Document Unitization: Where One Scanned Document Ends

Document unitization is the step that turns a run of scanned pages into separate documents by fixing where each one starts and ends. It matters for a scanned production because numbering, indexing, review and every later cite work on documents, and a scanner only knows about stacks of paper.

What is document unitization in a scanned production?

Document unitization in a scanned production is the decision, page by page, about which pages belong together as one document. The Sedona Conference Glossary defines unitization in one line: "The assembly of individually scanned pages into documents." The output is a set of files, or a list of page ranges, with one entry per document.

A sheet-fed scanner saves whatever was in the tray as one file. Feed it a folder holding a two-page letter, a nine-page contract and a single invoice, and you get a twelve-page PDF. Nothing in that file says page 3 begins something new. Unitization writes that fact down, either by splitting the PDF into three files or by recording three ranges: pages 1 to 2, 3 to 11, and 12.

The word describes a decision, and the decision can be wrong in two directions:

• Under-unitized. Several documents sit in one file. A search hit lands somewhere inside a 300-page block and the reader has to work out which letter it came from.
• Over-unitized. One document is cut into pieces. Each page of a ten-page agreement becomes its own record, and the signature page no longer travels with the terms it signs.

Unitization is separate from OCR. Recognition reads the characters on each page. It does not tell you that page 3 is the first page of a contract. A person, or a rule the person set at the scanner, makes that call.

Physical unitization versus logical unitization of a paper file

Physical unitization follows how the paper was fastened. Logical unitization follows what the pages say. The Sedona glossary describes both, and a paper file that has lost its staples leaves you only the second.

In the glossary's words, physical unitization "uses actual objects such as staples, paper clips, and folders to determine pages that belong together as documents for archival and retrieval purposes." Logical unitization is "the process of human review of each individual page in an image collection, using logical cues to determine pages that belong together as documents." The glossary lists the cues: "consecutive page numbering, report titles, similar headers and footers, and other logical indicators."

The two methods can disagree. Suppose a custodian stapled a cover memo to a report he did not write. Physical unitization gives you one document. Logical unitization gives you two, a memo and a report. Neither is wrong on its face. What matters is that one rule is chosen for the whole set, written down, and applied by everyone who touches the boxes. If the parties signed an ESI protocol or the request came with production instructions, read them for a unitization rule before you pick your own.

Write the unitization choice on the production memo before the next box is scanned. If the request for production asks for documents as kept in the ordinary course, keep the groups you found in the file. If you split a cover memo from the report under it, say that in the memo. Each file you send should stand as one item of documentary evidence.

A workable way to combine them: capture the physical boundaries while the paper is still fastened, then have a reader pass over the images and split only where a fastened bundle plainly holds more than one document.

How are attachments and parent documents kept together?

Attachments and parent documents are kept together by unitizing each as its own document and then recording that they belong to one family. In a production the family members are numbered one after another, parent first, and the index or load file carries the range of the whole family.

The Sedona glossary's unitization entry says the process "should also capture document relationships, such as parent and child attachments." Its entry for a document family gives a paper example: "a fax cover, the faxed letter, and an attachment to the letter, the fax cover being the 'Parent,' and the letter and attachment being a 'Child.'"

Take that example with page counts. A fax cover (1 page), a letter (2 pages) and an attached schedule (5 pages) arrive stapled together in the file of the fictional Calloway Marine Supply:

Three rows, one family. A reviewer who opens the schedule can see from the last column that it came with a letter and a cover sheet, and can pull all eight pages.

One agency's instructions show what a recipient may expect. The SEC's Data Delivery Standards state that "All document family groups, i.e. email attachments, embedded files, etc., should be produced together and children files should follow parent files sequentially in the Bates numbering." Your own protocol may say something different, so check it.

Paper families are harder to see than email families because no software recorded the link. The signs are on the page: an "Enc." or "Attachments" line at the foot of a letter, an exhibit tab, a cover sheet that lists what follows. The post on how to create a load file from scanned PDFs shows how boundaries and page counts become rows, and the comparison of types of load files covers which file format records them.

Slip sheets, separator pages and staples as unitization signals

Slip sheets, separator pages and staples are how a boundary survives the trip through the scanner. A staple marks a boundary on the paper. A slip sheet or separator page is a sheet you insert so that the same boundary shows up in the image file after the staple is gone.

Feeding a scanner means removing every fastener, and with it the only record of what was attached to what. Capture the boundary before that happens:

1. Work one folder at a time. Note the box and folder label on a log sheet before you open it.
2. Insert a separator at each fastener. As you pull a staple or clip, put a marked sheet in front of the bundle it held. A sheet printed with the word DOCUMENT in large type is enough.
3. Use a second kind of sheet for attachments. A sheet marked ATTACHMENT in front of an enclosure records that the pages after it belong to the document before it.
4. Mark folders too. A sheet carrying the folder label keeps the source of each bundle with the images.
5. Scan, then split at the separators. Break the file at each marked sheet. Some scanning software can split on a printed separator or a blank page by itself. Check your scanner's manual and test it on one folder first.
6. Decide what happens to the separators. Either delete them before numbering or number them as part of the set. Do the same thing for every one.

Blank sheets make poor separators. A duplex scan of single-sided paper produces blank backs on every page, and a rule that splits on blanks will cut each document into single sheets.

If the paper reached you already scanned, the fasteners are gone and you are doing logical unitization. Read for the cues in the image: a page 1, a new letterhead, a dateline, a signature block followed by a new title.

What goes wrong downstream when unitization is skipped?

When unitization is skipped, every later step inherits documents that are too large or too small. Page counts, number ranges, index rows, privilege calls and search results all take the document as their unit, so a wrong boundary produces a wrong entry in each of them.

The recipient's instructions can make it a compliance point as well. The SEC standards, where they allow PDF production, require that "All PDFs must be unitized at the document level, i.e., each PDF must represent a discrete document."

The repair is expensive in a specific way. Splitting a file after it has been numbered does not change the stamps on the pages, but every list built on the old boundaries has to be rebuilt. Splitting after an index exists means recoding every affected row. Do the split first, on a working copy, and keep the scan as it came off the machine in case a boundary is disputed.

Conversion fits after the split. If you upload each unitized PDF to Schema Law as its own file, each export covers one document, and the export formats include a searchable PDF and plain text. Schema Law reads pages into text and tables. It does not decide where a document ends. Once the boundaries are fixed, the guide to indexing discovery documents covers the fields to record for each one.