Tolerance: The Resource Every Envelope Detail Spends

The pattern recurs wherever something made in a controlled environment meets something built in the open. A window unit meets an opening. A precast panel meets a frame. A composite roof panel meets a purlin line. A door leaf meets paving. A profiled sheet meets a set of rails. In each case one side was produced to a dimension and the other was constructed towards one, and neither is at fault for being what it is.

What the records emphasise is that this is known in advance and therefore designable. The curtain walling record notes that brackets reach a slab edge whose position and level are known only within a range, and whose shortening continues after the facade is installed, so the bracket arrangement has to take up both the setting-out deviation and the movement that follows occupation.

The records name the absorbing component each time, and it is rarely the one that gets the attention. For a window it is the perimeter gap. For a hung panel it is the fixing, which is why the precast record says only the fixings can absorb the difference between factory accuracy and site accuracy. For an infill wall the flank restraint takes up the difference between where the column is drawn and where it is. For a pedestal terrace it is the support layer, which resolves the difference between a falling deck and a level surface at every support.

Naming it matters because it tells you what must not be used for something else. A perimeter gap doing tolerance duty is also the space the seal moves in and the space the insulation occupies. A fixing taking up deviation is also the load path. A pedestal absorbing deck level is also the point load path onto whatever lies below.

The records treat available adjustment as a quantity with a limit. The precast entry asks how much adjustment is built into the fixings and what happens when it runs out, and asks what the plan is if the structure turns out to be outside the assumed range. Those are questions for the facade engineer, and they are asked before the frame is set out rather than after.

The composite panel roof record shows the same limit from the rigid side. Panels are rigid and their joints have little tolerance, so frame and purlin setting-out is a panel issue as much as a structural one, and deflection under load reaches the joints because a panel cannot take up movement within its own thickness. Where the absorbing capacity is small, the accuracy demanded of the structure rises.

This is the sentence the whole subject turns on. In a facade of hung units the joints between units, not the faces, are where the assembly is designed to meet water. If the fixings cannot take up the difference, the deviation appears in the joint, and the joint is simultaneously the weathering element. A dimensional problem has become a moisture problem without anything being installed incorrectly.

The window perimeter shows the same transfer at a smaller scale. Squeeze the gap out and the seal has less room to move in and the insulation has less space to occupy, so a perimeter that looks neater on the day may have lost both. The record notes that sealants and tapes have finite movement capability, which the product literature states and general guidance cannot.

The precast record lists this among the common misunderstandings directly: uneven joints are read as poor workmanship, when they usually show structural tolerance appearing where the adjustment ran out. That reading matters because it points an investigation at the frame and the fixing layout rather than at the installer.

It also explains why a facade of this kind is largely decided before anything reaches the site. Cast-in fixings have to be positioned before the frame is poured, which puts a facade decision very early in the structural programme, and moving a fixing afterwards is difficult rather than routine.

The external door threshold record describes the most crowded version of the problem. Hardware clearance and drainage compete for the same vertical room, and whichever is fixed first constrains the other: raise the external surface to reduce a step and the door seal compresses differently or the leaf begins to catch; lower it to keep the channel draining and the step returns.

Because the door is manufactured to its own tolerances and the paving is laid to site tolerances, the record's conclusion is that the setting out has to be agreed before either is ordered, not reconciled afterwards. It adds that groundworks, floor, frame and paving trades each set part of the level here, usually weeks apart, so agreeing the datum once and recording it is what gives each trade a level to work to.

The practical discipline the records describe is to decide what the gap is for before anybody closes it up. On a window perimeter, is it carrying the difference between the unit and the opening, the movement of the joint over the life of the unit, or the insulation around the frame? It is usually all three, and each of them is reduced by anything that takes room out of the gap for appearance.

The same discipline applies to a facade joint, a head gap, a pedestal height and a panel joint. Once the adjustment has been used, it is used, and the next requirement arriving at that position has to be satisfied somewhere else or not at all.