Walk into any fabrication shop and look at the benches. There will be parts that are obviously what they are, and parts that are a tube with two bends and a bracket, which could belong to three different jobs running that week. The difference between those two situations is not skill. It is whether the shop decided, before cutting metal, how each piece would be identified and where that identity would live.
Identification is usually treated as administration, something that happens after the real work. In practice it is a process control. A part that cannot be identified cannot be measured against the right drawing, cannot be assembled in the right place, and cannot be replaced later without remeasuring the vehicle it came off.
A mark is only useful if it survives the journey. In a welded assembly that means four stages: the cut or formed blank, the welded subassembly, the finished and treated part, and the packed carton. Each stage has a different ability to destroy a mark. Cutting removes an edge. Welding covers a surface. Finishing removes a layer. Packing separates the part from its paperwork.
A shop that plans identification thinks about all four at once. A tag wired to a bundle is fine at the blank stage and useless after polishing. A marker pen line on a weld face is gone the moment the part is brushed. The question is never whether a mark can be applied. It is whether it can still be read at the point where somebody needs to know.
Exterior parts create a conflict. A titanium exhaust tip or a visible charge pipe is sold on its surface, and any mark on that surface is a defect. The identity still has to exist, so it moves: onto a flange face that will be hidden by a clamp, inside a bore, on a tab that gets trimmed, or onto the packaging with a matched count.
Where a permanent mark is required and no hidden surface exists, the practical compromise is a mark that can be removed without leaving a trace, combined with a record that cannot be removed at all. A photograph of the part with its tag, filed against the order, does more for later identification than a scribble that will be polished out.
The dangerous case is not a wrong part. It is two right parts that differ only in a way you cannot see. A left and a right with the same profile. Two wall thicknesses of the same diameter tube. A part for one model year with an extra boss that is not obvious until it is on the car.
In that situation, positional marking is not enough. The shop needs a rule that separates them at every stage, and the simplest rule is physical separation. Different bins, different colours of tag, different benches, different days. Human beings sort by appearance under time pressure, and two similar parts in one bin will eventually be mixed.
The mixed bin is the most common cause of a wrong part reaching a customer, and it is almost never a deliberate act. It happens when a job is paused, when a batch is part-finished at the end of a shift, when a trolley is reused without being cleared. By the time the parts are picked up again, the only thing connecting them to their order is a memory.
The fix is unglamorous. Bins that are emptied completely rather than topped up. A rule that a part leaves its bin only with its traveller. A check at packing that the count matches the paperwork rather than the pile. None of these require technology and all of them require somebody to care.
For a buyer receiving parts into their own assembly, the identification question moves downstream. A part that arrives correct but unmarked becomes an inventory problem: it cannot be issued against a build, it cannot be counted, and it cannot be returned if it is wrong.
This is worth raising before the order rather than after. Labels that carry the buyer's own part number, a consistent position on the carton, and a packing list that matches the physical count are all simple to arrange at the start and painful to retrofit across a shipment.
Custom metalwork has a long tail. A system fitted this year may still be on the vehicle in a decade, and the person servicing it then will have no memory of the build. What they will have is whatever the shop wrote down: the drawing revision, the material, the finish, the photographs taken at assembly.
That record is what makes a spare possible. Without it, replacing a section means measuring the installed part on the car, which is slower, less accurate and often impossible without disassembly. With it, the shop can make the same piece again, or a modified version of it, from the same starting point.
Most of this is decided by the buyer at the enquiry stage, usually without being asked. A few sentences in the purchase order will settle it.
None of these cost much to agree. All of them cost a great deal to discover later.
Identification failures rarely show up in inspection, because the part in front of the inspector is dimensionally correct. They show up months later, as a fit problem on one vehicle, or as a spare that does not match. By then the evidence has been polished, packed and dispersed.
Which is why the mark matters less than the habit behind it. A shop that treats identity as part of the process will get it right without being asked. A shop that treats it as paperwork will get it right until the week it is busy.
You can see the range of components this applies to on our other parts pages, and the smaller fittings on further fabricated parts are the ones most often confused with one another. Finished assemblies and replacements sit under fabricated assemblies, the welding that most marks have to survive is described at TIG titanium welding parts, and aluminium water tanks show why a hidden internal mark is sometimes the only option. How we organise a job from enquiry onward is set out on the company page.
