Tolerance: Who Absorbs the Difference Between Drawn and Built

The entries treat tolerance as a design quantity rather than as workmanship. A precast joint width is described as a design decision covering both deviation and movement. A cap design absorbs a certain amount of pile position deviation, and what happens beyond that is a question for the engineer. A base plate arrangement has to absorb the difference between where cast-in bolts landed and where they were drawn.

In each case somebody decided in advance that deviation would be taken up at a particular place. That place is a component of the building, even where it looks like a gap.

The sharpest cases are where work made to one standard of accuracy meets work made to another. A column base plate is the classic: the concrete beneath is placed to one standard, the steel above is fabricated to another, and the bedding beneath the plate is where that difference is resolved.

The pad foundation entry describes the same meeting from below. Holding-down bolts are cast into wet concrete and have to meet a plate drilled elsewhere, and correcting one that has moved is a structural repair, because the arrangement being altered is the one that transfers uplift and horizontal force.

Because the frame is fabricated to fixed dimensions, a base set out inaccurately cannot be absorbed later, and the reconciliation of the two tolerances is a design matter rather than a site one.

Several elements in the library exist to hand a line to whatever comes next, and each of them passes on its own error. The first course of an insulating formwork wall sets the plumb and line the rest of the wall inherits, so deviation in the substructure becomes deviation in the wall. The lowest joint in a precast structure sets the geometry for everything above, because deviation there is inherited by every storey.

In lighter work the same thing happens at a smaller scale. A framed wall is only as straight and as plumb as the plate or track it stands on, and correcting it later means correcting the whole wall. A volumetric stack inherits the accuracy of the plane the first row is set on, which cannot be recovered higher up.

This is the hardest part to argue about on site, because the evidence is not where the problem is. In a volumetric building, variation within each unit, the setting-out of the first row and the take-up at every connection add together up the stack, and where the connections run out of capacity the effect shows at the facade line or along a corridor, well away from the joint that caused it.

A precast joint dimensioned for the drawn gap rather than for accumulation behaves the same way: it will either refuse to close at one position or open too far at another, and by then the units either side of it have already been made.

Some of the worst outcomes come not from too little allowance but from somebody consuming it. A deflection head lets the structure above move without pressing on a partition, and the gap it depends on is invisible once the partition is lined; it is commonly lost to a screw through both legs of the track, a service bracket, or a board fixed hard to the soffit.

The same pattern appears around an inserted platform: a deliberate gap between a freestanding structure and the building is closed by a later finish, quietly creating a connection nobody designed. And under a base plate, a plate that is not fully bedded does not spread load as designed, with packing that leaves part of it unsupported driving the whole load through the remaining contact.

The bearing entry makes the general case. Construction tolerances mean the actual bearing achieved on site frequently differs from the one drawn, and that difference matters: a member can be entirely sound and still fail at its bearing, through insufficient length, through crushing of what it bears on, or because the support was not where the design assumed.

That is why the entries treat packing and bedding as design considerations rather than as workmanship. Where a surface is not true, how a plate is bedded or packed is part of the design of the connection.

Some deviations are absorbed and some reopen the design. Correcting a misplaced pile is not a matter of adjusting the wall above, since the redesign happens in the cap that connects them. Correcting a holding-down bolt that has moved is a structural repair. A pile that finished away from its intended position can leave a group off-centre beneath the load it was meant to receive.

Knowing which kind you are looking at is the practical value of this subject. The question is never whether the building matches the drawing; it is whether the difference landed somewhere that was arranged to receive it.

The useful conversations happen before the element that inherits a line is made. What deviation the joint or connection has been designed to absorb, how that compares with what is achievable on site, and who checks the setting-out before the next thing is fabricated, are questions with answers while there is still time.