Site Prep, Foundations & Tolerance

In conventional construction the foundation is forgiving. If the setting-out is a little off or the slab a little uneven, the trades above absorb it — a wall is packed out, a floor screeded level, a frame shimmed. The building heals the foundation's small sins as it rises. Modular removes that forgiveness. The modules are finished, rigid, precise and identical; they cannot be stretched, packed or fudged to fit a base that is in the wrong place. Whatever error is in the foundation is handed straight up to the first row of modules and, worse, accumulates as the stack grows, because each module sits on the last. A few millimetres of drift at the base can become a serious misalignment several storeys up.

So for modular the foundation is not a rough base to be corrected later; it is a precision datum that everything else trusts. It must be in the right place, at the right level, and flat and true where the modules bear, to an accuracy far tighter than ordinary groundwork — because there is no later opportunity to correct it. And it must be right first time, because of everything we saw in the last two lessons: the modules are already being made (or made) in the factory to a frozen design, the just-in-time choreography assumes a base ready to receive them, and a foundation discovered to be wrong *after* modules start arriving is a programme catastrophe — modules with nowhere to land, a crane and a delivery sequence stalled, expensive rework on the one element that cannot be reworked once loaded.

This inverts the usual attitude to foundations. On a modular job the substructure is arguably the highest-risk, highest-precision site operation, deserving the best setting-out, the most careful concrete work and the most rigorous checking, precisely because it is the interface between the imprecise world of the site and the precise world of the factory. Teams often build in adjustability at the base — levelling bolts, shims, packing, or a fine screed/grout bed — to take up the last small differences, but adjustability is a safety margin on top of an accurate foundation, not a substitute for one. You cannot shim your way out of a foundation that is simply in the wrong place.

The binding design — the type of foundation, the bearing capacity, the structural connection of module to base, the allowable tolerances — is, as always, engineering that belongs to the structural engineer and the manufacturer's system, against the ground conditions and the National Building Code of India and local rules. What the designer must carry is the attitude: for modular, the foundation is sacred, and it must be accurate, ready and right before the first module moves.

Modular rarely sits straight on the ground in anything but the simplest buildings. More often the repetitive, cellular modular accommodation sits on top of a podium — a site-built structure at the base that does a job the modules cannot. The reason is architectural as much as structural: the regular grid that makes modules efficient (small, repeated, cellular rooms) is usually the *wrong* geometry for the ground floor, which typically wants large, open, column-free spaces — a hotel lobby, shops, a reception, parking, a double-height entrance. Stacking small modular cells down to the street would deny the building the open ground floor it needs.

The answer is a transfer structure: a podium, usually in in-situ or precast concrete or steel, that carries the loads of the regular modular grid above and *transfers* them down to a different, wider column grid below, opening up the ground floor. The podium becomes the stage on which the modular building performs — and its top surface becomes the foundation datum for the modules, inheriting all the precision demands of the previous section, now raised into the air. This is a very common modular configuration: site-built, open, bespoke podium at the bottom; repetitive factory-built modules stacked above; the transfer structure reconciling the two geometries.

The podium is also where the prefab spectrum decision becomes vividly physical. The building is deliberately *hybrid*: the parts that are large, open and one-off (the ground floor, entrance, shared spaces) are built conventionally or from panels, where site construction's flexibility is an advantage; the parts that are small, cellular and massively repeated (the rooms above) are volumetric modules, where the factory's repetition pays. Far from a compromise, this is often the *optimal* use of off-site — concentrate the modular discipline where repetition rewards it and keep bespoke freedom where the architecture needs it. Many of the most successful modular buildings are exactly this: a rich, site-built base and a disciplined, factory-built tower.

For the designer, the podium raises two coordination tasks. First, the transfer itself — how the modular grid above lands on the podium's columns and beams — is a significant structural design, owned by the engineer, that must be coordinated early because it shapes both grids. Second, the podium top is a precision foundation for the modules and must be set out and finished to the tight tolerances the handshake demands, often with cast-in bearing plates or pockets located by survey exactly where each module corner will land. The binding structural design of the transfer and the connections is the engineer's and the manufacturer's; the designer's role is to resolve the hybrid logic early and coordinate the two worlds that meet at the podium.

