Revision review is required; reading everything is not possible
A maintenance or operations manual revision must be reviewed before it is put into use. The list of effective pages tells you which pages changed. It does not tell you what changed on them, and a changed torque value looks exactly like an unchanged one at reading speed.
Revision bars help, when they are there and when they are accurate. They are drawn by the author, and they mark lines the author thinks changed. A reviewer who only reads barred lines inherits every mistake in how the bars were applied.
A hypothetical revision
Consider a 240-page chapter of a maintenance manual, revised to update one inspection task. The list of effective pages shows 14 pages as revised. A comparison against the previous revision returns nine changes. Six are the intended task update: a new step, a changed inspection interval, a re-worded caution. One is a torque value in an unrelated task that changed from 25 to 28 Nm. One is a deleted note about safety wire. One is a changed part number in a figure callout.
The torque value sat on a page that was in the list of effective pages, because the page reflowed, but it had no revision bar. Read quickly, it would have passed. In the change list it is one line, with its page number in both revisions and the old and new text side by side.
Documents of that size need an Enterprise plan, which handles up to 300 pages per document (larger on request). Shorter procedures up to 100 pages work with a free account. Pricing lists the tiers.
What the comparison provides
- A change list with each altered step, limit, caution or warning, with the page in both revisions.
- A redlined PDF for the reviewer and a side-by-side view with synced scrolling for training.
- Repeated headers, footers and page numbers excluded, and a changed revision date in the footer reported once rather than on every page.
- Moved procedure steps reported as moves rather than as a deletion in one place and an insertion in another.
- Deterministic output, so the record of what was reviewed can be reproduced.
Why page-by-page comparison fails on manuals
Manuals are repaginated constantly. Insert one step near the start of a task and every page after it shifts, sometimes for the rest of the chapter. A tool that compares page 40 with page 40 reports the whole chapter as changed. The engine aligns content by paragraph, so an inserted step is one change, and the pages after it are recognised as the same text in a new position.
Tables of limits, clearances and torque values are a similar problem. If they are unchanged they are held out of the diff entirely, so a reviewer is not asked to re-read a table where nothing moved. If one cell changed, that cell is what gets reported.
Temporary revisions
When a temporary revision is incorporated into the permanent manual, compare the new permanent revision against the previous one. The change list should contain the temporary revision content and the planned changes for that revision, and nothing else. Anything outside that set is either a planned change that was not announced or something that slipped in, and either way it needs a look before the revision goes into use.
When something unexpected turns up
Expect the change list to be short. For most revisions it fits on a page or two, which is the point: the reviewer spends their time on the changes, not on finding them. A very long change list for a small revision usually means one of the inputs is wrong, for example an older revision than the one in use, and is worth checking before the review starts.
Do not quietly fix it in your own copy. Record the change with its page reference in both revisions, raise it with whoever issues the manual, and keep the redlined PDF with the query. If the change was intended, the issuer confirms it and the record shows it was reviewed. If it was an error, you caught it before anyone worked to it.
Scanned manuals, which some operators still hold for older types, need an OCR text layer before they can be compared. A pure image scan is refused. With OCR, formatting differences are ranked low so OCR noise does not hide real changes.
Deployment
Operators who cannot send manuals to an external service can run the engine inside their own network, with no outbound traffic. For everyone else the web tool processes files in memory and never stores them. Organisations that receive many revisions can use the API to compare each new revision against the last one automatically. The aviation industry page describes the options, and the general guide on how to compare two PDF files explains the method in more detail.
Try it on your own documents. Upload two versions on the homepage. Documents up to 30 pages need no account, and nothing is stored.
Industry page: PDF comparison for aviation