A finished grille wall. Everything that could go wrong with joint alignment has already happened by the time this photograph was taken.

The first three panels of a grille wall go up perfectly. The fourth is a fraction out. By the twelfth, the flutes no longer line up with the panel the installer started from, and the wall has developed a lean that nobody can explain and nobody can fix without taking it apart.

That progression is not a workmanship failure in the ordinary sense. It is tolerance accumulation, and grille panels are unusually good at exposing it. The PS wall panel range carries an interlocking profile on the long edges, which means every panel's position is set by the panel before it rather than by an independent fixing to the wall.

This piece works through where the error comes from, why a fluted face makes it visible, and what can actually be done about it on site.

Stand close to a single joint and run a finger along the flute line. Half a millimetre of step is almost impossible to feel. Step back three metres and it is obvious.

That is the first thing to understand about grille walls. The tolerance that matters is not the one you can measure at the joint. It is the one the eye reads across the whole run.

The flute lines are continuous across the panel face. When a joint is out of register, the line breaks, and the eye finds the break immediately.

Now do the arithmetic on a typical run. Take a wall of twenty panels. If each joint is out by half a millimetre in the same direction, the last panel sits ten millimetres off the line set by the first. On a panel with a ten-millimetre flute pitch, that is a full groove out of register. Nothing dramatic happened at any individual joint. The wall simply walked away from its own starting point.

The same arithmetic works in the installer's favour when the errors are random rather than systematic. Random error partly cancels. Systematic error accumulates in one direction and never recovers. Almost every badly aligned grille wall is a systematic error story, which is good news, because systematic errors have causes that can be found.

Four sources account for most of it, and they are not equally common. Ranking them by how often they turn out to be the culprit is worth doing, because installers tend to blame the last one on the list and it is usually the first.

Viewed at a shallow angle, a systematic step at the joints reads as a single continuous fault rather than a series of small ones.

The first row is the one that gets missed. A grille profile is formed into the board, and the decorative pattern is applied to the face. The distance between the interlocking edge and the first flute is a manufactured dimension with its own tolerance. If that dimension drifts through a production run, every panel carries the same offset, and every joint steps in the same direction. No amount of careful installation fixes it, because the installer is being handed a stack of panels that are individually consistent and collectively wrong.

The second row is more common on site than anyone admits. An interlocking profile has to be engaged along its full length. Pressing a panel home from the middle seats the centre and leaves both ends fractionally proud. The panel looks fitted. The joint is not closed.

A flat panel with a wood grain print forgives a great deal. The pattern is irregular, the eye has no reference line to follow, and a step at a joint disappears into the noise of the grain. A grille panel removes every one of those hiding places.

The shadow line is what the eye tracks. It is also what makes a misaligned joint impossible to overlook.

The flutes form a regular rhythm, and regularity is exactly what the human visual system is built to detect breaks in. A single line that steps by half a millimetre interrupts that rhythm, and the interruption is picked up before the viewer has consciously registered anything about the wall at all.

There is a second effect, and it is the one that turns a small problem into a complaint. The flutes cast shadows, and the shadow is what gives the wall its depth. At a joint that is out of register, two shadows meet at an angle instead of running as one line. Under directional light, that meeting point becomes the darkest spot on the wall. The defect does not merely exist. It draws attention to itself.

This is worth saying plainly to anyone specifying a grille wall for the first time. The product is less forgiving than a flat panel, not because it is made to a lower standard, but because it presents a regular geometric pattern that makes every deviation legible. The tolerance has to be managed rather than assumed.

Once the second panel is locked into the first, the pair behaves as a single rigid unit. Correcting anything downstream means releasing the interlock, and releasing an interlock that has already been pressed home risks damaging the edge that carries the finish. That is why the first panel is the only one that can be adjusted at no cost.

A plumb start panel produces vertical flutes and parallel joints. Every later panel inherits whatever the first one established.

Set the start line with a spirit level, not by measuring across from the adjacent wall or up from the floor. Floors are rarely level and corners are rarely square. A start line taken off either one imports that error into the wall and then multiplies it by the number of panels.

Check the first panel in two directions before anything else goes on it. Plumb along its length, and square across its width. Then check the flute line against the level rather than against the panel edge, because on some profiles the flute is not perfectly parallel to the interlock, and the flute is what the viewer will judge.

Checking against the panel immediately next to the one you just fitted tells you almost nothing, because adjacent panels can both be wrong in the same direction. Check back against the start.

On a long run, the only reliable check is against the start panel rather than against the previous joint.

Three practical methods, in order of how early they catch a problem. A laser line set parallel to the start panel gives a continuous reference across the whole wall and can be checked at any point in the run. A string line does the same job on a shorter wall and costs nothing. And on a small job, simply sighting along the wall from one end at a shallow angle will show a systematic step as a single continuous fault, which is easier to spot than a series of individual ones.

Whatever method is used, check every fourth or fifth panel rather than only at the end. A systematic error caught at panel five costs a few minutes. The same error found at panel twenty costs the wall.

There is a simple test that separates a factory fault from an installation fault, and it takes about two minutes.

Lay three panels from the same bundle side by side and compare where the first flute sits relative to the interlocking edge.

Take three panels from the same bundle and lay them flat, face up, aligned on their interlocking edges. Compare the distance from that edge to the first flute on each. If the three measurements agree and the wall is still stepping, the problem is on site. If they disagree by a visible amount, the bundle carries the fault, and no installation technique will correct it.

That test is also the one to run before accepting a delivery, on a sample pulled from more than one bundle. Registration between the pattern and the profile is a manufacturing control, and it either holds across a run or it does not. Asking a supplier for the tolerance on that dimension before the order is a more useful question than most of what appears on a typical quotation.

Grille joints do not open up because installers are careless. They open up because an interlocking system passes every panel's position down the line, and a regular flute pattern makes the accumulated result visible. The mechanism is predictable, which means it is also manageable.

Set the first panel plumb, check it in two directions, sight back to it every few boards, and keep the whole wall to one production run. Do those four things and the wall stays in register. Skip them and the fault will not appear at the start, which is what makes it expensive.