Tolerance and permitted deviation

Frequently asked questions

What is the difference between flat and level?

FLAT is about the surface's own local smoothness; LEVEL is about its relationship to horizontal. A floor that falls two centimetres across a room but has no bumps in it is perfectly flat and badly out of level. A floor that is dead horizontal on average but ripples every metre is level and not flat. They matter to different things: flatness decides whether a tile lippage appears, whether a long piece of furniture rocks, whether a forklift jolts; levelness decides whether doors swing open on their own, whether water runs to the drain, whether a run of units meets a worktop straight. This is why the American system quotes two numbers and the British straightedge check quotes one — a two-metre straightedge measures flatness only, and a floor can pass SR1 while sloping steadily. Anyone specifying a floor by surface regularity alone has said nothing about level, and that omission is a common source of dispute.

Why is a tolerance quoted over a length?

Because deviation without a gauge length is meaningless, and the same absolute error is trivial at one scale and unacceptable at another. Three millimetres out of plumb is invisible over a storey height and severe over a door frame; a five-millimetre dip is nothing across a room and a defect under a two-metre straightedge. So every honest tolerance names both numbers — so many millimetres in so many metres, or a gap under a straightedge of a stated length — and comparing two specifications means comparing both. It also explains why tolerances do not add up the way intuition suggests: a wall built to a permitted deviation per storey does not accumulate that deviation over four storeys, because each is measured from the reference rather than from the one below. Where they DO accumulate is the thing to watch — a series of components each cut to its own tolerance can leave a final gap far outside any of them, which is why setting out works from a datum and not from the last thing fixed.

Whose tolerance governs when two trades meet?

Neither, unless somebody wrote down the junction — and that gap is where most tolerance arguments live. Each trade works to a standard that is reasonable on its own: a concrete frame has a permitted deviation, a window has a manufacturing tolerance, a cladding panel has another, and every one of them may be within limits while the joint between them is impossible. The frame leans a permitted amount, the opening is a permitted amount out of square, and the panel that must cover both is made to a tighter tolerance than either. The answer used on well-run jobs is to design the JOINT to absorb the sum: a cover bead wide enough for the worst combination, a shim gap, an adjustable bracket. That is also why surveying the structure before ordering the components is standard on facades — the tolerance is not the problem, discovering it after manufacture is.