Keeping a record of what you checked
By the third proof of your second book, you will not remember whether you already fixed the running head on the part titles. You will think you probably did. You will not be certain, and the two options in front of you are to re-check the entire book or to assume. Assuming is how people stop checking properly, and re-checking everything every time is how they get there.
A proof record is the cheap fix. One page per proof, filled in while the book is on the table in front of you.
One entry per proof
Write it as you inspect, not afterward. Six things go in every entry.
The date the proof arrived. Not the date you ordered it. "Which week was this printed" is the question you will actually be asked if a fault turns out to be the printer's.
The printer and the binding. Amazon KDP, IngramSpark, Lulu, B&N Press — and paperback, hardcover or jacket. The same file produces a different object at each, so a fault on one is not evidence about the others.
Which files were uploaded, and their own dates. The exact file names of the interior PDF and the cover PDF, plus the EPUB if you proofed one. The proof in your hands is the output of two specific files; a record that does not name them cannot answer the only question that matters later, which is whether a fault is still in the current file.
What you found. One line per fault, each with a location. "p. 214, running head reads chapter 11 on a chapter 12 page" is a record. "Some running heads wrong" costs you an hour later.
What you changed. Against each fault, the fix and where you made it — the source manuscript, the cover, or a setting. Corrections made in the wrong place come back: a stray blank page deleted from the PDF and not from the .docx reappears on the next export.
What you decided to accept. The column people skip, and the one that pays for the whole exercise.
The accepted column
You will find things that are not worth fixing. A chapter opening sitting a hair lower than its neighbors. A spine title very slightly off center within the printer's normal drift. A cover that came back darker and flatter than the screen, which is what print does to every cover and not a fault at all — see The proof came back wrong. Now what?.
Write down the decision and the one-line reason. Without it you meet the same thing on the next proof with no memory of having weighed it, and you either re-argue it from scratch or "fix" something that was fine and introduce a real fault doing it. A decision you cannot find is a decision you have to make again.
Photograph faults rather than describing them
"Spine looked off" means nothing in six weeks. A photograph of the spine with a ruler laid along it means exactly one thing, to you and to anyone you show it to.
Shoot in daylight, get the whole cover in one frame and the fault in another, and include something for scale. Name the files with the same date as the entry so they file themselves.
Keep the proofs themselves
Write the date on the inside front cover in pen, and keep the copy. When the corrected proof arrives, put the two side by side.
Comparing a new proof against the last one is the fastest way to confirm a fix landed, and far more reliable than remembering what the old one looked like. It also catches the thing nobody looks for: a fix that worked and moved something else. A page count that changed to correct a widow has changed the spine width too.
In the tools. My Books keeps every project you make and can duplicate one into a new book with its design and settings intact, so the next book starts from the setup your last proof approved rather than from defaults. Start your book → · about the tools
What the log is worth beyond sanity
Evidence when the fault is manufacturing, not design. Binding that fails, ink that smears, a trim visibly outside the printer's own tolerance — these are the printer's to answer for, and they answer better to dated photographs of several copies from one order than to a description. Check the printer's current process for reporting a defective copy when you need it; it differs by printer and it changes.
A checklist tuned to the mistakes you personally make. Read down your own "what you found" column across four or five proofs and your failure modes fall out of it: you always miss the copyright page, or your images always creep into a margin. Use The proof checklist, front to back for coverage and your own list for the things you keep doing.
A head start on the next book. Most of what you fixed on book one is a setting or a habit, not a one-off. The log is where that knowledge is stored instead of evaporating. In a series it does more than that: the entries from volume one are what you measure volume two against, and Proofing a set of books together is the pass that finds the drift between them.
Where it sits with everything else
This is one drawer of the wider filing habit — the ISBN register, the files you actually uploaded, the licenses, the change log — described in The records a self-publisher needs to keep. Keep the proof record in the same per-book folder, named by date, and it is still there when a reader emails about a typo in a printing from two years ago.
Make your book
The tools are free to use: name your book, bring your manuscript or your artwork, and work as long as you like. You only pay when you want the files without the preview watermark.
Next: The proof checklist, front to back · More: Order the proof. Then read it in daylight · Measuring your proof with a ruler · The records a self-publisher needs to keep