A reproduction part can be dimensionally perfect in CAD and still refuse to go in. That gap between the model and the physical object is where most of the real engineering work happens, and it’s the difference between a part that clips home with a satisfying click and one that sits in a drawer.
Here’s what actually stands in the way.
Shrinkage is the big one
Thermoplastics are extruded hot and then cool. As they cool they contract. That contraction is not a rounding error — depending on material it runs roughly 0.3% to 2%.
On a 10 mm feature, 0.5% is 0.05 mm and nobody notices. On a 200 mm dash bezel, the same percentage is a full millimetre, and a millimetre of error across a mounting span means the clips don’t line up with the holes.
Rough figures by material:
| Material | Typical shrinkage | Practical note |
|---|---|---|
| PLA | ~0.3–0.5% | Low shrinkage, but unsuitable for interiors on thermal grounds |
| PETG | ~0.5–0.8% | Predictable and well-behaved |
| ASA | ~0.5–0.9% | Higher, and more sensitive to chamber temperature |
| ABS | ~0.7–1.5% | Highest of the common options; prone to warping on large flat parts |
Shrinkage also isn’t uniform. A part shrinks more along its longest axis, and it shrinks differently where walls are thick versus thin. You can’t simply scale a model up by a fixed percentage and call it solved — the compensation has to be applied with an understanding of where the critical dimensions are.
Not every dimension deserves equal attention
This is the shift in thinking that separates a working reproduction from an approximate one. On any given part, only a handful of dimensions actually determine whether it fits. The rest can drift by a few tenths and nobody will ever know.
On a typical dash vent, the dimensions that matter are:
- Clip spacing — centre-to-centre distance between mounting features. Get this wrong and nothing else matters.
- Clip cross-section — determines retention force. Too thick and it won’t insert or it cracks the housing; too thin and it rattles loose.
- Outer perimeter — the visible gap around the bezel. Errors here are cosmetic but immediately obvious.
- Duct interface diameter — where the part meets the ductwork behind it.
Louver thickness, internal rib geometry, and the exact radius of a decorative fillet are, functionally speaking, free. We measure everything, but we hold tolerance where it counts.
Elephant’s foot and first-layer squash
The first layer is deliberately pressed into the build plate to make it stick. That squash makes the bottom 0.2–0.4 mm of the part slightly wider than the model — the effect known as elephant’s foot.
On most geometry it’s cosmetic. On a part that has to slide into a tight recess, that flared base is exactly what stops it going in, and it’s maddening to diagnose because the part measures correctly everywhere except the very bottom edge.
The fix is a small chamfer on the bottom edge in the model itself, which gives the squash somewhere to go.
Holes always print undersized
A circular hole printed from straight-line extrusion segments is approximated as a polygon inscribed inside the true circle. Combine that with a small amount of material flowing inward as it cools, and printed holes come out consistently smaller than modelled — typically 0.1–0.3 mm on the diameter.
Two consequences worth knowing:
- Holes that take a fastener need to be modelled oversize, or drilled to size afterwards
- A hole that has to clear a locating pin needs deliberate clearance designed in, not assumed
Clearance is designed, not hoped for
Two mating parts modelled at exactly the same dimension will not assemble. They’ll interfere. Real assemblies need a designed gap, and the size of the gap depends on the fit you want:
- Sliding or free fit — roughly 0.3–0.5 mm of clearance. Moves without binding.
- Close fit — roughly 0.15–0.25 mm. Assembles by hand, holds position.
- Press or interference fit — 0.0 to slightly negative. Requires force, holds without fasteners.
A retaining clip that’s meant to snap in and stay is deliberately an interference fit. A bezel that has to sit flush against a curved dash is a close fit with a little compliance designed into the geometry so it can flex to the surface.
Print orientation changes the numbers
The same model printed in two orientations produces two dimensionally different parts. Accuracy in the Z axis is governed by layer height and is generally excellent. Accuracy in X and Y is governed by nozzle diameter, extrusion width and corner behaviour, and is slightly looser.
Orientation also determines strength. Layer adhesion is the weakest direction in any FDM part — typically well under half the strength measured along the layers. A mounting tab printed so that the load pulls the layers apart will fail at a fraction of the load the same tab handles when printed with layers running along the stress path.
So orientation gets chosen for two things simultaneously: putting the critical dimensions where the process is most accurate, and putting the layer lines where the load won’t peel them.
How we validate before anything ships
Theory only goes so far. The process that catches problems:
- Measure the reference — multiple samples where we can get them, because original parts have their own manufacturing variation and thirty years of distortion
- Model with compensation applied to the critical dimensions
- Print a test article — often a partial print of just the mounting interface rather than the whole part, which is faster and isolates the thing being tested
- Test fit on the actual vehicle, because a hole in a thirty-year-old dashboard is not necessarily where the factory drawing says it is
- Iterate — first revisions are rarely final
- Lock the parameters — material, printer profile, orientation and slicer settings all get recorded, because changing any one of them changes the dimensions
That last point is why we don’t sell files for parts where fitment is tight. A dimensionally compensated model is compensated for a specific process. Print it on a different machine with a different profile and the compensation is wrong.
Why the original part is a flawed reference
The thing being measured is usually broken, and it’s been through the same thermal and UV history described in our piece on what destroys interior plastic. Which means:
- It has shrunk from plasticizer loss, sometimes measurably
- It may have warped from heat cycling
- Broken edges have to be reconstructed by inference
- The factory part had its own tolerance band to begin with
Measuring several examples and cross-referencing against the vehicle itself is the only way through this. It’s slow, and it’s the reason a new part takes weeks rather than an afternoon.
If a part you need isn’t in the catalogue, tell us about it — and if you have a good original to lend as a reference, that’s genuinely the most valuable thing you can offer.