How I Buy Custom Parts: Shapeways vs. CNC Machining vs. a Raspberry Pi Printer

There's no single right way to buy custom parts

After five years of managing purchasing for a 30-person design firm—roughly $80,000 a year across eight vendors—I can tell you one thing: anyone who says "just pick the cheapest quote" hasn't ordered enough parts.

Custom parts aren't a commodity. A bracket that works fine as a prototype might fail in production. A design that's perfect for 3D printing could be a nightmare to CNC machine. And the "cheapest" option on paper often ends up costing more after shipping, setup fees, rework, and your own time.

In my experience, custom part buying splits into three scenarios. Different situation, different answer:

• Complex geometry, small quantities → use a 3D printing service like Shapeways
• Tight tolerances, production volumes → use CNC machining
• Design still moving, rapid iteration → use an in-house printer (a Raspberry Pi setup works great)

There's no universal winner here. But there is a right answer for your specific situation. Here's how I make that call.

Why total cost beats sticker price

When I took over purchasing in 2020, I compared quotes like everyone does. Sticker price. Cheapest wins. Then I found a vendor that was $2,400 cheaper than our regular supplier for a 200-part order. Looked like a win.

They couldn't provide a proper invoice. Handwritten receipt only. Finance rejected the expense report. I ate the $2,400 out of the department budget.

That was the wake-up call. I now calculate total cost of ownership (TCO) before comparing any vendor quotes. TCO is:

• Base part price
• Setup or tooling fees
• Shipping and handling
• Rush fees, if you need it fast
• Rework and redo costs, if quality misses spec
• Your time managing the order

The $500 quote that turns into $800 after shipping and revision fees is not cheaper than the $650 all-inclusive quote. I've made that mistake. It's not one I'll repeat.

Scenario A: Complex geometry, low quantities → use a 3D printing service

Your part has internal channels. Or complex curves. Or you need 15 pieces of something that can't be machined in one piece. That's the 3D printing scenario. We use Shapeways for most of this—they're a multi-process platform, so we can get SLS, MJF, or metal 3D printing without shopping around. Upload a CAD file, get an instant quote, receive parts in days.

Three things make them the TCO winner in this scenario:

Instant quoting saves real money. A traditional machine shop takes one to three days to quote. Shapeways gives me a number while my coffee's still hot. When you're processing 60-80 orders a year, that adds up to real dollars of admin time.

Design feedback catches problems before they cost money. We once uploaded a file with an unprintable overhang. Their system flagged it before production started. If it had gone to print, that was $600 in wasted material and a week of delay. That's the kind of "free" service that doesn't show up on any invoice but makes the TCO math work.

Reordering is almost too easy. We've saved our standard bracket and mounting plate files in our Shapeways store. When we need another batch, it's a couple of clicks. No re-uploading, no re-quoting. And when I'm browsing for new suppliers or off-the-shelf hardware, the Shapeways shop interface is genuinely the most reorder-friendly system I use across all eight vendors.

One caveat: if you're ordering functional parts where consistency matters—not just visual prototypes—check whether your supplier follows recognized standards. ISO/ASTM 52945 additive manufacturing guidance is a useful reference for how multi-step AM processes (printing, cleaning, post-curing) should be controlled. It's not something you'll see on every quote, but it's worth asking about when parts need to be trustworthy.

TCO math: per-part price is moderate, but setup costs are near zero, minimum quantities don't exist, and your time cost is minimal. For 1-50 parts with complex geometry, this route almost always wins.

Scenario B: Precision, tolerances, production volumes → use CNC machining

3D printing has limits. And those limits show up fast when you need a part flat to 0.005 inches, a threaded insert for a production housing, or 500 aluminum brackets with consistent dimensions. That's when CNC machining is the right call.

Here's something I didn't know until our machining vendor explained it: tooling selection is a big part of the cost equation. For angled surfaces, an adjustable angle milling cutter lets the machine cut compound angles in a single setup. Fewer setups mean tighter tolerances and lower cost per part. That kind of detail never shows up on the invoice, but it drives the TCO math.

But there's no getting around it: CNC quoting is slower and more conversational. You're not just uploading a file—you're answering design-for-manufacturability (DFM) questions. "Can you widen this tolerance?" "Do you really need this surface finish?" Those questions are annoyingly valuable. They're also where things break down if you don't have a review process.

We didn't have a formal DFM review process once. We approved a quote for 50 brackets without questioning the toolpath strategy. The parts arrived with razor-sharp edges and one tolerance out of spec. The redo cost us three weeks and $1,100 in rush machining fees. Now, every machining quote goes through a checklist before we approve it. That's a process gap I should have closed years earlier.

TCO math: setup costs run $100-500, but per-part prices drop below 3D printing at quantity. You'll spend more time on quoting and review—budget for it, and build a review process before you need one.

Scenario C: Rapid iteration, in-house agility → use a Raspberry Pi 3D printer

What if you're still designing the part? Three design revisions this week, and each one changes something fundamental? That's the DIY 3D printing scenario.

We bought a modest FDM printer for the office in 2023. Not everyone was on board. The budget committee asked why we'd spend money on a "toy" when we had Shapeways with instant quoting. The numbers didn't strongly justify it. My gut said we needed a design-test loop measured in hours, not days.

Learning how to use Raspberry Pi for 3D printing completely changed the cost calculation. OctoPrint on the Pi gives us remote monitoring through a webcam, temperature tracking, and push notifications when a print finishes—or fails. I don't babysit the printer anymore. The "maintenance time" concern turned out to be mostly theoretical.

In the first month, that $275 printer produced 12 functional prototypes. Outsourcing those through Shapeways or a local shop would have cost $40-120 each. Even after counting filament, failed prints, and the Raspberry Pi setup, it paid for itself in about six weeks.

But I'll be honest about the TCO: a cheap FDM printer isn't a replacement for professional manufacturing. Tolerances are looser. Surface finish isn't great. Some materials produce fumes you don't want in an office. Use it for iteration and learning—not for customer deliverables.

TCO math: equipment runs $250-500, filament is roughly $20 per kilogram, and your time cost drops significantly with OctoPrint. The speed of iteration is unmatched by any outsource vendor.

How to tell which scenario you're in

Alright, so which one applies to you? Here's the practical test I use. Four questions:

1. How many parts do you need? One to 50 with complex geometry → 3D printing service. Fifty or more with tight tolerances → CNC machining. Just a few, and still testing → DIY printer.

2. What's your deadline? Need parts in hand within three days? A 3D printing service like Shapeways is the most reliable play—their turnaround is consistent. With CNC, you're at the mercy of the shop's queue. DIY is fast only if you've already dialed in the print settings.

3. Could the design change before you order? If yes, don't buy a production process. Print it in-house, keep iterating. Lock the design, then think about CNC or a professional service.

4. What does a failed part actually cost you? A bad print in your office costs $5 in filament and a couple of hours. A bad batch from a vendor costs hundreds of dollars and an uncomfortable conversation with your VP. The higher the failure cost, the more you should lean on vendors with design feedback and proper quality standards.

One more thing: these scenarios aren't either-or. We've printed form-fit prototypes in our office, ordered functional prototypes through Shapeways, then sent the production version to a machine shop. Three processes, one workflow. The TCO thinking is what ties them together.

Bottom line

Stop comparing sticker prices. Start comparing total costs—including your time and your risk.

Prototyping complex geometry? Use a 3D printing service like Shapeways. Production with precision requirements? Use CNC machining. Still figuring out what to build? Get a Raspberry Pi-powered printer and iterate fast.

None of these is universally the best. But one of them fits your situation. Figure out which one, and the buying decision gets a lot easier.