A printed cube looks fine from above and leans when you view it from the front. Vertical walls are not vertical. Two parts that should mate flush meet at an angle, and no amount of slicer adjustment changes it.
What a lean means mechanically
If the X gantry is not perpendicular to the direction the Z axis travels in, then every time the machine steps up a layer it also displaces the toolhead very slightly sideways. Each layer is printed correctly relative to the one below it; the stack is what goes wrong.
The result is a consistent shear rather than a random defect. Every part on the plate leans the same way by the same proportion, which is the signature that distinguishes this from anything the slicer or the extruder could cause.
Measure it before adjusting anything
Print a tall, thin rectangular tower — something like 20 mm square and 100 mm high — then stand it on a known flat surface next to an engineer's square.
Measure the gap at the top of the tower between the part and the square's blade. That gap, over the tower's height, is the whole error expressed as a ratio you can work with. A tenth of a millimetre over 100 mm is negligible; a millimetre over the same height is a machine that needs attention.
Do it in both directions. A lean in X and a lean in Y have different causes, and knowing which one you have halves the work.
The two common causes
Dual Z motors out of sync. On machines with two independent lead screws, the two sides can end up at different heights — usually because the machine was moved with one side dropping under gravity, or because the motors were powered off at different points in their travel. Everything then tilts.
The standard fix is mechanical and takes a minute: with the motors disabled, wind the gantry up by hand until both sides bottom out firmly against their top limits, so the two screws are forced into agreement, then power on and re-home. Machines with automatic gantry levelling do this in firmware, and running that routine is the first thing to try if yours has it.
A frame assembled out of square. Kit machines and machines that have been transported are the usual candidates. Check the vertical extrusions against the base with a square, and check that the diagonal measurements across the frame match each other. A frame whose diagonals differ is not square, and no amount of gantry adjustment compensates for it.
Less obvious causes worth checking
- One Z nut mounted higher than the other, so the gantry is clamped into a tilt regardless of screw position.
- A bent lead screw forcing the gantry sideways at particular heights — which produces a lean that changes with height rather than a constant one.
- Rails or rollers adjusted unevenly, letting one end of the gantry sit lower in its carriage.
- A bed that is not perpendicular to Z, which produces a part that looks leaned relative to its own base even when the gantry is fine. Print the tower and check the sides against a square rather than the base, and the two cases separate immediately.
Two corrections that make it worse
Do not correct a lean by shimming the bed. That tilts the plate to match the gantry, which makes the first layer consistent again and leaves every vertical surface exactly as wrong as it was — because the two errors are in different planes.
Do not reach for slicer settings. There is no setting for this, and the search for one has cost people entire weekends. It is a geometry fault in the machine and it is fixed with hands and a square.
Re-check after any transport
This is the single most useful habit for this fault. A printer carried across a room by its gantry, or laid on its side in a car, arrives with its geometry changed. Fifteen minutes of checking squareness and re-levelling after a move is far cheaper than the three failed prints that otherwise diagnose it for you.
Where the error shows up worst
Tall parts, obviously — the lean accumulates with height, so a short part hides an error that a column reveals. Parts that mate with other parts, where a fraction of a degree becomes a visible gap along a joint. And any part that has to sit flat against a wall or a panel, where a lean means the top edge stands proud while the bottom sits flush.
Material makes no difference at all here. If a lean appears to be worse in one filament than another, what is really happening is that a high-contraction material is adding its own distortion on top — which is worth ruling out before dismantling anything, by printing the same tower in PLA.