half your specifications are wrong, and everyone works around them

Ask any experienced team, honestly and off the record, and they will tell you the same thing. There are specifications on their prints that nobody really holds, tolerances tighter than the function requires, callouts that made sense to someone once and have been quietly ignored ever since. The document says one thing. The floor has decided, through years of experience, what actually matters and what does not. That gap between the specification and the reality is one of the most common and least discussed risks in manufacturing.

the two documents

In a lot of operations there are effectively two sets of requirements. There is the official one, printed on the drawing and referenced in the control plan. And there is the real one, the shared understanding of which dimensions actually affect the product and which are there for reasons nobody remembers. Experienced people run to the real one. They know the tolerance on this feature could double without any effect on function, and that this other feature, loosely toleranced on the print, is the one that actually causes field problems. They are usually right. And the fact that they are running to an unofficial standard is a problem waiting to surface.

how specs get wrong in the first place

Specifications drift away from reality through ordinary means, not negligence. A tolerance gets copied from a similar part without anyone checking whether it fits this function. A designer applies a default because it is safer to be tight than to think hard about what the feature really needs. A requirement that mattered on the previous revision survives onto the next one after the reason for it is gone. Over time, a print accumulates a layer of requirements that do not reflect what the part actually needs to do, and everyone downstream is left to sort out which ones are real.

the cost of pretending they are all right

Living with wrong specifications is expensive in both directions. Over-tight tolerances that do not matter drive real cost, scrapping and reworking parts that would have functioned perfectly, chasing capability on characteristics that were never critical. Meanwhile the loose or missing requirement on the feature that actually matters lets real problems through. And the whole arrangement depends on undocumented judgment about which is which, judgment that lives in experienced heads and disappears when they do. A new engineer who takes every spec at face value will waste enormous effort holding tolerances that do not matter while missing the ones that do.

fixing the print instead of working around it

The healthier path is to close the gap rather than manage it informally forever. When the floor is quietly ignoring a requirement, that is not a discipline problem to crack down on. It is information. It usually means the requirement is wrong, and it is worth the effort to confirm that and correct the document, so the print says what the part actually needs. Where a loose spec is letting real problems through, tighten it deliberately, based on function. The goal is a specification you can actually trust, where holding the print means making a good part and the unofficial second standard is no longer necessary.

trust the print, or fix it

The principle underneath all of this is simple. A specification should mean something. If your people cannot run to the print as written, because they know parts of it are wrong, then your document is not controlling your product, and the real control is a set of unwritten understandings that no audit will ever see and no new hire will inherit intact. That is a fragile way to run. The work of getting specifications right, tolerances that match function and requirements that reflect what the part actually does, is unglamorous and easy to defer. It is also what separates a print you can trust from a print everyone has quietly agreed to ignore.

Where is the gap widest between what your prints require and what your team actually holds? And how much of that gap is written down anywhere?