This is where most of our tooling budget goes, and where I've made the most expensive mistakes.
We run Sandvik Coromant turning inserts on most of our lathes. The reason isn't brand loyalty. It's that when I need to hold ±0.02 mm on a 4140 pre-hard diameter, I need the grade and coating to behave the way the catalogue says. Sandvik Coromant products — inserts, holders, and the tooling systems around them — are designed to work as a system, and that matters more than most people expect.
Which brings me to my worst assumption. In 2021 I assumed "equivalent ISO grade" meant interchangeable. Didn't verify. Turned out the cheaper inserts were the right shape and the wrong substrate — same ISO code, different cobalt content, different coating thickness. We ran a batch of 180 bushings. About 40 came out with chatter marks that passed a visual check and failed the customer's CMM. Roughly $1,800 in scrap plus a week of schedule. I learned never to assume the designation on a box tells you what's inside it. To be fair, the savings on those inserts were real — about 30% per piece — but one scrapped batch ate the entire annual delta in an afternoon.
Now we do a first-article cut on every new insert lot before it goes into production. Not a formal process — just a test cut on scrap stock. That's a process gap I should have closed after the first bad batch, not the third.
One more thing, and this took me years to accept: Sandvik Coromant is a cutting tool company — not a machine shop, and not an additive house. Twice now, their application engineer has told me "this geometry is going to fight you in turning; have you looked at EDM or a printed insert?" They didn't lose the sale. They kept my trust for the next three years of insert orders. From my perspective, a supplier who names their boundary is worth more than one who claims to do everything.
Scenario B: complex geometry, small volume, plastic or pattern
If the part has internal channels, lattice structure, or a shape that would take five setups on a mill — and you need one to fifty of them — resin printing is usually the answer. For small work I use an Elegoo Mars 5 resin 3D printer: 7-inch 9K mono LCD, layer heights down to 0.01 mm, and cheap enough that we bought two.
But the printer is the least important decision. The best industrial 3D printing materials for your part are a separate conversation from "which resin is on sale this week." I learned that in early 2023: I assumed any resin labeled "tough" would hold a threaded insert under load. Didn't check the creep behavior. The part passed fit check, went into a fixture, and sagged 0.6 mm over a weekend at 40°C in the shop. Nobody was hurt. The fixture was scrapped and we lost a day.
Put another way: a resin print is a plastic part, and plastic has a memory. If the part sees sustained load or heat, read the datasheet for the specific resin — heat deflection temperature, flexural modulus, creep — not the marketing name on the bottle. For patterns and jigs, standard resin is fine. For anything functional, pay for the engineering resin and print a coupon before you commit.
Scenario C: flat parts, fast turnaround, sheet stock
Laser cutting is the process people most often misapply in both directions — they use it for things it's great at, and they assume it's cheap for things it isn't.
So, how does a laser cutter work step by step? Roughly: you send a DXF, the shop nests your parts on a sheet, a focused beam burns or melts along the toolpath, and the head moves at a controlled speed while assist gas blows molten material out of the kerf. The kerf — the width of material the beam removes — typically runs 0.1–0.3 mm depending on material and thickness, and good shops compensate for it in nesting. The edge you get has a small heat-affected zone, which matters on thin aluminum and on any material that hardens at the cut line.
Where I went wrong: in 2022 I assumed laser cutting was always cheaper than machining a bracket. It was — for the flat plate. But the bracket needed three bends and two tapped holes, and by the time we paid for forming, fixturing, and a second op, we were within 15% of just milling it. The laser quote was accurate and still led me to the wrong decision, because I compared the wrong scopes. The lesson: compare finished-part cost, not process cost.
The case that breaks the pattern
The counterintuitive one, and the reason I wrote this down: for a recent one-off bracket that needed to look like machined aluminum, we resin-printed it, sanded it, and painted it. Total cost about $70. The CNC quote was $410 with a two-week lead time. It wasn't a functional part — it was a fit-check mockup for a customer review. If your part doesn't need to bear load, conduct current, or survive heat, the "right" process is the cheapest one that produces the geometry — and that's often not the process the material suggests.
How to tell which scenario you're in
Run these in order:
- Does the part carry load, heat, or electrical current? If yes — metal. Go to Scenario A.
- Is it flat stock with few or no bends? Go to C, but quote the finished part including forming and secondary ops.
- Is the geometry 3D, quantity under ~50, and no structural requirement? Go to B, and specify the resin by datasheet number rather than marketing name.
- Does it need to do two of those things at once? Split it. I've saved more money making a part in two processes and assembling it than by forcing one process to do everything.
This was accurate as of early 2025. Insert grades, resin formulations, and laser service pricing all move, so verify current specs and rates before you budget.
Last thing. The supplier who told me "this isn't our strength — here's who does it better" has gotten every turning insert order we've placed since. If you're evaluating vendors, that sentence is worth more than any capability brochure. In my opinion, it's the single best signal you'll get.