Two errors, two different fixes
A printed part that comes out the wrong size is usually wrong in two independent ways at once, and the reason people go round in circles is that they try to fix both with the same adjustment.
The first error is proportional. Plastic extruded hot and cooled to room temperature contracts, and the contraction is a percentage of the length. A part designed at 100 mm that measures 99.4 has lost 0.6 percent; the same material on a 200 mm part would lose about 1.2 mm. The correction is a scale factor: scale = nominal divided by measured, here 100 over 99.4, or 100.60 percent. Apply it in the slicer or in CAD and every dimension moves in the right proportion.
The second error is not proportional and does not respond to scaling. Printed holes come out undersized, and the shortfall is roughly the same in millimetres whether the hole is 4 mm or 20 mm. There are two mechanisms and both are additive. A circle in an STL is a polygon, and the printed toolpath follows the flats, so the inscribed circle is smaller than the intended one by an amount set by the facet count and, to a first approximation, independent of scaling. Separately, the extruded bead spreads slightly inward on the inside of a curve, and the first few layers spread more because of elephant foot, so material intrudes into the hole. Neither of those is a percentage of diameter.
So the compensation is additive: modelled diameter = wanted diameter plus (nominal minus measured). If an 8 mm hole prints at 7.72, the compensation is 0.28 mm, and a 6 mm hole you want to come out at 6.00 should be modelled at 6.28. Scaling the part to 100.6 percent would have added only 0.048 mm to that 6 mm hole, which is a sixth of what it needs.
How much shrinkage to expect
Material shrinkage figures published for injection moulding are a poor guide to printing, because a printed part is anchored to a heated bed while it cools and is built from thin beads that each solidify almost immediately. Constrained cooling suppresses most of the free shrinkage. In practice, in-plane shrinkage on a well-adhered part is usually well under one percent even for materials whose raw shrinkage figure is several times that.
| Material | Typical in-plane error, printed and constrained | Notes |
|---|---|---|
| PLA | 0.1 - 0.4% | Lowest of the common materials, largely amorphous behaviour on cooling |
| PETG | 0.2 - 0.5% | Consistent, but bead spread is generous so holes suffer more |
| ABS / ASA | 0.4 - 0.8% | Higher, and strongly dependent on chamber temperature and part size |
| Nylon | 0.5 - 1.5% | Semi-crystalline and hygroscopic; keeps moving after the print |
| Filled grades | Lower than the base polymer | Fibre or mineral content restrains contraction |
These are ranges from measurement, not specifications, and your machine and geometry matter more than the material label. The only figure worth acting on is the one you measured on a part shaped like the parts you print. A thin flat plate and a chunky block of the same material do not shrink the same amount.
When the number you measured is not shrinkage
Before applying any scale factor, check that what you measured is actually material behaviour. Three things masquerade as shrinkage and none of them should be corrected with a scale.
If X and Y disagree by more than about 0.2 percentage points, that is machine calibration or warp. Material shrinkage in the print plane is the same in both directions; belts and steps-per-millimetre are not. Applying a single scale factor will then leave one axis wrong in the other direction.
If Z is out, suspect first-layer squish and layer division before shrinkage. Squish removes a fixed amount from every part, so a 20 mm block and a 100 mm block are both short by the same tenth of a millimetre, which is a five times larger percentage on the small one. Scaling Z to fix one will ruin the other. The layer height helper shows the Z a given part actually reaches, which usually explains the residue.
If the error changes with flow settings, it is over-extrusion, not shrinkage. An over-extruding printer makes outside dimensions large and holes small at the same time, which is the classic signature and is entirely fixable at the source. Run the e-steps and flow calibration first; a dimensional compensation applied on top of an uncalibrated extruder will be wrong the moment the flow changes.
Measuring in a way that means something
Let the part cool fully. Nylon and other semi-crystalline materials continue to move for hours and can keep relaxing for days, and nylon in particular absorbs moisture from the air and grows slightly as it does. Measuring a warm part gives an optimistic shrinkage figure and a scale factor that undercorrects.
Measure holes with pin gauges if you have them. A caliper inside a printed hole is measuring across whatever part of a polygon the jaws happen to land on, and readings will vary by a tenth of a millimetre depending on rotation. Where a hole exists to take a fastener or a bearing, what matters is the largest pin that passes, which is the inscribed diameter, and that is the number to compensate against.
Measure away from the first few layers. Elephant foot flares the base outward, so a caliper across the bottom 1 mm of a block reads large and tells you nothing about the rest of the part. Most slicers have an elephant foot compensation setting that handles this locally, and using it is better than distorting the whole model to correct one edge.
Finally, decide whether you need the compensation at all. A hobby bracket with 0.5 mm of clearance around every fastener does not care about 0.3 percent. A part mating with a bearing does. Set the tolerance field to what the job actually requires, and note that a fixed percentage error means you stay in tolerance only up to a certain part size — at 0.6 percent shrinkage, plus or minus 0.15 mm holds to about 25 mm and no further.
Questions people ask
Why are my printed holes always too small?
Two additive causes stack up. First, a circle stored in an STL is a polygon, and the nozzle follows the polygon, so the largest pin that fits is the inscribed circle rather than the intended one. Increasing the export resolution reduces this but never removes it. Second, the extruded bead spreads slightly inward on the concave side of a curve, and the bottom layers spread more because of elephant foot, so material intrudes. Both effects are roughly constant in millimetres, which is why the compensation is additive on the diameter rather than a percentage. Measure one hole, get the offset, and apply that same offset to every hole regardless of size.
Should I apply the scale factor in the slicer or in CAD?
In CAD if the part has holes and mating features, in the slicer if it does not. A slicer scale multiplies everything uniformly, which is correct for shrinkage and wrong for holes, since holes need an additive offset that scaling will not deliver. Modelling the compensation means you can scale the outside dimensions and offset the holes independently, which is the only way to get both right. The slicer scale is a reasonable shortcut for decorative parts, large single-dimension prints, and anything where nothing has to fit anything else.
Does shrinkage compensation change between filaments?
Yes, and between colours of the same product often enough that it is worth rechecking. Pigment and filler loading affect both cooling behaviour and flow, so a black and a natural spool from the same maker can differ measurably. The bigger jumps are between material families: PLA sits low, ABS and ASA higher and more chamber-dependent, nylon higher again and unstable over time. If you hold a real tolerance, print a test coupon with each new material and remeasure rather than carrying a factor across.
Is elephant foot the same as shrinkage?
No, and they need different fixes. Elephant foot is the flare at the bottom of a part, caused by the first layers being squashed against the bed while still hot and by the weight above them spreading the soft plastic outward. It is local to the bottom couple of millimetres and it makes the base larger, whereas shrinkage makes the whole part smaller. Scaling the model down to compensate for elephant foot makes everything above the base wrong. Use the slicer setting designed for it, or add a small chamfer to the bottom edge in CAD, which achieves the same thing and is more predictable.
My part is right in X and wrong in Y. What is that?
Not material shrinkage, which is the same in both directions in the print plane. The usual causes are mechanical: belt tension differing between the axes, a steps-per-millimetre value wrong on one, a frame that is out of square so moves are skewed, or a part that warped and pulled one dimension in during cooling. Applying a scale factor at this point would encode the machine fault into every model you ever print. Square the frame, check the belts, print a large square test and measure both diagonals, and only reach for a scale factor once X and Y agree with each other.