Checking a Built Stair: Riser Tolerance and What Guards It

A flight you did not cut, and half an hour to say something about it

The stair went in during the first fix, before the screed and before anybody had chosen a floor covering. The screed followed in May. The engineered board went down in the hall in July, and the carpet fitter came in August. Now it is September, a warranty inspector or a purchaser's surveyor is standing at the bottom with a tape, and the question is no longer what the flight was set out to. It is what the flight measures today, with everything that has been added to both floors since the strings were cut.

Two things get written up on stairs more than anything else, and they are unrelated defects with unrelated remedies. One is that the risers are not uniform enough. The other is that the guarding is wrong — too low, too open, or fixed to something that will not hold it. The first is arithmetic performed on a list of measurements. The second is a comparison against a document that changes depending on whether the building has been handed over and who is currently walking up the flight. Both are settled in an hour, with a tape, a straightedge, a level and something round.

None of what follows is about setting a stair out. Lifting the total rise off a datum, dividing it, deciding the riser count and cutting the strings all happen before the flight exists, and that work belongs to the page on building stairs. This page starts where that one ends: the flight is in, the finishes are on, and somebody wants a defensible answer.

Where you hold the tape decides what the flight measures

A riser height is the vertical distance from the leading edge of one tread to the leading edge of the next, taken on the surfaces that people actually walk on. Every clause of that sentence is somewhere to get a different number. Measure to the substrate under an 18 mm board and every riser except the bottom one comes out right while the bottom one comes out wrong. Let the tape lean back over the nosing and you have added the cosine of the lean to a dimension whose whole tolerance is nine millimetres. Take a reading in the middle of a tread that bows, and what you have written down is the tread's deflection under its own span.

So fix the method before the first reading and do not vary it. One instrument for the whole flight, one position across the tread width, one operator, one continuous pass. Mixing a laser at the top with a folded rule at the bottom manufactures a difference that belongs to the instruments rather than to the stair, and the rule you are testing against is a difference. Carpet deserves its own decision: pile compresses under a foot and under a tape, so a flight already carpeted cannot be measured to a repeatable surface, and a flight about to be carpeted should be measured now with the intended underlay and pile thickness written down beside the figures as an assumption rather than a fact.

1. Lay a straightedge or throw a laser line down the nosings first and confirm the flight is a plane. A twisted string shows up here, and it invalidates every reading taken afterwards.
2. Number the risers on masking tape, bottom to top, so no reading can later be attributed to the wrong step.
3. Take riser one from the finished lower floor to the leading edge of the first tread, tape plumb and clear of the nosing overhang.
4. Take every intermediate riser nosing to nosing, on the same vertical line across the width of the flight.
5. Take the last riser from the leading edge of the top tread up to the finished upper floor, including any landing nosing that projects over it.
6. Repeat one riser at the far end of the tread width. A difference here is a flight out of level across its width, which is a separate defect with a separate remedy.
7. Write every figure down before deciding anything. The two the rule uses are the largest and the smallest, and neither can be identified until the list is complete.

Three-eighths, three times over

Riser spread is one of three uniformity limits sitting in the same part of the code, and the other two get skipped because nobody thinks to bring a second tape. IRC R311.7.5.2 applies the same three-eighths to tread depth, deepest against shallowest across the flight. IBC 1011.5.4 states riser and tread uniformity together in one sentence. And IRC R311.7.5.3 puts the same figure on nosing projection, which matters more than it sounds: the nosing is what a descending foot actually meets, and an inconsistent projection changes the effective going without changing any dimension that appears on a drawing.

The figure itself is not universal, and assuming it is the fastest way to sign off the wrong thing. OSHA's construction stairway rule at 29 CFR 1926.1052(a) is tighter than the model codes — riser height and tread depth uniform within each flight, with variation capped at a quarter of an inch across the stairway system — and it governs the stair the trades are climbing while the building is still a site, which may well be the same physical flight in front of you. Approved Document K in England and Wales does not express the requirement as a tolerance at all: the risers in a flight are to be equal, with the workmanship allowance a matter for the standard the stair was made to, BS 5395-1 for straight flights and winders. Australian work is judged under the National Construction Code, with AS 1657 covering fixed industrial access.

So the check has two questions before it has any measurements. Which document was this flight built under, and which edition has the authority having jurisdiction adopted? And is this a finished building or a live construction site? The same set of readings produces two different verdicts, and only one of them is the one being asked for.

The checks a passing spread proves nothing about

Headroom is the one that ends flights. It is measured vertically from the sloped plane touching the tread nosings, not from a tread, and both model codes put the floor at the same place: IRC R311.7.2 and IBC 1011.3 both require 2032 mm, six feet eight inches, with Approved Document K working to two metres over the pitch line. Take it at the worst point rather than at the entry, which on almost every domestic flight means where the pitch line passes under the trimmer or the bulkhead, and take it plumb — a tape angled to reach a convenient soffit reads long.

Handrail height carries its own datum and its own band. IRC R311.7.8.1 and IBC 1014.2 both put the gripping surface between 34 and 38 inches above the nosing line, which is a range rather than a minimum, so a rail can fail for being too high as readily as too low. The two that actually fail on site are continuity and grip. A rail that stops at the last baluster instead of running the full flight has failed continuity; a 100 mm square newel-section timber called a handrail has failed grip, because the grip-size provisions describe a cross-section a hand can close around, not a shape a hand can rest on.

None of these are downstream of the riser check. They are parallel to it, and a flight that passes on spread by a comfortable margin can still be written up for a soffit, a rail that dies into a wall, or a width measured over a bulky moulding. Do them in one visit and record them as separate lines, because they will be remedied by different people.

