Site Measurement vs Shop Drawing Tolerance: Where Railings Go Wrong

A concrete slab edge on a cast-in-place tower is built to a tolerance measured in tens of millimetres. An aluminum extrusion is cut in the shop to roughly a millimetre. Somewhere between those two numbers is the reason railing packages generate site issues, and it is not because anyone did anything wrong. It is because two industries with radically different precision expectations have to meet at a physical joint, and the design has to absorb the difference. Projects that plan for that absorption install cleanly. Projects that assume the building matches the drawing spend their last month cutting parts on a balcony.

The tolerance gap, in real numbers

Exact allowable tolerances depend on the governing standard and the project specification, and they vary — always check the actual contract documents rather than a rule of thumb. But the orders of magnitude are consistent, and the orders of magnitude are the point:

• Cast-in-place slab edge location: commonly allowed to deviate by a couple of centimetres from plan position, sometimes more on a tall building where formwork is climbing.
• Floor-to-floor height variation: several millimetres per floor is normal, and it accumulates. Over 30 floors, guards that are all fabricated identically will meet cladding datums that have drifted.
• Slab surface flatness and levelness: variation across a single balcony deck is expected, and topping placement adds another layer of it.
• Building plumb: towers lean slightly and move seasonally. Over full height, the deviation from a true vertical line is real and measurable.
• Aluminum extrusion fabrication: cut lengths and hole locations held to roughly a millimetre in a competent shop.
• Tempered or laminated glass panel dimensions: also roughly a millimetre or two, and critically, glass cannot be modified after fabrication. A panel that is 8 mm too wide is scrap.

That last line is the crux. Aluminum is forgiving — a post can be trimmed, a rail can be cut, a bracket can be slotted. Glass is not. Once a panel is tempered and laminated, it is a finished object. This asymmetry should drive every decision about how much you rely on field measurement versus how much you design in adjustability.

Aluminum forgives a bad dimension. Glass just becomes an expensive rectangle in a crate.

Two strategies: measure everything, or design for adjustment

There are only two honest ways to bridge the gap, and most good projects use both in different places.

Strategy one is field verification: measure the as-built condition, then fabricate to it. This produces the tightest, cleanest result — panel gaps consistent, end conditions snug, no visible shimming. It costs schedule, because you cannot measure a floor until it exists and is stripped, and it costs flexibility, because every panel becomes floor-specific and mis-shipping one becomes a real problem. On a tower it also means fabrication cannot begin until the structure is well underway, which pushes the railing package later into the schedule.

Strategy two is designed adjustability: build tolerance into the system so it absorbs variation without measurement. Slotted base plates, adjustable spigots, telescoping or overlapping trim, oversized post caps that hide a range of positions, and end conditions detailed with a deliberate reveal rather than a tight butt joint. This lets fabrication proceed from the contract documents on a repetitive tower and lets shipments be interchangeable between floors. It costs some visual precision — the reveal that absorbs the tolerance is visible — and it needs the architect's agreement early, because a designer expecting a seamless line will not accept a 20 mm shadow gap discovered at first-floor review.

Practically, the good pattern on a repetitive tower is: adjustable systems for the typical repeating balconies, field-verified fabrication for the one-off conditions — podium terraces, amenity levels, stairs, and anywhere the guard meets an existing or irregular structure. Repetition rewards adjustability; uniqueness rewards measurement.

How field measurement should actually be done

Field measurement is not one person with a tape measure and a notebook, and treating it that way is how a project ends up with 40 balconies fabricated to a dimension that was recorded wrong on floor 12.

• Measure from established survey control, not from adjacent construction. If you dimension a guard off a cladding line that itself moves, you inherit that movement.
• Take heights from the finished floor elevation the guard will actually sit on — after topping, after pavers, after whatever else is going down. Guard height compliance is measured from the walking surface, so a guard set to the required height above the structural slab — commonly 1,070 mm on multi-unit residential balconies, though the governing figure varies with occupancy and province — ends up short once 40 mm of finished assembly goes in on top of it.
• Record at multiple points per bay, not one. A single dimension per balcony hides slab edge curvature, and curvature is exactly what produces a tapering gap that looks like a defect.
• Photograph every measured condition with the tape in frame. When a dimension is later disputed, the photograph settles it in a minute.
• Note what is not yet built. If the topping is not placed when you measure, that fact has to be flagged and the assumed thickness recorded, because it will be wrong on at least one floor.
• Have the same crew or the same protocol measure all floors. Two people with different habits about where they take the reading produce a data set with a systematic error in it.

Digital tools help — laser distance meters as a baseline, and on complex podium or retrofit conditions, laser scanning to produce a point cloud that can be brought straight into the shop drawing model. Scanning is genuinely worth the cost on irregular existing structures, where the difference between the drawing and the building is unknown and possibly large. On a repetitive new-build tower it is usually overkill.

What happens when the gap is not managed

The failure modes are predictable. Glass panels that do not fit and have to be re-ordered — a 6 to 12 week procurement hit landing at the worst possible point in the schedule. Site-cutting of aluminum, which destroys the factory finish at the cut and creates a corrosion initiation point that will show in a few years, particularly on coastal projects where marine exposure is already a design driver. Shims stacked under base plates to correct for a floor that is out of level, which compromises the anchor's load path and is the kind of thing that surfaces in a failure investigation. Visibly inconsistent panel gaps across a façade, which is a cosmetic defect that an owner will absolutely put on a punch list. And the worst outcome: guard height that measures under the required minimum once the finished floor is in, which is a code compliance failure, not an aesthetic one.

Every one of those is cheaper to prevent than to correct, and the prevention is a conversation at shop drawing stage about which strategy applies to which condition. When engineering and drafting sit in the same building as the fabrication floor, that conversation includes the person who will actually cut the metal — which is how a tolerance strategy stops being a note on a drawing and becomes something the shop is set up to deliver.

Common questions