Print finishing / Collating, assembling and signing off
Proofing a document on the stock and device it will actually use
The check before a run, which is only worth anything if it is done on the material and the machine the real thing will use.
What is physically going on
Almost every failure in the production clusters is invisible on a screen and obvious on the right piece of paper. A proof is the step where that happens, and it is routinely undermined in the same three ways: it is viewed on a screen rather than printed, it is printed on the wrong stock or the wrong device, or it is printed flat when the finished thing is folded, bound and trimmed. Each of those removes exactly the failures the proof existed to catch. A proof on the real stock, produced the way the run will be produced, and physically finished, is a different artefact from a PDF someone approved in a viewer.
The decisions
Reasons rather than a description of the code. Where a number depends on the stock or the press, the instruction is to get it from the printer rather than to quote one here, because spine width, creep and bulk are not constants.
- Proof on the actual stock. Paper is the white in every light area and it changes tints, detail and show-through.
- Proof on the actual device, or as close as the process allows, since the difference between an office laser and a press is most of this cluster.
- Finish the proof: fold it, bind it, trim it, staple it. The unfinished proof cannot show a gutter problem, creep, or a staple through a logo.
- Proof the worst case, not the typical one: the longest document, the longest name, the largest amount, the most line items.
- Have somebody who did not build it read the proof, because the person who built it knows what it is supposed to say and reads that instead of what is there.
- Record what was approved and against which version, so the thing that goes to the run is the thing that was signed off.
What it means for the page rule
The page rule is the part of this a document author controls. The binder decides the gutter and the printer decides the creep allowance; the document's job is to leave the room those decisions need.
/* Nothing in a stylesheet makes a proof good. What a stylesheet can do
is make the proof reproducible: pin the version and the date into the
document so an approval refers to something specific. */
@page {
@bottom-left {
content: "QMS-PR-014, rev C, issued 14 March 2026";
font-size: 7pt;
color: #666;
}
}
/* A proof marker, applied by a build flag rather than by hand, so it
cannot be left on in the run or forgotten in the proof. */
.proof-mark {
position: fixed;
top: 50%; left: 50%;
transform: translate(-50%, -50%) rotate(-45deg);
z-index: -1;
font-size: 96pt;
font-weight: 700;
color: #dcdcdc;
print-color-adjust: exact;
-webkit-print-color-adjust: exact;
}What people do instead
Approving a PDF in a viewer and calling it a proof. It catches spelling and it catches nothing else in this cluster, and the approval it produces creates the impression that the document was checked when the checks that mattered were not run.
How to find out rather than assume
Hold the finished proof. If it is not on the right paper, not from the right device, and not folded, bound or stapled the way the real one will be, it has not proved the things this cluster is about. Then hand it to somebody who has not read it.
Where this API sits in it
Outside it. This API renders HTML to PDF: it produces a sequence of pages at the size and margins your CSS and the render options set, and everything on this page happens afterwards, to paper, in a bindery. There is no imposition option, no binding option and no finishing option, and there could not be, because none of those are properties of a file. What the render does decide is the two things finishing cares about most: how many pages there are, and how much room is left at the edges. Both of those are set before anything is printed, which is why a finishing constraint discovered late is expensive and one designed for is free.
Frequently asked
Can the API impose the pages into printer spreads?
No, and you should not want it to. Imposition is done by the printer with tooling built for it, against the specific press, stock and folding scheme, and the input that tooling wants is single pages in reading order, which is what a render naturally produces. Supplying anything else means somebody has to undo it first.
How do I get the page count to a multiple of four?
By deciding where the padding goes rather than letting it land at the end. Add the blanks deliberately, style them with the blank page rule so they carry no folio or running header, and mark them as intentionally blank if the document is formal. Counting the covers into the total is the step people miss.
Do any of these apply to a document that is only ever read on screen?
No. This cluster is about physical operations on paper, so if a document is genuinely never printed, none of it is your problem. It is worth checking that assumption rather than making it: delivery notes, permits, method statements and invoices all end up on paper more often than the people generating them expect.
Related finishing topics
The operations that constrain each other, then the rest of the same group.
Corner stapling, and the content the staple goes through
A single staple through the top corner, which is how most office documents are fastened and which occupies a specific area of every page.
Creep in a folded booklet, and the margin that shrinks in the middle
The way inner sheets of a folded booklet push outwards past the outer ones, so trimming the stack flush takes more off the inner pages than the outer.
The inner margin a bound document needs and the outer one it does not
The extra width on the bound edge, which exists because a fastened document does not open flat and the innermost part of the page curves away from the reader.
Blank pages in a double-sided document, and making them deliberate
The empty sides that appear in any double-sided document, which are either chosen by you or produced by accident.
Tab dividers in a printed manual, and where the tabs actually sit
Divider sheets whose tabs extend beyond the edge of the document, cycling down the page so that every tab is visible at once.
Every finishing topic
The full list, grouped by binding, folding and trimming, and collation.
None of this is a render option, and none of it could be. What the render controls is that the page count and the margins are the ones the finishing needs.