What guards it is decided by the drop, not by the stair

The trigger for a guard is a height, not a step count. IRC R312.1.1 and IBC 1015.2 both require guarding where the walking surface sits more than 30 inches, 762 mm, above the floor or grade below, measured at any point within 36 inches, 914 mm, horizontally of the open edge. That second dimension is the one that gets missed. A landing only 500 mm above a lower deck can still need a guard if that deck itself falls away within three feet of the edge, because the code measures the drop that a person could reach, not the drop directly underneath.

Height then splits by code family. IRC R312.1.2 sets 36 inches as the residential minimum; IBC 1015.3 sets 42 inches. On a flight, both measure it from the line connecting the leading edges of the treads rather than from the tread surface, and that single detail produces the most common measuring error on an open string: stand on a tread, run the tape up to the rail, and the figure is generous by most of a riser, because your datum sits behind and below the code's sloped one.

There is an interaction between the guard and the handrail worth checking rather than assuming. Where the top of a handrail serves as the top of a stair guard, the model codes permit that combined element to sit within the handrail band rather than at the full guard height, which is why a compliant open-string balustrade can measure less than the landing guard immediately above it without either being wrong. The permission is conditional and the conditions differ by edition, so read the one that has been adopted rather than the one in your head.

Elsewhere the datum survives and the numbers do not. Approved Document K tabulates guarding heights by building type and by location — a flight in a dwelling, a landing, the edge of an internal floor, a balcony — and they are not all the same figure, so the table is the answer rather than any one value remembered from it. The National Construction Code likewise measures a barrier on a stair from the nosing line, which means a rail that measures correctly with the tape stood on a tread can still be short of the requirement in either jurisdiction.

The half of the guard check that a tape cannot reach

A guard has to resist force as well as satisfy geometry, and this is the part nobody performs on site because it genuinely cannot be performed on site. In the United States, ASCE 7, Minimum Design Loads and Associated Criteria for Buildings and Other Structures, sets the design loads for handrail and guardrail systems in its Section 4.5: a uniform load along the top of the system and a concentrated load applied in any direction at any point along that top, the two not applied simultaneously, plus a separate load on intermediate rails, balusters and panel fillers taken over a small square area. In the United Kingdom the equivalent horizontal line loads on barriers come from BS EN 1991-1-1 with its UK National Annex, graded by the use of the space so that a domestic stair and a grandstand are not asked for the same thing. Australia and New Zealand work to AS/NZS 1170.1.

What is available to a visiting inspector is evidence rather than proof. Where is each post fixed, and to what — a newel bolted through a trimmer is a different structure from one screwed to the face of a stringer, and the difference is the whole capacity. Are the fixings the ones on the manufacturer's detail, in the quantity that detail shows? Does a glass balustrade have a specified channel with a documented interlayer, or a shoe and an assumption? And on commercial work, was a calculation produced at all, because there the calculation is a deliverable and its absence is itself the finding.

The honest field test is a hand and a note, and it must be described as exactly that. Push the top rail firmly outward at mid-span between posts and again at the top newel, and watch the base rather than the rail — flex in a timber rail means very little, rotation at a post base means the fixing is carrying the moment badly. If it moves, record where it moved and how far. That is not a load test and should never be written up as one, but a guard that visibly rotates under a shove is not carrying a defensible concentrated load either, and saying so in those terms is both true and useful.

Which failures are a morning's work, and which ones are the flight

A single outlier at one end is the cheap failure and it is by far the most common. The bottom riser short, or the top one short, almost always means a floor finish arrived after the strings were cut: the lower floor gained a build-up nobody fed back, or the stringer was dropped by a tread thickness that then changed on the schedule. Packing the string base, re-cutting it, or bringing the bottom riser up with the same material the floor gained will all resolve it, and the arithmetic for that adjustment belongs to the page on building a flight rather than to this one.

A spread distributed across the whole run is a different animal. It means the layout accumulated rather than slipped, there is no single step carrying the error, and there is therefore nothing to adjust. On a housed and wedged timber string that is a new flight, and pretending otherwise wastes a week before arriving at the same place. On a cut string with loose treads it is occasionally recoverable by re-housing two or three steps, and it is rarely worth what it costs. Name which of the two you have found, because a report that says only that the risers are uneven gets priced as the expensive one every time.

Guarding failures usually run the other way. A height shortfall on an open string can often be taken up in the rail rather than in the posts. An opening failure is more infill in the same span, which is joinery. A load concern is a post fixing, and post fixings are reachable from below more often than people assume once a soffit is opened. The expensive guarding failure is the one where the posts land on nothing at all — no trimmer, no blocking, no bolt through anything structural — and that is a floor-structure job wearing a balustrade, which is why the fixing is worth finding before the height is argued about.

What the record has to say, and how long it stays true

Write down the list, not the conclusion. Every riser in order with its measured height, then the tallest, the shortest, the spread between them, and the permitted variation you compared it against with the code and edition named beside it. For the guard: the height with its datum stated explicitly as the nosing line or the walking surface, the sphere sizes used and the specific gaps each was applied to, and what you could see of the post fixings. A verdict with no readings underneath it cannot be re-checked by anybody, which means it has to be re-done from scratch the moment it is questioned.

Then date it and describe the state of the finishes on the day, because a stair check is only true for that day. Between a visit and a handover, a tiler can add twenty millimetres at the foot of the flight and move the bottom riser without going near the stair, a carpet fitter can put thick underlay on the treads and nothing on the lower floor, and a balustrade can come off for decorating and go back with two of its four post bolts. Say so in the record in those words. The next person to measure this flight will be doing it because something moved, and the list you left is what tells them what moved.