What is the single most common mistake in a bulk run?

Sizing the mark in pixels. It is invisible on one file and enormous on the next, and because nobody looks at three hundred outputs, it is usually discovered by a customer. Percentage sizing is not a convenience here — it is the thing that makes the run repeatable.

The mark looks right on file one but wrong on file forty. Why?

Almost always because file forty has a different subject under the same corner. A white mark in the bottom right is perfect over a dark table edge and gone over a white backdrop. An outline or a shadow fixes it for both; a cleverer colour fixes it for neither.

Should I run everything at once or in batches?

At once, unless the tab struggles. Files are worked a few at a time and released as they are written, so a long queue costs time rather than memory. If a run does stall, halve it — two archives extract into the same folder just as happily as one.

Can I mark a folder that already contains marked files?

You can, and you will get two marks on those files. There is nothing in a file that says whether it has been marked, so nothing here can tell. Keeping marked output in its own folder — which the naming template makes easy — is the whole answer.

Does the run change the original files?

No. Every original is opened read-only and never written to. What comes back is a separate archive of new files, and if you dislike the result you delete the archive and nothing has happened.

Marking hundreds of pictures without checking hundreds of pictures

The hard part of a bulk run is not the running. It is choosing a setting that is still correct on files you will never look at — which comes down to percentages, queue order, naming, and five minutes spent on file one.

The part that is actually hard

Running a mark over four hundred files is a loop, and loops are easy. What is hard is that you will look at one of those four hundred pictures and then agree to whatever happens to the other three hundred and ninety-nine. Every real problem in a bulk run is a version of that: a decision made against one file that turns out not to have been about that file at all.

So the useful skill is knowing which settings generalise and which only looked like they did. That list is short and it is the same every time.

What carries across a mixed folder
Carries	Mark width as a percentage of each picture's width.
Carries	Inset as a percentage of each picture's shorter edge.
Carries	Repeating gaps, because they are multiples of the mark's own size.
Carries	Rotation, opacity, blend mode, output format, naming.
Does not	Inset in pixels, once the folder holds more than one picture size.
Does not	A colour chosen against one background, when the backgrounds differ.
Does not	A corner choice, when that corner is empty in some pictures and not others.

Why percentage sizing is the whole trick

Take a folder with a 1200-pixel web export and a 6000-pixel original in it, which is what almost every real folder looks like. Ask for a mark forty pixels wide. On the web export it occupies three and a half per cent of the width; on the original it is two-thirds of one per cent — a line of text you would need to zoom in to find. Neither is what you chose. There is no pixel figure that is right for both, and the folder is not going to become uniform to help you.

Ask instead for a mark that is twenty-four per cent of the width and both files get a mark occupying a quarter of their frame. That is the same decision applied twice rather than one decision applied to two different problems. The same reasoning runs through the inset — a percentage of the shorter edge keeps a corner mark the same distance in, proportionally, on a square and on a panorama.

Seven steps, in order

Put the whole folder in at once. Queue order is the order a file manager would show, with digits read as numbers, so photo-9 lands before photo-10 rather than after photo-100.
Set the mark against the first picture. It is decoded and shown for exactly this reason. Judging a mark against a real file from the folder is the only check that transfers to the rest of it.
Leave every size as a percentage. Mark width against the picture's width, inset against its shorter edge. Switching the inset to pixels is occasionally right and is the one setting that will not survive a mixed-size folder.
Zoom the browser out to the size the pictures will be seen at. Most marks are chosen at full size and viewed at feed size. Checking at the size that matters takes ten seconds and changes the answer often.
Pick the naming template before you run. {name}-wm keeps the original stem and makes the marked copy obvious in a folder listing. Keeping the original name is supported and is a good way to lose track of which is which.
Save the setting under a name. It is kept in this browser. The next folder starts where this one finished, which is the entire return on the five minutes you just spent.
Run it, then read the list before you unzip. The completion line counts what was written and what was left out; the lines under it say why. Four skipped files with reasons is information. Four missing files is a mystery.

Checking file one properly

Check it at the size it will be seen

A mark is chosen at a hundred per cent on a large screen and looked at on a phone inside a feed, where the picture is maybe four hundred pixels wide. Zoom the browser out until the picture on screen is roughly that wide and read the mark again. If it has become a grey smear, it needs more size or more opacity, and it is much cheaper to learn that now.

Check it against the worst background in the folder

Not the first file — the worst one. Drag in the picture with the palest area where your mark sits, or the busiest. If the mark holds there it will hold everywhere, and if it does not, an outline at three or four per cent of the text height is usually the cheapest fix.

Check the output format against what you will do next

Keeping the original format means a folder of PNG screenshots stays PNG and a folder of camera JPGs stays JPG, which is almost always what you want. Forcing JPG on a folder that contains cut-outs with see-through backgrounds will fill every transparent area with white, and that is a decision worth making on purpose rather than discovering in the archive.

What the list is telling you

Every file gets a line. A marked file shows what it was called, what it is called now, and how large it is. A file that was left out shows the reason in the same sentence as what to do about it — a picture past the megapixel ceiling, a file over the size limit, or something the browser could not read as a picture at all.

Lines appear in queue order, not in the order files finished, because the archive is built in queue order too. A gap in the sequence means a file is still being written, not that one has gone missing.
A file that was worked on at a reduced size says so on its own line. Nothing is quietly downsized.
An animation says that the mark went on frame one and the rest of the frames did not come through. A CMYK JPEG says it was converted to screen colour. A multi-page scan says which page survived.
The completion line counts written and skipped separately. If the second number is not zero, the reasons are directly beneath it.

What is the single most common mistake in a bulk run?: Sizing the mark in pixels. It is invisible on one file and enormous on the next, and because nobody looks at three hundred outputs, it is usually discovered by a customer. Percentage sizing is not a convenience here — it is the thing that makes the run repeatable.
The mark looks right on file one but wrong on file forty. Why?: Almost always because file forty has a different subject under the same corner. A white mark in the bottom right is perfect over a dark table edge and gone over a white backdrop. An outline or a shadow fixes it for both; a cleverer colour fixes it for neither.
Should I run everything at once or in batches?: At once, unless the tab struggles. Files are worked a few at a time and released as they are written, so a long queue costs time rather than memory. If a run does stall, halve it — two archives extract into the same folder just as happily as one.
Can I mark a folder that already contains marked files?: You can, and you will get two marks on those files. There is nothing in a file that says whether it has been marked, so nothing here can tell. Keeping marked output in its own folder — which the naming template makes easy — is the whole answer.
Does the run change the original files?: No. Every original is opened read-only and never written to. What comes back is a separate archive of new files, and if you dislike the result you delete the archive and nothing has happened.

Back to the tool

Marks are applied here. Nothing is sent anywhere.