12 Design Rules That Decide Whether Your Part Prints Well

Most files that arrive at a print bureau were drawn for a different process. They were designed for machining, or for injection moulding, or by someone who has never watched a printer work. They're usually printable, but they print worse than they need to, weaker, uglier, more expensive, and sometimes they fail in service for reasons that were designed in from the start.

These twelve rules fix nearly all of it. None of them require special software or advanced technique. They're just the things worth knowing before you export the STL.

1. Design walls as multiples of the nozzle

Your printer lays down beads about 0.4 mm wide. A wall is built from a whole number of those beads, you can't have two and a half.

Ask for a 1.0 mm wall and the slicer fits two 0.4 mm perimeters and then has 0.2 mm left over. Depending on settings it either leaves a gap, over-extrudes to fill it, or drops to a single perimeter. All three produce a weaker, uglier wall than if you had asked for 0.8 mm or 1.2 mm.

Design walls at 0.8, 1.2, 1.6 or 2.0 mm. For anything structural, 2.0 mm and up. Below 0.8 mm you're asking for a single-bead wall, which is fragile and prints unreliably.

2. Fifty degrees is the overhang limit

Each layer needs the layer below it for support. When a surface leans out past roughly 45-50° from vertical, each new layer is hanging over air and the extruded plastic droops before it solidifies.

Below 50° from vertical, the part prints cleanly with no support at all. Beyond it, the slicer generates support structures, which cost material, cost time, and leave marks on the surface where they were removed.

Design overhangs at 50° or shallower wherever you can. Where you can't, consider chamfering the transition instead of leaving a sharp horizontal ledge. A 45° chamfer under a boss costs almost nothing in material and eliminates the support entirely.

3. Bridges are fine; unsupported horizontals aren't

There's one useful exception. A bridge, a horizontal span between two points that are both already printed, works fine up to about 50 mm because the extruded plastic stretches taut between the anchors and cools under tension.

A horizontal surface anchored at only one end isn't a bridge, it's a cantilever, and it will droop.

Design holes in vertical walls as teardrops rather than circles where appearance permits. The top of a circular hole is an unsupported horizontal; a teardrop replaces it with two 45° faces that self-support.

4. Holes print undersized, always

This one catches everybody. Ask for a 5 mm hole and you'll measure 4.7-4.9 mm.

The cause is geometric. The nozzle traces a circle and extrudes a bead of finite width. On the inside of a curve, that bead squeezes inward slightly. The tighter the radius, the worse it gets, so small holes lose proportionally more than large ones.

Oversize holes by 0.2-0.3 mm in CAD if a fastener must pass through, or plan to drill them out afterwards. Drilling is often the better answer for anything that needs to be accurate, because a drilled hole in a printed part is round in a way a printed one isn't.

5. The first few layers spread

The nozzle squashes the first layer into the build plate to make it stick. That squash spreads the plastic slightly wider than nominal, and the effect persists for the first two or three layers. An "elephant's foot" of maybe 0.1-0.2 mm around the base.

For most parts this doesn't matter. For a part that must sit flat in a pocket, or a press fit that starts at the bottom face, it absolutely does.

Add a 0.5 mm chamfer to the bottom edge of parts with critical base dimensions. It gives the spread somewhere to go.

6. Orientation decides strength, and it isn't subtle

An FDM part is anisotropic. Along the layers it's close to the material's real strength. Across the layers, where each layer is a thermal weld to its neighbour rather than continuous polymer, it can be half as strong or worse.

This is the single biggest determinant of whether a printed part survives, and it's entirely free to get right.

Picture a hook. Printed flat on the plate, the layers run along the hook's length and the load pulls along them, strong. Printed standing upright, the layers stack across the hook and the load pulls the welds apart. It will snap at a layer line, usually at the point of highest stress.

Tell us the load direction. We orient every part deliberately, but we can only optimise for a load case we know about. If you say nothing, we optimise for surface finish and support minimisation and guess at the loading.

7. Fillet every internal corner

A sharp internal corner is a stress concentration in any manufacturing process, and in a layered part it's worse because a crack that starts there has a ready-made path to propagate along a layer boundary.

