The flat pattern that came out wrong in both directions

A part developed without accounting for material stretch.

A flat pattern built on a single wrong assumption about how much a bend actually stretches the material can come out wrong in two directions at once rather than just one, because a real box needs bends running in more than one orientation to close up, and the identical flawed assumption feeds into every one of those bends regardless of which way they happen to run.

Why one wrong assumption about bend allowance affects every bend it touches

A flat pattern is built by working backward from a finished part's dimensions, subtracting from the naive sum of each face's outer measurement the specific amount of material every individual bend consumes through its own stretch, the bend allowance the earlier articles in this set have already described. That subtraction depends on a single underlying figure for the material and thickness in use, and if that figure is wrong, whether copied from the wrong material, calculated with the wrong bend radius, or simply mistyped, every bend built from it inherits the identical error, not just one bend in isolation. A box formed with bends running along its length and separate bends running along its width applies that same wrong figure to both sets of bends without any way of noticing partway through that one direction happened to come out closer to correct than the other, since the underlying mistake was never direction-specific in the first place, it was a single number wrong at the source, repeated faithfully everywhere it was used. This is a genuinely different category of mistake from a single miscut edge or a single wrong hole position, both of which show up as one isolated discrepancy on an otherwise correct part, since a bend allowance error instead rides along inside every dimension the flat pattern ever touches, quietly wrong in exactly the same proportion wherever it appears rather than concentrated in one obviously suspect spot. This is also why the error is so easy to miss while the flat pattern is being programmed, since the person entering the bend allowance is working from a single figure for a single material and a single radius, and nothing in that step distinguishes between a bend that will end up running along the part's length and a bend that will end up running along its width, the software applies the same formula either way without ever pausing to ask whether the figure it has been handed is actually the right one for the job in front of it.

The tailoring-seam-allowance comparison

A dressmaker cutting fabric for a garment works from a seam allowance, a fixed extra margin added around each cut piece specifically to be consumed by the stitching itself, and if that seam allowance figure is subtly wrong, every seam sewn using it inherits the identical error, the shoulder seams as much as the side seams. The finished garment does not come out wrong along only one direction, it comes out too narrow across the chest and too short in overall length simultaneously, since the same flawed assumption about how much fabric each seam actually consumes was applied uniformly to every seam regardless of which direction that particular seam happened to run through the garment. Nothing about the stitching itself was done incorrectly, every seam was sewn exactly to the seam allowance the dressmaker was working from, and the finished garment's error traces back entirely to that one number being wrong at the source rather than to any inconsistency in how carefully each individual seam was executed. A dressmaker who later lets out one seam to claw back some of the missing width only patches the single dimension that happened to be checked first, while the shoulder seams and the hem, sewn from the same flawed allowance and never separately inspected, carry the identical error forward into the finished garment untouched. A sheet metal flat pattern built on a wrong bend allowance figure fails in exactly the same uniform, direction-independent way, and a part quietly reworked along one edge to bring it back into tolerance still carries its uncorrected bend allowance into every other bend nobody thought to re-check.

Why a single flat pattern error is invisible until the part is folded

The flat pattern itself, before any bending happens, offers no visible sign that anything is wrong, since a flat sheet of metal with the wrong bend allowance baked into its dimensions looks and measures no differently from one built on the correct figure, the entire discrepancy hidden inside a number that only reveals itself once the part is actually folded and its finished dimensions can be checked against the drawing. This is exactly the trap the earlier article in this set on flat pattern length already warned about, an error that only becomes visible at the very last stage of the process, well after the flat sheet has already been cut, and correcting it means going all the way back to the original bend allowance calculation rather than simply adjusting the finished part directly. Adjusting the finished part instead of the underlying figure is the tempting shortcut and also the one most likely to fail quietly, since trimming a single out-of-tolerance edge on one bad part does nothing to the bend allowance figure still sitting uncorrected in the file, and the very next part cut and folded from that same file arrives with the identical error still fully intact, waiting to be discovered all over again. A flat sheet passed on for bending typically travels with nothing more than its cut profile checked against the drawing, a step that confirms the outer dimensions are correct while saying nothing about whether the bend allowance folded invisibly into those dimensions is right, so a sheet can clear every inspection it is likely to face on the way to the press brake and still be carrying an error that will not surface until well after the sheet itself has already been formed and consumed.

The number that matters here

A bend allowance figure wrong by even a small amount compounds across every bend it is used in, so a box with bends running in two different directions can come out wrong in both its length and its width at once from a single incorrect assumption, the total error in each direction scaling directly with how many bends that direction actually contains. A direction with three bends in it accumulates three times the single-bend error, while a direction with only one bend shows only a third of that same total, which is exactly why a genuinely wrong bend allowance can leave one dimension of a finished part looking only mildly off while the other looks badly wrong, despite both errors tracing back to the identical root cause.

My key error with this

The rear body panels came back and did not line up, not by a large amount anywhere in particular, but by a millimetre here and a millimetre there in a way that accumulated across the assembly until the last holes did not meet anything. The cause was upstream in the flat pattern, in an assumption about how much material a bend consumes that was applied identically to every bend regardless of direction, so each individual error was small and every one of them pointed the same way. Waiting for corrected parts meant a fortnight, which the schedule could not absorb, so we bent one panel back slightly by hand and drilled additional holes to pick up what we could, which worked and is not something I would like to defend. What replaced the belief is that a bend allowance is an assumption rather than a constant, that an error in it is never isolated because the same figure feeds every bend on the part, and that on any developed sheet part the cheap check is a single test piece before the whole set is cut.