Bates numbering software for a law firm that is not an e-discovery department

One continuous sequence, and the arithmetic that goes with it

The sequence runs across the whole set: the last page of the first document and the first page of the second are consecutive. Prefix, suffix, digit padding, start value and step are all set before the run, and the padding is what keeps the numbers sortable as text — ABC-000123 sorts before ABC-001230, which is how production software compares them more often than as integers.

Two things a real production needs and a generic stamper does not have: you can reserve a round thousand at each document boundary, so a document that grows later still has room before the next one starts; and you can declare the ranges already used in earlier productions, so the run refuses to hand out a number you have already given the other side.

Placement measured page by page, not once for the batch

A real set is not uniform. Some files are born-digital Letter pages, some are scans whose crop box is much smaller than the media box, and some pages carry a rotation of 90 or 270 from the scanner feed. The number is placed against each page’s own crop box and its own rotation value, so it lands where the reader looks rather than at a coordinate that happens to be the lower right of an unrotated page. Mixed Letter, Legal, A4 and oversized drawing pages in one run are handled individually.

A number that would sit within a quarter inch of the trim edge raises a print-safety warning before the run, because that is the one the copy shop cuts off.

What proves the run to somebody who was not watching

With verification on, every stamped file is re-opened after it is written and the expected number is looked for on the pages that should carry it. A file that will not re-open is reported as a failure and never delivered. Alongside the PDFs you get a manifest CSV — source name, output name, page count, BegBates, EndBates and a SHA-256 of the bytes written — and, optionally, a Summation-style .dat load file and an Opticon .opt image load file. The .opt is refused rather than written wrong if any document is missing a per-page row, because a loader trusts the .opt over the files.

The SHA-256 column is the one that matters months later: it is how you show that the file in the case folder is the file you produced.

The three checks that belong in the same pass

Numbering is the visible half of getting a production out. The other half is what is still inside the files: a metadata pass finds the author name, the original file path and the revision history sitting in the PDF, and a redaction pass finds black bars with extractable text underneath. Both are the ordinary way confidential material leaks out of a document that was stamped perfectly. Both run read-only, on the same machine, over the same queue.

Why it can be this cheap

There is no upload endpoint. The PDFs are opened, modified and written by your own browser using pdf-lib, so there is no storage bill, no bandwidth bill and no per-gigabyte hosting to pass on to you — which is why forty documents cost the same as one and why the price is a single $19 rather than a seat. You can verify the claim rather than trust it: open the network panel and run a batch, or disconnect the network entirely and watch it finish.

The trade-off is honest: the ceiling is your own machine’s memory, and there is no shared workspace, no review, no productions history across a team. If you need those, you need a platform.