A 2 mm fillet on internal corners is close to free and dramatically improves fatigue life. This costs nothing in material and no extra print time. It's the highest return-on-effort change in this entire list.

External corners benefit from a small chamfer too, mostly for handling. A chamfered edge doesn't lift when the nozzle passes and doesn't cut your fingers.

8. Design in tolerance for parts that fit together

Two printed parts that mate need clearance, and how much depends on the fit you want:

• Free-running or loose fit: 0.4-0.5 mm total clearance
• Sliding fit: 0.3 mm
• Snug or press fit: 0.15-0.2 mm
• Interference (needs force): 0.05-0.1 mm

These are total clearances across the joint, so split them between the two parts. If you model both parts to nominal and expect them to fit, they won't. They will be too tight, because of the same bead-width effect that shrinks holes.

If a fit is critical and you only get one attempt, print a small test coupon of just the mating feature first. It costs almost nothing and beats reprinting the whole assembly.

9. Don't print threads below M8

Printed threads work, but only at coarse pitches. Below about M8 the thread form is smaller than the printer can resolve and you get a spiral ridge rather than a thread.

Better options, in order of preference:

1. Heat-set threaded inserts. A brass insert pushed in with a soldering iron gives you a real metal thread in a printed part. Cheap, fast, and vastly stronger than any printed thread.
2. A captive nut pocket. Design a hexagonal recess that a standard nut drops into.
3. Tapping. Print a pilot hole and cut the thread with a tap. Works well in ABS, PC and PA-CF.
4. A printed thread, only if it's M8 or larger and lightly loaded.

10. Watch the build envelope before you finish the model

Our largest single part is 300 × 300 × 300 mm on the K1 Max, or 256 mm cube on the P1S. Discovering at export time that your part is 340 mm long is a bad moment.

Larger parts get split and bonded, which is a normal operation and works well. But the split should be designed in, at a seam you choose, with registration features so the halves self-align. A split we've to improvise at the last minute is never as good as one you planned.

Check the longest dimension early. If it's close to the limit, design the split yourself or tell us so we can advise on where to put it.

11. Infill isn't the strength lever you think it's

The instinct when a part needs to be stronger is to raise the infill. It's usually the wrong lever.

Bending stiffness comes overwhelmingly from material at the outside of the section, the perimeters, not from the infill in the middle. This is the same reason an I-beam works.

Adding perimeters beats adding infill, almost every time. A part with four perimeters at 20% infill is stiffer and lighter than the same part with two perimeters at 50% infill, and it prints faster.

Raise infill when you need compressive strength through the body, when you'll be tapping threads into the material, or when the part must be machined afterwards.

12. Consolidate assemblies

This is the rule that most changes what's possible, and the one people forget because it requires thinking differently.

Injection moulding forces you to split designs into mouldable pieces with draft angles and no undercuts, then assemble them with fasteners. Printing has none of those constraints. Undercuts are free. Internal channels that no drill could reach are free. A nine-piece welded assembly can often become one printed part.

Fewer parts means fewer fasteners, fewer tolerance stack-ups, less assembly labour, no leak paths at joints, and usually less mass.

When you port a design to printing, ask which parts exist only because moulding demanded them. Those are the ones to merge.

In short

If you only apply four of these, apply these:

1. Fillet internal corners: free, and it prevents the most common failure.
2. Tell us the load direction: free, and it determines strength more than anything else.
3. Oversize holes by 0.25 mm: free, and it saves a reprint.
4. Use wall thicknesses that are multiples of 0.4 mm: free, and it makes every wall stronger.

All four cost nothing. Together they eliminate most of the problems we see.

We check this anyway

Every file that arrives gets a design-for-additive review before it's printed, free with every quote. We check wall thickness against the nozzle, look for unsupported overhangs, flag holes that will come out undersized, check the envelope, and choose an orientation.

If something won't work, we tell you before printing rather than after. But a file that arrives already designed for the process prints better, faster and cheaper than one we've to work around. So these rules are worth the twenty minutes they take to apply.