Setting Out Terrazzo Divider Strips

The Grid Is Ordered Long Before the Terrazzo Is

A lobby measuring 23.4 m by 8.4 m — near enough 77 ft by 27 ft 6 in, about 2,120 sq ft of finished floor — will swallow somewhere north of two hundred metres of divider strip. None of that is stock you pull off a rack. Divider strip is a manufactured section ordered in a named alloy, at a stated top width, at a depth that has to match a topping thickness nobody has poured yet, and it arrives in sticks cut to the supplier's standard length. On a fit-out programme the strip order routinely has a longer lead time than the terrazzo itself, which means the panel layout has to be settled and approved while the slab is still being power-floated by somebody else.

That is the awkward part of this job. The setting-out drawing is not a record of what you did; it is a purchase document, issued and signed off before a single line is chalked. Once the architect has approved a strip layout, the grid stops being yours. Moving a line 150 mm to clear a lift threshold after the strip has been cut to length turns a five-minute decision into a variation, a re-cut and a fortnight.

So the sequence runs backwards from the way the floor gets built. Survey the substrate and find the lines the slab has already drawn. Confirm which terrazzo system is actually specified, because that decides what a strip is for. Set the grid against the building's structural lines rather than its finishes. Count the strip runs, count the exposed edges at every other floor covering, price the movement-joint assemblies separately, and only then let anyone put a pour date in the programme.

A Strip Is Not the Same Object in Every System

In the cementitious systems the divider strip is a crack-control device wearing a decorative hat. A cement-and-chip topping shrinks as it cures, and shrinkage that has nowhere to go arrives as a random crack across the middle of a panel. Strips give that movement a straight, metal-edged place to arrive at, which is why panel size and panel proportion are constrained in cementitious work and why the strip layout is a technical submission rather than a graphic one. BS 8204-4, Screeds, bases and in situ floorings — Part 4: Cementitious terrazzo wearing surfaces: Code of practice, and the National Terrazzo and Mosaic Association's system specification guides both address panel sizing; read the current edition rather than the figure you remember from the last job, because the limits differ between systems and between the two documents.

In a thin-set resin system — epoxy or polyacrylate — shrinkage is small enough that crack control is not the strip's job at all. There the strip is a design line, a colour change and a screed rail, and panels can run considerably larger than any cementitious grid would tolerate. That difference is the single biggest swing in this package. Halving the panel spacing in both directions on the lobby below takes its strip length from 226 m to a little over 420 m, and the difference lands entirely on your material cost. If the specification is still open when you are asked for a number, run the quantity twice and say which assumption the price sits on.

One thing does not change with the system: the strips are the screed rails. The topping is laid to their tops and ground back to them, so a strip top is the finished floor level, in metal, permanently. Setting divider strip is therefore a levelling operation carried out to a datum, not a decorative one carried out to the underbed surface, and every millimetre of error you leave in a strip line is a millimetre the grinder has to take out of the surrounding floor.

Alloy is a compatibility question, not a taste question. Brass is a copper-zinc alloy and a portland cement matrix is strongly alkaline; strip suppliers and NTMA guidance both flag the risk of discolouration and attack where the two sit together, which is why white alloy and plastic strips are so common in cementitious work and why brass appears more freely in resin systems, where the matrix chemistry is different. Put the question to the strip supplier in writing against the matrix that is actually specified, and get a sample cast in that matrix if the client is buying the look rather than the line.

Set the Grid to the Building, Not to the Plaster

Lobby walls are late work and they lie. Feature panelling, glazing lines, the reception joinery and the lift architraves all move during a fit-out, sometimes by more than the tolerance you are working to, and a strip grid squared off a wall that has not been built yet is a grid that will be wrong. Take your control lines off the structural grid — column centres, established from the setting-out engineer's marks — and offset from there. With three bays of 7.8 m down the length of this lobby, the grid lines themselves are the strongest datum on the floor and they are the only ones that will not have moved by the time the topping goes down.

Then decide what the grid must hit, because a divider strip is a straight metal line in a polished floor and the eye finds it from across the room. Lift thresholds, the centrelines of the main entrance doors, the edges of the entrance mat well, the face of the reception plinth and the boxes around the columns are all places where a line either lands correctly or visibly does not. A strip running 80 mm off a lift sill is the detail everybody notices and nobody can unsee, and it is far cheaper to shift the whole grid at the drawing stage than to argue about it on a floor that has already been ground.

Squareness is the next thing to spend. Measure both diagonals and check the long walls against a stringline; a lobby out by 15 or 20 mm over its length is unremarkable. That error has to go somewhere, and the choice is only ever which panel absorbs it. Push it into the perimeter panel against a wall, where a base or a skirting will cover the taper, and never into a panel whose edge runs along a threshold or the front of a joinery run, where it sits beside a straight machined line and reads as bad work.

