Bates numbering — stamp pages the way a court expects.

What a Bates number actually is

A Bates number is a fixed-width sequential identifier stamped on every page of a document set, so that any page can be cited unambiguously by anyone holding the same set. It is the reason a lawyer can say “turn to ABC-001482” in a deposition and every person in the room arrives at the same page. The practice is named after a Boston firm that started using a consecutive-numbering stamp in 1892; the machinery is gone, the convention survived.

The format is nearly always a prefix plus a zero-padded counter. The prefix identifies the production — a party, a custodian, a case — and the padding keeps every identifier the same length. ABC-000123 is not decoration: the three leading zeros are what make a plain text sort agree with a page sort.

Numbering a set, step by step

1. Decide the prefix before you start. Once a production has gone out as ABC-, a second batch numbered XYZ- looks like a different production and has to be explained. Pick something that identifies the set, keep it short, and keep it.
2. Set the starting number from the last set, not from one. If the previous production ended at ABC-001206, this one starts at ABC-001207. Continuous numbering across batches is the whole point; restarting at 1 creates two pages with the same identifier and no way to tell them apart in a citation.
3. Choose enough digits for the biggest document you will ever produce. Six digits covers a million pages and is the usual default. Four digits runs out at 9,999 and silently starts producing five-digit identifiers, which breaks sorting exactly when the set is large enough to matter.
4. Put the number in the footer, outside the text block. Bottom right is the convention in most US productions. The number has to sit clear of the page content: a stamp that lands on top of a paragraph has just altered the evidence, and opposing counsel will notice.
5. Number every page, including blanks. The sequence is what proves completeness. A gap between ABC-000418 and ABC-000420 invites a written question about what was removed.

When this comes up

• Document production in litigation. The core use. Every page of every produced document gets a number, and the production log lists ranges per document rather than page counts.
• Contract exhibits and appendices. An agreement with twelve exhibits is easier to assemble and to reference when each exhibit page carries a continuous number rather than restarting at page 1 twelve times.
• Closing binders and audit files. Any set assembled from documents of different origins needs one identifier that survives being photocopied, scanned and re-assembled by someone else.
• Batch scans before review. A scanner produces files in whatever order the feeder took them. Numbering the merged set fixes the order in a way that survives the file being renamed.

A worked example: 3,412 pages, six digits

A production of 3,412 pages from a single custodian, prefix ABC-, six digits, starting at 1. The first page becomes ABC-000001, the last becomes ABC-003412. Every identifier is nine characters, so a text sort, a spreadsheet sort and a page sort all produce the same order — which is the only reason to pad at all.

Now the same set with four digits. Page 9,999 is ABC-9999 and page 10,000 is ABC-10000. In a plain sort, ABC-10000 lands before ABC-9999, and the production log stops matching the pages. Nothing is corrupt; the identifiers simply stopped being comparable. Repadding means renumbering the whole set and reissuing it, which is why the digit count is worth deciding once, generously.

The starting number deserves the same care. A second production that restarts at 1 has two pages numbered ABC-000001. A citation to that identifier is now ambiguous, and the fix is a written correction to everyone who received the first batch.

Where numbering goes wrong

• Restarting at 1 for each new batch. Continuous numbering across productions is the convention; restarting destroys the only property the identifier had.
• Skipping blank pages. Leaves gaps that have to be explained, and makes the range in the production log wrong.
• Stamping over the text. A number that overlaps a paragraph has altered the document. Footer, outside the margin, always.
• Numbering the original. Always number a copy. The set you produce is evidence; the file you received from the client is the record of what you received.
• Forgetting the OCR step. A number drawn on a pure image scan is not searchable, and neither is the rest of the page. If the set needs to be searchable, OCR first, then number.

Questions