Keeping a Bates sequence honest across a set of files

Bates numbering exists so that a page can be pointed at unambiguously. "The third page of the fourth exhibit" invites argument. "ACME-004471" does not.

That is the whole value, and it rests on one property: every page in the production has a different number, and no number is used twice. Everything else is formatting.

Where sequences break

The failure is almost never in the stamping. It is in the bookkeeping between documents.

Restarting by accident. Twelve files, each numbered starting at 1. Now there are twelve page 000001s and the numbering conveys nothing. This happens when a tool defaults to 1 and a person processes files one at a time on a Friday afternoon.

Overlapping ranges. Two people number two halves of a set, both estimating where the other will finish, and the estimates overlap by forty pages. Worse than restarting, because it looks correct until someone cites a number that resolves to two pages.

Outgrowing the padding. You chose four digits, expecting a few thousand pages. At page 10,000 the number becomes five digits, which sorts differently in every file browser and text search you will subsequently use. ACME-9999 and ACME-10000 do not sort adjacently as text.

Practical defences

• Pick the padding for the set you might end up with, not the one in front of you. Six digits costs nothing and covers a million pages. Four digits will eventually cost you an afternoon.
• Record the last number used. A sequence continued from a written-down figure is reliable. One continued from memory is not.
• Number a whole production in one pass where you can. Batch processing exists precisely so the sequence has a single origin.
• Check the first and last number after every run, and keep them. If your tool reports them, it is saving you the job of opening the file to look.

What it does not prove

A Bates number is drawn onto a page. It is not a seal, a signature, or evidence of anything. It does not show that a page is unaltered, that it came from where it says, or that it was produced when it says.

If you need to show that a document has not changed since you handed it over, that is a different mechanism — a checksum, recorded separately, that anyone holding the file can recompute. Also not proof of authorship, but it does answer the question "is this the same file", which numbering cannot.

The two work well together and are often confused for each other. Numbering makes a page citable. A checksum makes a file checkable. Neither makes anything provable.