The setting out is *vital* as all the risers *MUST* be of the same height
and the treads the same thickness and depth to avoid trips and falls - and
this applies to whatever materials you use.

Of course all real world stairs will have some variation, however
small. I'd be interested to hear what the limits in acceptable error
is.

I've just built my first stair, in a chapel conversion - so there's a
lot of variation in the building to accommodate.

Mine is a u-shaped stair, climbing 3.2m in 5,2,11 steps (superimposed
made up treads and risers on top of rough carcassed carriages), and
the error on the going probably 3mm variation (in 285mm), error in
rising 4mm (in 178mm) - apart from the bottom step where a twist in
the ground floor relative to the upper floor resulted in a 10mm
variation at one side (of a 1200mm wide stairway). With hindsight, I
would have made more effort to spread that 2mm at a time over the
bottom 5 steps.

Now I've tried to detect the error by walking up and down in every
conceivable pattern - but other than measuring it, I can neither see
it nor detect it walking up and down.

I seem to vaguely recall 10mm as being the figure for variation in the
rise that's likely to be detectable.