Custom Part Ordering: A 3D Systems Engineer’s 7-Point Mistake-Prevention Checklist

Why this checklist exists

I’m a manufacturing engineer on the 3D Systems on-demand production team. For nine years, I’ve handled custom part orders—CNC machining, metal and polymer 3D printing, and the occasional short injection-molding run—for customers in aerospace, medical, and industrial equipment.

I’ve also personally made and documented 23 significant mistakes. Scrap, rework, and expedite fees add up to roughly $38,000. Some of that was my own bad judgment. A lot of it came from approving a customer’s file instead of asking what the file didn’t say.

This is the checklist I run before an order gets released. Not “get three quotes”—anyone can write that. These are the seven checks I wish someone had made me do before my first expensive lesson.

1. Treat the tooling note as a design detail, not an afterthought

The note in the CAD file said “end mill ball nose for cast iron.” The job was forty cast-iron housings with a curved face and a surface-finish callout that left no room for interpretation. That single line was supposed to carry the whole machining strategy.

I approved it without asking the obvious questions. Which coating? Climb or conventional? What finish pass? How many flutes? Cast iron is abrasive and unforgiving. The uncoated end mill broke down somewhere between the roughing pass and the finish pass, and the first article came off the machine with a smeared, burnished surface. Looked fine from three feet. Under a loupe, it looked terrible.

We scrapped the first article, called the customer, and started over with a coated carbide ball nose end mill at the right chip load. Damage: roughly $1,800 and six days. All because “end mill ball nose for cast iron” should have been the start of a conversation, not the end of one.

If you see a tooling note that specific, expect a follow-up question from the shop. That question is not incompetence—it’s how good parts happen.

2. “Do 3D printers mix colors?” is the wrong question

Every few weeks a customer asks: do 3D printers mix colors? It’s a fair question, but the mental model is off. A 3D printer isn’t an inkjet laying down cyan, magenta, and yellow on paper.

Production 3D printers do not mix colors that way. FDM machines can pause and swap filament. Resin machines print one color unless you purge and refill. Powder-bed machines are single-material by design. Some multi-material systems can place different materials voxel by voxel, and full-color binder jetting exists—but those are for visual prototypes and concept models more often than engineering parts. None of them mix a gradient like fountain-pen ink.

I learned this the hard way. I told a customer we could handle their multi-color request on an engineering material. What I should have said: we could make a multi-color prototype in a different material, or a single-color part in the material they actually needed, but not both. They needed both. We ended up printing twice.

When color comes up, ask what the part has to do, then work backward. If a functional part needs brand color, print it in a base color and paint it. Supply vector artwork or a 300 DPI raster at final size for the decoration step. If the color has to match a brand standard, the commonly used tolerance is Delta E below 2 for brand-critical colors, according to Pantone’s color-matching guidance. And always approve a painted sample on the actual material before production—screens lie.

3. The “best 3D printer 2025” lists are not production plans

January rolls around and someone sends a link titled “best 3D printer 2025.” I work at 3D Systems, so I’m not going to pretend we don’t make excellent machines. But I’ve also watched good machines sit underutilized because they were chosen from a ranking, not from a production plan.

If you search for “3D Systems best 3D printer 2025,” you’re probably deciding whether to buy our equipment. Just don’t decide based on a list that does not know your part. At an earlier job, a machine shop bought the mill that topped a similar list. The mill was fine. The plan was missing workholding, training, tooling, and maintenance. Utilization stayed near 45 percent for a year. The machine wasn’t wrong—the process around it was.

So read the best-3D-printer-2025 articles if you want a survey. Skip the rankings and steal their evaluation criteria: What materials are qualified? What post-processing does the process need—powder handling, supports, heat treatment? How long until you actually produce certified parts? Can you get service in your region? If those questions get complicated, an on-demand production partner like 3D Systems may make more sense than a machine purchase. The best machine is the one your supply chain can support seven days a week, not the one with the prettiest demo part.

4. Design for the process after the build plate

A 3D printer or CNC machine produces a part. That part usually goes somewhere else afterward: into an assembly, an anodizing rack, or a weld fixture.

The welding lesson hit in late 2023. A customer ordered about sixty printed stainless-steel fittings that they planned to join with a 1000W laser welding machine. The printed parts were dimensionally perfect. The welds were not—porosity everywhere.

It wasn’t the printer. It wasn’t their welding equipment. The problem was a small closed pocket at the root of the weld joint, where expanding gas had nowhere to escape. If the part had been machined from solid, the pocket wouldn’t have existed. But we were printing the part, and we could have designed a vent into the model. We just hadn’t thought about the welding step when we reviewed the design.

Now the checklist asks: what happens after this part is made?

• If it’s welded, specify weld prep such as a chamfer or root gap, and make sure there are no sealed cavities behind the joint.
• If it’s painted or anodized, include holes for racking and drainage.
• If it’s press-fit, the tolerance has to be on the drawing, not in someone’s head.

Design for the last process, not the first one.

5. Verify the supplier, not just the price

A customer recently asked us to re-run a functional part that had failed in their testing. They’d bought “genuine” parts from a third-party storefront that used the 3D Systems logo in its header. The parts looked right, but they had no material certification and no process traceability. When the part failed, no one answered the phone.

Check the source: look at the official 3D Systems logo on the actual corporate site, but more importantly, confirm the vendor is an authorized partner. Call the main line if a quote looks too good. Real suppliers do not vanish when a quality question arrives.

The re-run cost more than the original deal saved, plus a week of schedule. Thirty seconds of verification would have prevented the whole detour.

6. Tolerance is a number, not a hope

One of the most expensive phrases in manufacturing is “use standard tolerances.” Which standard? Shop-standard might be ±0.1 mm. That same shop-standard might be useless for a press-fit hole.

I approved an order once where the drawing called out a critical bore and then said “standard tolerance.” The part was made to our usual tolerance. The customer pressed in a bushing that was supposed to be an interference fit; it spun. The drawing lacked the number, and I failed to ask. Both sides lost.

Write the tolerance block into the model or drawing for every critical feature. When the number is missing, ask before production, not after inspection. And schedule a first-article inspection—a cheap step that catches most of the expensive mistakes on this list.

7. Sanity-check the file: units, format, revision

STL files are unitless. If the CAD model was drawn in inches and the shop’s software assumes millimeters, the printed part will be 25.4 times smaller than intended. The opposite direction gives you a part 25.4 times too large. People know this and still don’t check the bounding box.

STEP files preserve units better and are usually the right choice for machined parts. STL is fine for 3D printing, as long as the export resolution is fine enough. Either way, open the part in the quoting or slicing software and confirm the overall bounding box before releasing. And for the love of your supplier’s sanity, name the file with the revision and units: bracket_revC_mm.step, not final_v2_new.stl.

Mistakes that still catch me

Even with the checklist, the failure mode changes rather than disappears. Three things still cause trouble:

• Paying more does not buy quality. Quality comes from clear specifications plus qualified review. Price only buys attention, and even that has limits.
• Choosing the process before defining the part. “We need 3D printing” is sometimes really “we need 100 units; which method is best?” The need should come first.
• Skipping the DFM conversation because it feels slow. A five-minute call about a draft design can prevent a five-week correction later.

There is something satisfying about watching a first article pass inspection on the first run. After nine years and twenty-three documented mistakes, that moment hasn’t gotten old. It usually happens when someone ran this list before releasing the order, not after the parts landed on a bench.

The point isn’t to memorize rules. It’s to get comfortable asking the unglamorous questions. If you’re ordering custom parts, run these seven checks. You’ll still pay for your own mistakes eventually, but these seven won’t be among them.