Used 3D Printer, Ultimaker S3, or Injection Molding? A Quality Manager’s Guide

I'm the person who signs off on first articles before they go to customers. In 2024, I reviewed roughly 250 unique parts and rejected about 9% of first deliveries because the supplier's “standard” tolerance didn't match the spec on our drawing. So when I talk about buying 3D printers, I'm not going to pretend there's one machine that's right for everyone.

The real question isn't “Should I buy an Ultimaker S3?” or “Should I buy a used 3D printer?” It's “What happens if the part is late, out of spec, or impossible to trace?” Once you answer that, the process choice gets easier.

No universal answer: four manufacturing scenarios

In my role, production decisions tend to fall into four scenarios:

• Short-run functional parts, with a real deadline: an industrial FDM system like the Ultimaker S3 plus Digital Factory is often the right mix of control and speed.
• High volume or qualified polymer material: injection-molded polyamide 6 is still the most deterministic route for many PA6 parts.
• Metal features and tight edges: CNC machining with a Seco indexable end mill is easier to validate than an extruder-based process.
• Non-critical models and experiments: a used 3D printer can be a smart purchase, as long as you keep it out of customer-critical work.

Scenario A: Short-run functional parts and a deadline — Ultimaker S3 plus Digital Factory

When someone asks me for Ultimaker S3 3D printer specifications, I usually tell them to look at three things: build envelope, material profile, and workflow history. The Ultimaker S3 has a 230 x 190 x 200 mm build volume, dual extrusion, and swappable print cores. Those numbers matter, but not as much as whether the printer can produce the same result on Tuesday as it did on Monday.

This is where Digital Factory - Ultimaker’s fleet management and job logging platform - earns its place. Cura gives me profile control, and Digital Factory gives me the audit trail. If a customer asks “can you prove this was printed in the approved material?” I can pull up the job and show them the file, profile, printer, and timestamp. That traceability is a spec, not a bonus.

I'll give you a real timeline. In March 2024, we paid $400 extra for a guaranteed delivery slot at a machine shop. The alternative was missing a $15,000 customer demo. The $400 didn't buy speed. It bought a fixed date. If you have a deadline that cannot move, pay for the process you can verify. A new Ultimaker S3 with a maintenance contract costs more than an old open-frame printer, but the cost is predictable. An unplanned failure is not.

Scenario B: High volume and regulated material — injection-molded polyamide 6

Here is where I sound anti-additive. If you need thousands of identical parts from glass-filled polyamide 6, I would probably choose injection molding over FDM. The injection molding polyamide 6 market exists because PA6 has decades of published material data. Producers know its shrinkage, moldflow behavior, and mechanical test history. I can qualify a mold and lock a process window.

I don't have current market numbers memorized, and I won't pretend otherwise. But the engineering point is simpler: once the mold is approved, the injection molding process repeats with less operator-dependent variation than a desktop FDM batch. That consistency is what quality people mean by capability. A printer can make a PA6 part. A well-run mold can make 5,000 PA6 parts without silently changing layer adhesion.

Scenario C: Metal features and tight edges — CNC with a Seco indexable end mill

If the critical feature is a 90-degree shoulder in aluminum, or a slot width that has to hold across 1,000 parts, I don't argue with the production engineer who prefers CNC. We keep a Seco indexable end mill in the tool cabinet for exactly that work. The insert is a replaceable, certified cutting edge; when it wears, I replace it without resetting the program.

This doesn't make additive bad. It makes additive wrong for that scenario. The first question should be: where is the tolerance? If the answer is metal, CNC with an indexable end mill gives you a more deterministic toolpath than an extruder. I'd rather explain to an engineer why we didn't use a 3D printer than explain to a customer why their aluminum bracket failed.

Scenario D: Non-critical models and experiments — a used 3D printer can make sense

I don't want the earlier sections to sound like I'm against buying used. For concept models, tooling that never touches a product, or experiments in a lab, the economics are different. If someone asks me where to buy used 3D printers, I tell them to start with certified refurbishers and surplus dealers who allow a test print. Then check four things: print head hours, heated bed condition, firmware and calibration history, and documented service records.

A missing service record is a red flag. I once skipped that check because a printer looked clean and I thought “what are the odds?” The first real job shifted 0.4 mm. I know the odds better now. Treat that machine like a laboratory instrument, not production equipment. Keep it away from customer jobs until you have validated it against a test part with critical dimensions.

How to tell which scenario you're in

Take out the drawing and ask four questions in order:

1. What material is required? If it says PA6-GF or another qualified polymer, injection molding or CNC will be easier to validate than FDM.
2. How many parts do you need in the next 30 days? Under a hundred? FDM is practical. Thousands? Call a molder.
3. Which dimension makes you say no? For complex internal channels or low-volume stress tests, industrial FDM can work. For a tight metal pocket, use CNC.
4. What happens if delivery slips by one week? If the answer is a lost customer, buy certainty: a new S3 with service and Digital Factory, or a machine shop with a confirmed slot. A “pretty good deal” on used equipment is not a delivery plan.

If you have time to experiment, used can be the right way to learn. But for a part with a deadline, a known process with a log and support wins every time. The cost difference between “probably” and “verified” is rarely visible in the purchase price. It shows up later, usually as a rework order or an apology.

If you ask me, the best process is the one whose next part looks like its last part. That's why I'm okay with a range of answers: Ultimaker S3 for functional short runs, injection-molded polyamide 6 for volume, Seco end mills for metal, and used machines for experimentation. The key is to stop asking which printer is better and start asking which process you can stand behind when the deadline is real.