With the fixed points settled, the arithmetic falls out. Quartering the 7.8 m structural bay gives strip lines at 1.95 m, and dividing the 8.4 m width into four gives 2.1 m, so the field lands as forty-eight panels of about 1.95 by 2.1 m — roughly 6 ft 5 in by 6 ft 11 in — which reads as a deliberate grid rather than an arbitrary one because it is derived from the frame. Panel proportion matters as much as panel area in cementitious work: a long, thin panel shrinks predominantly along its length and concentrates that movement at the ends.

Curves and inlays are a separate conversation and a separate line on the order. Radius strips are formed off site to a stated radius and quoted per piece, not per metre, and a heavy-top section cannot be bent to a true arc on the floor with anything you own. A logo, a compass rose or a border in a second colour adds strip length that a panel-grid calculation never sees, so measure the developed length of every feature line separately and add it in deliberately.

How Deep Is the Strip You Just Ordered

A divider strip is ordered by five things: alloy, finish, top width, gauge and depth. The first four are visible on a sample board and get argued about at design stage. Depth is the one that goes wrong quietly, because nobody sees it until the strips are on site and the underbed is already screeded.

Depth is the finished topping thickness plus whatever embedment the system needs below it, which means the topping thickness has to be fixed and confirmed against the approved section before the order goes out. In a sand cushion floor, the strip stands on the underbed and rises to finished floor level, so an underbed screeded 6 mm thicker than the section says leaves every strip on the job 6 mm short — and there is no way to recover that other than lifting them or drowning them. In a thin-set resin floor the strip depth is effectively the nominal topping thickness, and one ordered deeper than that sits proud of the trowel and has to be ground down as a separate operation.

Setting them is a levelling job. Establish finished floor level from the building datum, not from the slab, transfer it around the lobby, and set the strips to a laser or a taut line with the anchoring lugs bedded in the stiff paste or adhesive the system calls for. Check straightness along the whole run rather than between adjacent panels: a 23.4 m line that wanders 10 mm in the middle will be read against the mullion spacing of the glazed façade behind it and will lose that comparison every time.

Before the strip goes on order, open a check hole through whatever is already down and measure the build-up rather than trusting the section drawing. Slab levels, an unexpected screed from a previous trade, or a depressed slab at the mat well that was never poured to the depression shown all change the depth you need, and they change it across the entire order. This is a fifteen-minute exercise that has saved entire packages.

The Grinder Arrives and Edits Your Work

Grinding is where setting-out errors become permanent. The floor is rough ground, the voids are grouted, and the surface is taken down through progressively finer passes — and the strip tops go down with it, because they are part of the surface. A strip left standing proud gives the machine two options and neither is good: flatten the strip, taking thickness off a section that was never meant to lose much, or take the surrounding floor down to meet it, which on a thin-set topping means grinding a long way into a layer only a few millimetres deep to begin with. A strip set low is worse, because it never fully surfaces on the first cut. The line appears in some panels and breaks in others, the cure is to grind the whole floor down to the lowest strip on the job, and until somebody does that it is a ghosting line the client finds in the first week and never stops finding.

Top width and gauge earn their price at this stage. A heavy-top section survives repeated regrinding over the life of a lobby floor and holds a crisp line; a light section, ground twice in fifteen years, gets thin. That is an argument to have with the specifier while the sample board is on the table, because it is invisible in a photograph and obvious underfoot a decade later. The polished finish itself carries the other check nobody in the strip conversation raises: the slip performance of a polished terrazzo entrance floor is measured by ANSI A326.3, the standard test method for dynamic coefficient of friction on hard surface flooring, and an entrance bay that gets wet in November is exactly where that number gets tested in anger.

Grinding terrazzo is also silica work. OSHA 29 CFR 1926.1153, Respirable Crystalline Silica, governs it, and wet grinding with the slurry contained is the normal control — which in an occupied or partially handed-over lobby means water management, edge protection around the wet area and a conversation with the building manager that is easier to have a fortnight early. The programme consequence is real: the strip layout you approved determines how many linear metres of edge the machine has to work up to by hand, and hand edging is the slowest hour on the floor.

What Actually Goes on the Order

The order is not two hundred and forty metres of brass. It is a schedule of sections against the approved drawing, each with its own alloy, finish, top width, gauge, depth and stick length, and each with a run length that somebody can check against the layout. Splitting it that way is what makes a delivery verifiable on the tailgate rather than three weeks later when a run comes up short.

Add spare stock deliberately, in the same alloy and the same batch, and keep it labelled with the drawing reference after handover. Terrazzo lobbies get altered — a new security line, a reception desk moved, a door blocked up — and the repair that matches is the one cut from the offcut box rather than reordered five years later against a finish that has quietly changed.