Even with the most accurate foundation, the site will never match the factory for precision — so the remaining difference must be designed into the joint. This is the essence of the tolerance handshake: you do not try to make the site as precise as the factory (you cannot, affordably); instead you design the interface between them to absorb the mismatch. The joint where module meets foundation, and where module meets module, is detailed with deliberate, engineered tolerance — room to take up the difference between a base set out to centimetres and a module made to millimetres.

This is a core idea from the grids-and-tolerance discipline of Module 5, made concrete at the most critical interface in the building. Tolerance is not sloppiness; it is the planned allowance for the fact that real, made things are never exactly their nominal size or in exactly their nominal place. A good interface detail states how much movement it can absorb in each direction — up/down (levelled on shims, bolts or a grout bed), side to side, and rotation — and is designed so that anywhere within that range the connection still works structurally, keeps out weather, and meets its fire and acoustic duties. The art is to give the site a realistic target it can actually hit, and a joint generous enough to forgive the rest, without the gap becoming so large that it is ugly, weak, or impossible to seal.

There is a deep design principle hiding here: locate the tolerance where it is cheapest to provide. It is far cheaper to build adjustability into a grout bed and a levelling detail at the base than to demand impossible accuracy from a concrete pour, and far cheaper to absorb movement in a designed joint than to discover it as a crack later. Good DfMA pushes the tolerance into the interface on purpose, so the precise parts stay precise and the imprecise parts are allowed to be imprecise within known limits. The handshake works because both sides know their role: the site promises a realistic accuracy, and the joint promises to swallow the rest.

The binding numbers — how much tolerance each connection actually provides, what the foundation must achieve, how the joint performs for structure, fire, acoustics and weather across its whole range of movement — are engineering, set by the manufacturer's tested system and the structural, fire and acoustic engineers for the specific product. The figures in this lesson are illustrative of the principle, never a specification. What the designer must own is the *thinking*: that a precise product and an imprecise base meet through a designed-tolerance joint, and that deciding where tolerance lives is a design decision with real consequences for cost, quality and buildability.

The handshake only works if both promises are kept, and keeping the site's promise comes down to setting out and survey. Setting out is the act of translating the design's coordinates onto the real ground — marking exactly where every grid line, every module corner, every bearing point and every cast-in plate must go. On a modular job this is a precision operation done with survey instruments (total stations, often GNSS), not a tape and a pencil, because the positions must be right to the tolerances the handshake assumes and because an error here propagates into every module above.

The discipline is to measure, mark, and then independently check before committing. The foundation or podium is set out from an agreed survey control; the critical points — especially the cast-in plates or pockets where modules land — are placed and then surveyed again to confirm they are where the design says, within tolerance, *before* the concrete is final and certainly before the first module arrives. This verification step is not optional on a modular job: because the modules are already made and the sequence already committed, the site's one chance to catch a setting-out error is *before* loading begins. A survey that finds a base ten centimetres out the day before delivery is a bad day; the same error found after three storeys of modules are stacked is a disaster.

This is where dimensional metrology — the science of measurement — stops being an abstraction and becomes the thing that makes modular work. The whole system depends on everyone measuring from the same control, to agreed tolerances, with checks that catch error before it is locked in. BIM helps enormously: the model carries the exact coordinates of every setting-out point and every interface, so the survey team sets out from the same single source of truth the factory built to, closing the loop between design, manufacture and site. The more the site and factory share one coordinated model and one control system, the tighter and more reliable the handshake.

For the designer, the practical stance is to respect setting-out and survey as first-class parts of the design-to-assembly chain, not site housekeeping: to provide clear, coordinated setting-out information (ideally from the BIM model), to specify the critical interfaces and their tolerances with the engineers and manufacturer, and to build the verification step into the programme so the foundation is proved right before it is trusted. The binding survey, the metrology, the tolerance specification and the foundation design all belong to the engineers, surveyors and the manufacturer's system; the designer's job is to understand the handshake well enough to design interfaces that can be set out and checked, and to insist that, for modular, the foundation is got right first time — because there is no second time.