A printed tolerance is not a property of the material or of the design. It is a property of your machine, your nozzle, your material and your first-layer settings taken together, which is why two people following identical advice get different fits. Everything below is a starting point that a twenty-minute test replaces with your own number.
A clearance ladder to start from
Clearance here means the gap designed into the model, per side, between two mating features.
| Fit | Gap per side | Behaviour |
|---|---|---|
| Interference | 0 to 0.1 mm | Needs force or heat to assemble. Holds without glue. |
| Snug | 0.1 mm | Assembles by hand with effort, does not move afterwards. |
| Close sliding | 0.2 mm | Moves smoothly, minimal play. Lids, drawers, panels. |
| Free running | 0.3 mm | Rotates or slides freely. Hinges, axles, moving assemblies. |
| Loose | 0.4 to 0.5 mm | Tolerates paint, debris and print-to-print variation. |
Two clarifications that save arguments. These are per side, so a shaft in a hole needs twice the figure on the diameter. And they assume both parts are printed; a printed part mating with a bought one — a bearing, a bolt, an aluminium extrusion — needs the printed side to carry the whole allowance.
Both errors run the same way
The reason printed fits are tight far more often than loose is that the two mating features err in opposite directions, and both of those directions close the gap.
Holes come out undersized. A circle is approximated by straight segments that sit inside the true circle, and the extruded bead bulges inward on a tight radius. The effect is worse on small holes and worse with a wide nozzle.
Outside features come out oversized. The bead bulges outward at corners, the first layer squashes outward into elephant's foot, and the walls can bulge under pressure. A printed peg is a little fatter than drawn.
Design a 10 mm peg into a 10 mm hole and neither error is on your side. This is why printed assemblies so often need a scalpel.
Thermal contraction is the smaller effect, and people blame it first
Take a 25 mm bore in PETG and apply a mid-range contraction of 0.4%. The bore finishes at 24.9 mm, and correcting for it means modelling at 25.1004 mm — a correction of 0.1004 mm. A tenth of a millimetre on a 25 mm bore is the same order as the elephant's foot on a first layer, which is precisely why blaming contraction first is usually wrong. The shrinkage calculator will size the effect for your own bore, so you can rule it in or out before changing anything else.
Now compare that to what people routinely measure on printed holes, which is a few tenths of a millimetre undersize on a hole of this size. The extrusion-geometry error is comfortably the larger of the two, and it does not scale the same way, which is why a global scale factor cannot fix holes and outer dimensions at the same time. Shrinkage and tolerance covers when the scale factor is the right tool; for fits, it usually is not.
If your holes specifically are the problem, holes printing undersized sets out the fixes in order.
The test that replaces this page
Print a clearance comb: one part with a row of pegs or slots at 0.1 mm increments, its mate with a single matching feature. Assemble each pairing and note which one behaves the way you want.
Do it once per material and once per nozzle size. It takes twenty minutes, it is good until you change something, and it converts every table on the internet — including the one above — into a number that is actually true for your machine. Write the result somewhere you will find it.
Orientation changes the answer
The same hole is a different hole depending on how it was printed.
A vertical hole, its axis along Z, is drawn as a circle on every layer and is the most accurate kind. It is still undersized, but predictably so.
A horizontal hole, its axis in the print plane, is built as a series of stacked chords, and its upper region has to bridge. The top of the bore droops inward and the profile is not round. Expect a horizontal hole to need more clearance than a vertical one, and design it as a teardrop if it does not have to be circular.
The same asymmetry applies to threads, splines and any feature whose accuracy you care about. If a fit matters, orient it deliberately rather than letting the slicer's default placement decide.
Design features that forgive
The most robust answer to printed tolerance is a design that does not depend on it.
- Chamfer every entry. A lead-in chamfer on a hole and on a peg turns an interference fit into something that starts square and assembles. It also hides elephant's foot at the base of a bore.
- Use slots instead of holes wherever a fastener passes through. A slot forgives positional error in one direction for free.
- Split clamps and screws instead of press fits, for anything that has to be adjustable or reassembled.
- Heat-set inserts instead of printed threads, for anything that will be undone more than a few times. A brass insert melted into a printed boss is stronger and repeatable in a way a printed thread is not.
- Design in an adjuster — a small grub screw, a wedge, a springy tab — where the fit genuinely matters and reprinting is expensive.
When a fit fails, diagnose before reprinting
Measure the two parts rather than guessing which one is wrong. If the hole is undersized and the peg is on size, the fix is hole compensation. If both are off in the same direction, it is a flow or line-width calibration problem, not a tolerance one. If it fits at the top of the bore and not at the bottom, it is elephant's foot and the fix is a chamfer or an elephant's-foot compensation setting.
Reprinting a whole assembly at a different scale is the expensive way to discover which of those it was.