I've Been Ordering Parts on Xometry for 4 Years — Here's What I've Learned About Tolerances, Materials, and When to Go Off-Platform

There's no single 'right' way to use Xometry (and anyone who tells you otherwise hasn't made enough mistakes)

I've been handling custom part orders for about four years now. Not as a purchasing agent for a Fortune 500 — I'm part of a small R&D team, and we prototype and produce short-run parts constantly. We use Xometry a lot. Probably 60-70% of our outsourced CNC and 3D printed parts go through them.

But I've also made some expensive, embarrassing mistakes. I'm not a manufacturing engineer. I learned by doing, and by failing. So I'm going to walk you through the three main scenarios I see engineers and buyers face when they're deciding how to use Xometry — and what I've learned the hard way about each one.

I'm not a metallurgist or a materials scientist. So I can't speak to the ultimate grain structure of a 5-axis CNC'd aluminum part. What I can tell you from a procurement and practical engineering perspective is how to avoid the pitfalls I fell into.

Let's get into it.

Scenario A: You're price sensitive & need decent parts in moderate quantities

This is probably the most common scenario. You need, say, 50 to 500 pieces. They're not for a fighter jet. They're for a production run of a consumer product, an enclosure, a bracket. You want them to be good — but you're not holding every dimension to a micron.

What I've learned: Don't over-spec your tolerances

This is mistake #1 I made. My first big order on Xometry was for 200 machined aluminum brackets. I specified every hole at +/- 0.005 inches. Looked good on my screen. But the quote tripled.

I assumed 'tighter is better'. Turned out I was paying for precision I didn't need. Standard Xometry tolerances for CNC machining are typically around +/- 0.005 inches (or 0.13 mm) for metal, and they handle +/- 0.010 inches for plastics easily. For 90% of what we do, that's perfectly fine.

Reference: Xometry's published standard tolerances align with general ISO 2768-m for machined parts, though they offer tighter options for a premium.

The rule I follow now: Only tighten tolerances on mating surfaces or critical features. Let the rest run at standard. You'll save money and not see any functional difference.

Material selection for this scenario

For most general parts, 6061 aluminum is the sweet spot. It's cheap, machines well, and is strong enough for most applications. For plastics, I've had good luck with ABS and Delrin (acetal) for structural parts, and we use a lot of nylon 12 for 3D printing on our own FDM printers — but that's another story.

Scenario B: You need precision parts with very tight tolerances

Maybe you're working on a precision assembly, a medical device component, or something that interfaces with existing, tightly-toleranced hardware. You need +/- 0.002 inches or better.

What I've learned: Xometry can handle it, but you need to be specific

I once ordered a series of alignment pins for a jig. I assumed Xometry's default tolerances would be fine. They weren't. The first batch had variation of up to 0.008 inches on critical diameters. They were useless.

This is where the instant quoting engine is both a blessing and a curse. It gives you a price. But it can't read your mind about which dimensions are critical. You need to annotate your drawings. Use callouts. Mark critical dimensions with explicit tolerances.

Total cost of my original oversight: $800 wasted on bad parts, plus a 1-week delay. That's when I learned: For tight-tolerance work, you have to use Xometry's 'CNC Machining with Critical Dimensions' option or similar, and you must communicate clearly on the drawing.

Also, for really tight stuff (like bearing fits), I've found it's worth getting a manual quote or talking to their support. The automated system is great, but for complex or very tight parts, a human conversation is unbeatable.

Scenario C: You're prototyping or need something very fast

You have an idea. You need it in your hands in 3 days. You're not sure about material, and you might change the design tomorrow.

What I've learned: Use their on-demand manufacturing and wide material selection for speed, but don't expect perfection

For prototyping, Xometry is hard to beat. Their platform and network of partners mean they can usually turn parts around in a few days. We use their 3D printing services constantly — SLA for form/fit, SLS for functional—and their CNC for aluminum prototypes.

But here's the nuance: 'Fast' sometimes means 'fast because they have capacity'. It doesn't always mean 'fast because it's optimized'. I've had prototypes come back with tooling marks that were fine for testing, but wouldn't fly in a final product. That's acceptable for prototyping, but you need to know that upfront.

For this scenario, I've also bought a couple of FDM printers for in-house use. We have a Bambu Lab X1 and an older Prusa. For quick iterations, it's cheaper and faster to print in-house. For anything with tight tolerances, metal, or advanced materials like carbon-fiber-filled nylon, we send it to Xometry or another service.

How to tell which scenario you're in (the judgment guide)

Here's a simple mental checklist I use before every order:

1. How many parts do I need? Under 10? Probably prototype (Scenario C). 50-500? Scenario A.
2. What's the most critical dimension? If it's a hole that a shaft goes through, that's critical. If it's an overall envelope, it's probably not.
3. What's my timeline? If I need it in 2 days, I'm in Scenario C. If I can wait 2 weeks, I'm in A or B.
4. What's the cost of failure? If the part fails, does the whole product stop? That's Scenario B. If it's just a test, that's C.

I can't give you a universal answer because there isn't one. Our needs change order to order. What I can tell you is what worked (and failed) for me.

To be fair, Xometry is a great platform for most of what we do. The instant quoting engine is genuinely useful. But the 'best practice' of just uploading a step file and clicking 'buy' can be costly if you don't think about your specific scenario first.

I still keep a checklist from the $1,200 initial mistake I mentioned earlier. It's saved me from repeating that error a dozen times over.