Safety data sheet problems are almost never problems of absence. Ask any chemical distributor for an SDS and you will get one. The question that causes trouble is narrower: was it the current revision on the day that shipment left, and can you show it.
Why revisions are structurally hard
A revision arrives from a supplier, usually by email, sometimes as a portal notification, occasionally as an attachment to something else. It goes to a person. That person has to recognise it as a revision rather than a copy, find where the current version is stored, replace it, work out which customers received the product under the old revision, and distribute the new one to them.
Every step there is manual, and the whole chain depends on the first one — somebody noticing. Meanwhile nothing in the order or shipping process is aware that anything changed. The system that decides whether a shipment can go out has no relationship to the folder where the document lives.
That is the actual defect. Not the filing, the disconnection. Document control that lives beside the transaction rather than inside it can only ever be as good as the attention of whoever is watching it.
The four failures that show up at audit
Shipped against a superseded revision. The most common finding, and the hardest to argue with, because the supplier's own revision date is on record.
Customer never received the update. Distribution obligations do not end at the first shipment. If a revision is issued, downstream customers who received product under the old one generally need the new one, and proving you sent it is a separate problem from having sent it.
No sheet at all for a specific lot. Usually a new product or a substituted supplier where the material arrived faster than the paperwork.
The right sheet, no evidence of when. You have revision 5.0 today. The question is what you had in March. Without version history there is no answer, only an assertion.
What enforcement looks like
Three things convert this from a filing discipline into a control.
Versioning, not replacement. A new revision supersedes the old one rather than overwriting it, so the record of what was current on any given date survives. That is what answers the fourth failure, and it costs nothing except not deleting things.
A block at the transaction. If the current revision is not attached, the shipment does not go. This sounds aggressive and is the entire point: it converts a documentation task into a condition of shipping, which is the only mechanism that reliably survives a busy week. The same principle applies to restricted-use products, where licence verification belongs at order entry rather than at the dock.
Redistribution as a queue, not a memory. When a revision supersedes another, the set of customers who need the update is a query — everyone who received that product since the previous revision. Generating that list should be automatic; sending it should be tracked.
Why this belongs with classification
SDS control and DOT hazmat classification are usually treated as separate compliance areas, and operationally they fail together. Both depend on knowing what a product is before it ships, and both are commonly applied at the dock — the last moment, under the most time pressure, by the people with the least context.
Moving both to order entry changes the economics. Classification is known when the order is taken, the placarding requirement is known before the truck is loaded, and the document check has happened while there was still time to do something about it. How that works against batch records is on the chemical, ag and plastics page.